There is a version of the Blockbuster story that most people know. A company that dominated home video rental for two decades, failed to see what Netflix represented, and declined steadily until there was nothing left. The story is tidy. It fits a familiar pattern — the incumbent that couldn’t adapt, the disruptor that moved faster, the market that rendered a business model obsolete. It is also, in its most important detail, wrong.

Blockbuster saw Netflix. It understood the threat specifically and early. By late 2006 it had built a direct competitive response — a hybrid rental model that combined online convenience with physical store access — that caused Netflix’s subscriber growth to measurably slow. Netflix acknowledged the threat in its own filings. The program was working.

It was pulled apart before it could finish the job. Not because leadership changed its mind about the strategy. Not because customers stopped responding. Because the capital required to sustain it had already been committed elsewhere — to a debt obligation placed on Blockbuster’s balance sheet at the 2004 Viacom spinoff, before the competitive battle had fully begun.

What follows is not a story about a company that failed to adapt. It is a story about a company that adapted, found a response that worked, and ran out of the financial runway required to see it through.

The System

Blockbuster’s model at its peak was built on a simple competitive logic: physical scarcity, resolved at scale. A customer who wanted a specific movie on a Friday night in 1995 had limited options. Blockbuster had spent a decade building the infrastructure to be the reliable answer to that problem — wide store coverage, computerized inventory tracking, broad title selection, and a family-oriented environment that mass-market consumers trusted.

By the end of 2004, the company operated approximately 9,100 stores across the United States and 24 other countries. Total revenues for 2004 were $6.1 billion. The domestic rental market represented roughly $8 billion in annual consumer spending that year. Blockbuster’s share of that market was substantial and defensible — or appeared to be.

The model had a structural feature that would later become its most visible liability. Late fees — extended viewing fees in the company’s formal language — were not an incidental revenue line. They were embedded in the economic logic of a system built around high-demand titles with limited copy depth. When a popular new release sat unreturned past its due date, it was unavailable to the next customer. The fee was the mechanism that cleared inventory back into circulation. The revenue it generated — estimated at $400 to $450 million annually in the company’s own disclosures — funded a material portion of the operating income that supported the store network’s cost base.

The model required one condition to function as designed: that physical presence remained the primary mechanism through which consumers accessed home video. As long as that condition held, the store network was a genuine competitive asset. Title availability, store proximity, and the browsing experience were things a mail-order service could not replicate. The density of the network was the business.

That condition had held for nearly twenty years. By 2004, it was no longer guaranteed.

The Shift

Netflix launched its DVD-by-mail subscription service in 1998. The founding narrative of the company — that its CEO Reed Hastings had been charged a late fee on an overdue rental and decided to build an alternative — was pointed directly at Blockbuster’s most resented feature. The model Netflix offered was structurally different from anything in the physical rental market: a monthly subscription, no due dates, no late fees, a catalog far larger than any store could stock, delivered by mail.

What Netflix demonstrated over the following years was not simply that consumers would use a mail subscription service. It demonstrated something more significant: that a meaningful segment of consumers would accept a delay in exchange for eliminating the friction that late fees, store trips, and limited selection created. Physical presence was not a prerequisite for the rental transaction. Convenience could be redefined.

By 2004 this was no longer a hypothesis. Netflix had over one million subscribers. The domestic rental market was declining — from approximately $8.2 billion in 2003 to an estimated $8.0 billion in 2004, with projections showing continued contraction through the decade. Blockbuster’s own 2004 annual report acknowledged that its core rental business had continued to decline during the year.

The condition the Blockbuster model required — that physical access remained primary — was no longer guaranteed. The shift had already happened. The question was what Blockbuster could do about it, and what it had available to do it with.

The answer to the second question had been determined months earlier, in a transaction Blockbuster did not choose.

The Response

In October 2004, Viacom completed the spinoff of Blockbuster as an independent publicly traded company. Viacom had owned a controlling stake in Blockbuster since 1994, acquiring it primarily to help finance the Paramount Pictures acquisition. For a decade, Blockbuster had operated as a subsidiary within a larger media conglomerate, its capital allocation subordinated to parent company priorities.

The spinoff structure placed approximately $1 billion in senior subordinated notes — high-interest debt instruments due in 2012 — on Blockbuster’s balance sheet at separation. This was the capital structure within which every subsequent decision would have to operate.

Blockbuster’s response to the Netflix threat unfolded in two moves, both of which were rational given what management understood at the time.

The first was the elimination of late fees. Effective January 1, 2005, Blockbuster ended extended viewing fees at all company-operated domestic stores. Chairman and CEO John Antioco described it as the biggest customer benefit in the company’s history. The competitive logic was sound: late fee resentment was the most documented driver of customer attrition to Netflix. Removing it addressed the most visible competitive disadvantage directly.

The financial consequence was immediate. In the company’s own words, extended viewing fees had contributed $400 to $450 million in revenues and $250 to $300 million in operating income the prior year. That income was gone. The cost base built around its presence was not.

The second move came in late 2006. Blockbuster Total Access launched as a hybrid rental model that allowed online subscribers to return DVDs by mail or exchange them at a participating Blockbuster store for free in-store rentals. It was a direct competitive response to Netflix built around the one structural advantage Netflix could not replicate: a physical store network already in place. The program combined Netflix’s convenience with a capability Netflix structurally lacked.

The competitive effect was real and measurable. By the end of 2006, Blockbuster had added approximately 700,000 online subscribers in a single quarter. By Q1 2007, Antioco reported that Total Access had exceeded 3 million subscribers and that the first quarter of 2007 had been the company’s highest subscriber growth quarter ever. Netflix noticed. In its own FY2007 annual filing, Netflix stated directly that its slowing growth appeared primarily to be the result of the rapid growth of Blockbuster Online.

Total Access was working. And it was expensive.

Each in-store exchange — the feature that differentiated Total Access from Netflix — cost Blockbuster more than the subscription revenue it generated at competitive pricing. The subsidy was the mechanism of competitive traction. The program’s success at attracting subscribers made the aggregate subsidy cost larger with each passing quarter. By Q2 2007, with the subscriber base at 3.6 million, newly appointed chairman and CEO Jim Keyes stated directly that “the costs associated with the program affected our profitability” and that modifications were planned before year end.

The modifications were rational from a near-term financial perspective. The debt service obligation was fixed. The operating losses were widening. The board pressure — from Carl Icahn, who had acquired a significant stake and pushed consistently for cost reduction over competitive investment — was real. Reducing the per-exchange subsidy and pulling back marketing spend on Total Access improved the near-term P&L. It also removed the program’s competitive effectiveness at the moment Netflix was most vulnerable to it.

Netflix’s own filing acknowledged what followed: Blockbuster’s pricing changes contributed to an acceleration in Netflix’s subscriber growth resuming.

The strategic response was not a failure of vision. It was a response that worked until the capital structure made sustaining it impossible.

The Constraint Layer

Five constraints were compounding simultaneously by 2007. None was solvable without addressing the others. Together they formed a set that closed off the available options faster than any single response could open them.

The debt service obligation. The $1 billion in senior subordinated notes issued at the 2004 Viacom spinoff carried a fixed annual cash obligation regardless of operating performance. In a business generating sufficient free cash flow, this would have been manageable. Blockbuster was simultaneously absorbing the late fee revenue loss, investing in Total Access subscriber acquisition, and closing underperforming stores — all of which consumed cash. The debt service was senior to all of it. Every dollar committed to interest payments was a dollar unavailable for competitive investment.

The late fee revenue gap. The $250 to $300 million in operating income that late fees had contributed was gone. The cost base — store leases, labor, title acquisition — had been calibrated to its presence. Eliminating late fees was competitively necessary. Absorbing the loss without a corresponding reduction in costs created a structural gap that the balance sheet had to bridge at the same time it was funding debt service and Total Access investment. The three draws on available capital were simultaneous and non-negotiable.

The store lease portfolio. At peak, Blockbuster operated more than 9,000 locations globally. Lease obligations are multi-year commitments that cannot be exited without cost. As revenue declined and store traffic shifted to online channels, locations that had been marginally profitable became loss-generating — but the lease obligations continued on their original terms. Store closure programs reduced the count over time, but the pace of closure required to materially reduce the fixed cost base would have triggered restructuring costs the balance sheet could not absorb.

The Total Access subsidy constraint. The program that was winning subscribers was doing so at a per-unit cost that required capital to sustain. The board intervention that restructured Total Access pricing in the second half of 2007 reduced the subsidy and with it the program’s competitive draw. By Q3 2007 the subscriber base had declined to 3.1 million from the Q2 peak of 3.6 million. The company announced it would no longer report online subscriber counts going forward, framing the shift as a broader focus on total membership rather than the online channel specifically. Netflix’s growth resumed.

The franchise network misalignment. Blockbuster’s domestic store base was not entirely company-operated. As of late 2005, approximately 1,079 of the 5,696 domestic stores operating under the Blockbuster brand were franchise locations. The no-late-fee program, when implemented at the start of 2005, was adopted at all company-operated stores. Approximately 475 franchise stores were participating as of Q3 2005 — fewer than half. The franchise stores that did not participate were operating under a different economic model, one in which late fees remained a revenue component their unit economics depended on. Whether those franchisees could have sustained a full transition to the Total Access model — with its in-store exchange costs layered on top of the late fee elimination — is not directly documented. What the participation data shows is that the store network Blockbuster’s hybrid model depended on was not fully available for that model’s execution. The competitive differentiator that made Total Access superior to Netflix was a physical network that, in practice, was operating under at least two different economic arrangements simultaneously.

A third competitor had also entered during this period. Redbox had begun placing low-cost DVD rental kiosks at grocery stores and fast food locations, introducing price-based competition that the in-store rental model had not previously faced. This added a second front to the competitive pressure at the same time the balance sheet had no capacity to respond to the first.

The five constraints did not operate independently. The debt service consumed capital the Total Access subsidy required. The late fee gap widened the operating losses the debt service was already straining. The lease obligations created a fixed cost floor that could not be reduced quickly enough as revenue declined. The franchise misalignment created execution gaps in the network the hybrid model needed. Each constraint made the others harder to address. The window in which any single intervention might have been sufficient was already closing before the competitive response was fully deployed.

The Compression

By 2008 the option set had become structurally binding. The actions required to escape the constraints each required resources the constraints themselves had already consumed.

Sustaining Total Access at competitive pricing required ongoing per-exchange subsidy capital. The debt service had consumed it. Reducing the fixed cost base required lease termination payments and restructuring costs. The operating losses had consumed the liquidity those payments required. Investing in streaming — the next competitive arena Netflix was already positioning for, having introduced its instant-watching feature in January 2007 — required technology infrastructure and content licensing capital. Neither was available. Refinancing the debt on terms that would restore operational flexibility required a credible growth story. The competitive position, following the Total Access restructuring, no longer supported one.

Creditor negotiations began in 2009. The bankruptcy petition was filed on September 23, 2010, listing assets of $1.02 billion against debt of $1.46 billion. Dish Network acquired Blockbuster out of the bankruptcy proceedings in April 2011 for approximately $320 million. Remaining corporate stores closed through 2013 and 2014.

The compression point — the moment at which the available option set became unresolvable — was not 2010. It was 2007, when the board restructured Total Access and the balance sheet lost the capacity to restore it. The bankruptcy was the administrative resolution of a structural position that had become unresolvable two years before the filing.

It is worth holding this sequence precisely. Total Access was launched in late 2006. It reached 3.6 million subscribers by mid-2007. Netflix’s own filings acknowledged that Blockbuster’s competitive response was slowing its growth. The program was restructured in the second half of 2007 under board pressure driven by near-term profitability concerns. Netflix’s growth resumed. Blockbuster filed for bankruptcy in 2010.

The sequence matters because it locates where the story actually ends. Not at the filing. Not at the Netflix launch. At the moment in 2007 when the one program generating competitive traction was pulled back, and the balance sheet had already lost the capacity to restore it.

What Blockbuster Left Behind

Three framings of this story circulate widely and each of them is wrong in a specific way.

The first is that Blockbuster failed to see Netflix coming. The record does not support this. Blockbuster launched its own online subscription service in 2004. It eliminated late fees in 2005 in direct response to the competitive disadvantage Netflix had built its brand around. It launched Total Access in 2006 as a hybrid response designed around the one structural advantage Netflix could not replicate. The company understood the threat. It built a response. The response was working.

The second is that Blockbuster was too slow to adapt. The timeline runs in the other direction. Total Access launched less than a decade after Netflix. It reached competitive parity — and by Netflix’s own account, competitive advantage — within months. The constraint was not pace. It was capital.

The third is that this was a story about technological disruption. Streaming did not end Blockbuster. Netflix’s instant-watching feature launched in January 2007 with a catalog of around 6,000 titles and was available only on personal computers. What ended Blockbuster’s ability to compete was a balance sheet obligation incurred in 2004 that consumed the capital every subsequent competitive response required.

What the market lost when Blockbuster exited was not simply a retail category. A viable hybrid rental model — one that combined the convenience of mail subscription with the immediacy of physical exchange — was extinguished before it could demonstrate whether it could have sustained competitive pressure on Netflix over the longer term. That question was not answered by the market. It was foreclosed by a capital structure.

The store in Bend, Oregon that remains open today operates as it always has — late fees and all. It is, among other things, a reminder that the model itself was not the problem. The problem was what it cost to transform it, and what had already been spent before the transformation began.

Deliberate Drift analyzes how structural positions accumulate over time — decisions that appear rational while constraints build beneath the surface, or advantages that appear ordinary while they compound into something durable.

Blockbuster saw Netflix. It built a direct response — Total Access, a hybrid rental model that combined online convenience with physical store exchange. Netflix acknowledged in its own SEC filing that Blockbuster's response was slowing their growth.

Total Access was pulled apart before it could finish the job. Not because the strategy failed. Because a billion dollars in debt, placed on Blockbuster's balance sheet at the 2004 Viacom spinoff, made sustaining it structurally impossible.

This episode works through the five constraints that compounded simultaneously — and locates the actual compression point, which wasn't 2010.

Subscribe for new episodes on a biweekly schedule at deliberatedrift.com.

Takeaways

• Blockbuster's failure is often misattributed to its inability to adapt to Netflix.

• The company actually had a working response to Netflix before its financial issues escalated.

• Blockbuster's debt from the 2004 spinoff significantly impacted its ability to compete effectively.

• Late fees were a crucial part of Blockbuster's revenue model, which they later eliminated.

• Total Access was Blockbuster's hybrid model that briefly succeeded against Netflix's growth.

• Franchise participation in Blockbuster's initiatives was inconsistent, weakening their competitive advantage.

In 1983, a warehouse opened in Seattle and charged people money to shop there. The fee was $25 a year. The idea was not new — Sol Price had been doing it in San Diego since 1976 — and the logic was straightforward: charge members upfront, use the revenue to subsidize lower merchandise prices, give members a reason to come back. It was a funding mechanism. A way to make the math work.

Forty years later, Costco Wholesale Corporation collects approximately $4.8 billion in membership fees annually. Its merchandise operations — the warehouses, the pallets, the $1.50 hot dogs, the Kirkland olive oil — produce almost no profit on their own. The fees are not subsidizing the merchandise. The merchandise is justifying the fees.

That inversion did not happen by accident, and it did not happen quickly. It accumulated through a series of decisions that looked, at the time, like operational discipline. Thin margins. Limited product selection. Unusually high wages for retail workers. No advertising. Each of these looked like a constraint the company had accepted, or a risk it was running. What they were actually doing, in combination and over time, was building an asset that competitors could not replicate — not because the structure was secret, but because replication required accepting years of sub-market returns while waiting for something that only time could produce.

The System

The warehouse club model that Costco inherited from Price Club rested on a specific exchange. Members paid an annual fee. In return, they received access to merchandise priced at margins thin enough that the savings, over a year of shopping, would more than cover the cost of membership. The fee was the entry price. The value proposition was the spread between Costco’s prices and what the member would otherwise pay.

This required Costco to keep merchandise gross margins deliberately low — in the range of 10 to 14 percent, compared to 25 to 30 percent for conventional grocery and considerably more for department retail. The 10-K filings from the early years document this consistently. Costco was not failing to extract margin from its suppliers; it was choosing not to. The margin that a conventional retailer would have kept was being returned to members in the form of lower prices.

The model required one condition to function: members had to believe the fee was worth paying. As long as that belief held, the fee revenue was stable, and stable fee revenue made the thin merchandise margins viable. The whole structure rested on member satisfaction — not as a nice-to-have, but as a load-bearing element.

What made this more than a clever pricing strategy was what happened when members renewed. A member who paid $25 in year one and returned in year two had already decided the exchange was worth it. The second renewal required less persuasion than the first. The third less than the second. Each renewal cycle, the decision to pay shifted incrementally from an evaluation to a habit. The fee became a standing commitment rather than an annual deliberation.

This dynamic was present from the beginning. It was not yet understood as the structural core of the business.

The Shift

The Price Club merger in October 1993 created PriceCostco, a combined entity with more than 200 warehouse locations across the United States, Canada, and several international markets. Jim Sinegal became chief executive of the combined company. The Price family, who had built the original warehouse club concept, exited management within a year. By 1997, the PriceCostco name was retired and all locations rebranded as Costco.

The merger is significant not because of its scale — though the combined buying power mattered — but because of what it required Sinegal to defend. PriceCostco was publicly traded from its formation. Quarterly earnings pressure was now structural. Analysts covering the company could see that Costco was sitting on merchandising scale it was not monetizing. The question, asked consistently through the mid-1990s and into the 2000s, was straightforward: why not expand margins?

Sinegal’s answer was always the same. Expanding merchandise margins would damage the value proposition members were paying to access. If the prices rose toward conventional retail levels, the fee stopped making sense. Members who stopped believing the fee was worth it would not renew. And if they did not renew, the fee revenue that made the thin merchandise margins viable would erode. The merchandise margin cap — held below 15 percent, and in most years well below — was not operational conservatism. It was the mechanism by which the membership asset was being built.

The discipline held. And it outlived him. On a Q3 FY2024 earnings call, decades after the original margin defense was first articulated, analyst Chuck Grom of Gordon Haskett put the question to current CEO Ron Vachris directly: the 14 to 15 percent margin ceiling had been in place for years — did Costco intend to maintain it? Vachris answered without equivocation. That ceiling, he said, had been part of how Costco measured return for many years. There were no plans to move it.

The question had been asked, in various forms, for three decades. The answer had not changed. What was shifting during this period, largely beneath the surface of the financial reporting, was the quality of the membership base. Each year that a member renewed was a year in which the renewal decision moved further from a price comparison and closer to an automatic commitment. The base was not just growing. It was maturing.

The Response

Competitors watched Costco’s growth and drew the obvious conclusion: the warehouse club format worked, and scale was the route to competing in it. Sam’s Club, operating within Walmart’s structure, expanded aggressively through the 1990s and 2000s. BJ’s Wholesale Club established a regional presence in the Northeast. Both followed the same basic template — large-format warehouse, membership fee, limited SKU count, bulk merchandise.

Neither closed the gap with Costco on the metric that mattered most.

Sam’s Club’s membership renewal rates have consistently trailed Costco’s by a significant and persistent margin — a gap documented in analyst and industry coverage across multiple decades. The structural reason is not difficult to identify. Sam’s Club operates within Walmart, which means its financial performance is reported as part of a consolidated enterprise with its own earnings expectations. Holding merchandise margins at Costco’s floor — accepting near-zero merchandise profit in exchange for membership asset accumulation — is a commitment that a standalone operator can sustain more easily than a division of a publicly traded conglomerate whose shareholders are measuring consolidated returns. Sam’s Club could replicate Costco’s format. It could not fully replicate Costco’s incentive structure.

BJ’s faced a different version of the same problem. A regional operator building a membership base from scratch in the 2000s was not just competing against Costco’s current value proposition. It was competing against the accumulated renewal history of a membership base that had been renewing for twenty years. A new member at BJ’s was making a first-year decision. A member renewing at Costco for the fifteenth time was not.

The responses confirmed rather than disrupted Costco’s position. Format replication without operational replication produced a weaker version of the model. And operational replication without temporal replication — without the years of renewal history that produced habituated behavior — produced a membership base that had not yet developed the inertia that Costco’s had.

The Accumulation

Three reinforcing mechanisms were operating simultaneously through the 2000s and into the 2010s, each compounding the others.

The first was the renewal flywheel. Thin merchandise margins delivered genuine value, which produced high renewal rates, which produced stable fee revenue, which funded continued investment in buying power and employee quality, which sustained the value that produced high renewal rates. Each cycle of this loop did not just maintain the position — it deepened it. Members who had renewed multiple times were not evaluating Costco against alternatives each year. The evaluation had already occurred, repeatedly, and the answer had been the same each time. The renewal was becoming automatic.

The second was the stock-keeping unit (SKU) trust loop. Costco’s warehouses carry fewer than 4,000 active SKUs. A conventional supermarket carries tens of thousands. The constraint is also a service: because Costco selects so few products, the selection itself is a curation signal. A member who trusts Costco’s judgment about which olive oil or which mattress to carry does not need to comparison shop. That trust accumulates with each satisfactory purchase. The Kirkland Signature private label, expanding steadily through this period, deepened the loop further. A household that relies on Kirkland for a meaningful portion of its regular purchasing has built a dependency that does not dissolve when the membership fee increases by five dollars.

The third was the employee quality loop. Costco has paid above-market wages for retail workers since its founding — a policy that drew consistent analyst criticism on cost grounds and was defended by management on retention grounds. The Q4 FY2024 earnings call documents average US and Canada wages just above $30 an hour. The structural consequence of above-market compensation is lower turnover. Lower turnover produces more experienced warehouse staff. More experienced staff produces more consistent member experience. Consistent member experience is a component of the renewal value proposition. This loop operates on a longer cycle than the others, but its output — the institutional knowledge and service consistency embedded in a stable workforce — is the hardest element of the model for a competitor to replicate, because it requires sustained above-market labor investment for years before the benefit compounds into measurable renewal behavior.

By fiscal year 2013, the structural consequence of these three loops running simultaneously for two decades was visible in Costco’s income statement. Membership fee revenue came in at approximately $2.3 billion. Operating income came in at approximately $2.8 billion. The merchandise operations — the warehouses, the inventory, the supply chain, the employees — produced operating profit of roughly $500 million on net sales exceeding $100 billion. The fee revenue was not supplementing the merchandise profit. It was the profit. The merchandise existed to justify the renewal.

The Lock-In

The temporal barrier is the one that cannot be purchased. A competitor entering the warehouse club market in 2013 could raise capital, build warehouses, hire buyers, and price merchandise at thin margins. What it could not do was compress the years required for a membership base to mature from deliberate evaluation into habitual renewal. That compression is not available at any price.

The other requirements follow from it: establishing a sufficient initial membership base to fund operations at thin merchandise margins; delivering enough value, consistently enough, over enough renewal cycles to begin producing habituated behavior; sustaining above-market labor costs before the retention benefit appears in the service quality data; building a private label brand from zero against a Kirkland Signature that had been a household staple in millions of homes for fifteen or twenty years. None of these is individually impossible. In combination, over the time horizon required, they constitute a commitment that no publicly traded competitor has demonstrated the organizational capacity to sustain. The barrier is not capital. It is time — and the organizational discipline required to accept below-market returns while time does its work.

The membership asset is not a list of names and credit card numbers. It is a behavioral state — the accumulated habit of renewal across millions of households. A member renewing for the first time is making a deliberate cost-benefit evaluation. A member renewing for the twelfth time is largely not. The fee has been absorbed into the household budget. The shopping pattern has been established. The switching cost is not the annual membership fee — it is the disruption of a deeply embedded routine.

Sam’s Club, which opened its first location in 1983, has operated the same warehouse club format for as long as Costco. In fiscal year 2024, Costco’s US and Canada membership renewal rate was 92.9 percent. The gap between the two companies is not a gap in execution. It is a gap in accumulated time, organizational incentive alignment, and the depth of habituated renewal behavior those two things produce.

The Compression

The fee increase in September 2024 — from $60 to $65 for Gold Star membership and from $120 to $130 for Executive membership, the first increase in seven years — was the most direct test of the structural position Costco had built. A fee increase is an explicit request for members to reaffirm the value proposition. A membership base that is renewing out of habit rather than current-period evaluation absorbs the increase without significant attrition. A membership base that is still making annual price comparisons does not.

The Q4 FY2024 earnings call documents the result. US and Canada renewal rate: 92.9 percent. Worldwide: 90.5 percent. CFO Gary Millerchip, responding to analyst questions about member attrition, stated that the company had not seen a significant member reaction and that renewal rates remained stable.

The minor decline in the US renewal rate — down 0.1 percentage points from the prior quarter — was attributed specifically to a 2023 digital promotion that had generated more than 200,000 new sign-ups. Digital promotions attract a younger, less habituated cohort that renews at a lower initial rate. As that cohort entered the renewal calculation, it dragged the aggregate rate down marginally. The underlying renewal behavior of the established base was unchanged.

The fee increase is not just a revenue event. It is a measurement. Each time Costco raises the fee and absorption is confirmed, the structural durability of the membership asset is documented. The 2006 increase was absorbed. The 2011 increase was absorbed. The 2017 increase was absorbed. The 2024 increase was absorbed. The pattern is not a coincidence. It is the accumulated result of two decades of delivered value compounding into behavioral lock-in.

Costco entered fiscal year 2025 with 76.2 million paid household members. At the current fee levels, the annual fee revenue run rate exceeds $4 billion. That revenue is largely independent of merchandise sales fluctuations, largely immune to the margin compression dynamics that govern conventional retail competition, and largely secured by behavioral inertia that has been compounding for forty years.

What Costco Built

Three framings of Costco’s position are common, and all three miss the structural point.

The first is that Costco succeeds because of scale. Scale matters — purchasing leverage is real, and operating a high-volume warehouse efficiently requires genuine capability. But scale alone does not explain a 92.9 percent renewal rate. Walmart has more scale than Costco. Sam’s Club has been running the same format for as long. Scale is a necessary condition for the model. It is not the structural advantage.

The second is that Costco succeeds because of operational discipline — the margin cap, the SKU restraint, the refusal to advertise. The discipline is real and has been maintained against persistent pressure. But discipline is a behavior, not an asset. What the discipline produced over time — the membership base whose renewal behavior has become habitual — is the asset. Costco is not valuable because its management is disciplined. Its management has been disciplined long enough that the result of the discipline has become self-sustaining.

The third is that Costco succeeds because members love it. Customer satisfaction scores are consistently high. The brand generates genuine affection. But affection is not a structural position. What looks like affection from the outside is, in structural terms, the behavioral expression of accumulated switching costs. Members do not renew at 92.9 percent because they are enthusiastic. They renew because the renewal has become automatic — because the shopping pattern is established, the private label dependencies are real, and the disruption cost of switching exceeds any plausible benefit from doing so.

What Costco actually built is a revenue stream that does not behave like retail revenue. It does not fluctuate with consumer sentiment in the way merchandise sales do. It does not compress under competitive pricing pressure. It does not require quarterly promotional investment to sustain. It compounds — not dramatically, not visibly, but year over year, renewal cycle by renewal cycle, as the membership base matures and the habit deepens.

The 2009 financial crisis tested this. Through a period of significant consumer spending contraction, Costco’s renewal rates held. The Q2 FY2009 earnings call documents the US and Canada renewal rate at approximately 87.5 percent — in the middle of the worst consumer spending environment in a generation. Merchandise sales declined. Fee revenue did not.

That durability is not a product feature. It is the structural output of twenty-five years of operational decisions that, taken individually, looked like restraint, and taken together, built something that conventional retail analysis had no framework to value.

DISCUSSION QUESTION

Costco’s renewal rates have held above 90 percent through recessions, fee increases, and the expansion of digital retail alternatives. The warehouse experience — the limited selection, the discovery dynamic, the physical scale — continues to justify the fee for the overwhelming majority of members who encounter it.

The experience that produces renewal is a physical one. It depends on members showing up, navigating the warehouse, encountering products they did not plan to buy. That is what the Kirkland brand is built on. It is what makes Costco structurally different from a subscription service that delivers boxes to a door.

The question the structural analysis opens, but does not resolve, is this: at what point does the shift in consumer behavior toward digital purchasing begin to erode the warehouse visit frequency that the renewal motivation depends on — and whether Costco’s response to that shift, when it becomes necessary, will require changes to the model that the model’s own incentive structure makes difficult to make?

If you have a view on this, I’d like to hear it. Leave a comment below.

The podcast version is available wherever you listen to podcasts or at deliberatedriftpodcast.com

This episode traces how that inversion accumulated — through decisions that looked, at the time, like operational discipline. Thin margins. Limited product selection. Above-market wages for retail workers. No advertising. Each of these looked like a constraint the company had accepted. What they were actually doing, in combination and over time, was building a membership asset that competitors could not replicate — not because the structure was secret, but because replication required accepting years of sub-market returns while waiting for something that only time could produce.

What this episode covers:

• How the warehouse club model worked and what condition it required to function

• Why Costco's margin discipline held through thirty years of public market pressure — and through multiple leadership generations

• Why Sam's Club has operated the same format since 1983 and still trails Costco's renewal rate by a significant margin

• The three reinforcing mechanisms that compounded the membership asset through the 2000s and 2010s

• What the 2024 fee increase — the first in seven years — revealed about the structural durability of the model

Full written analysis at deliberatedrift.com

In April 1998, Netflix launched as a DVD-by-mail service. You paid per rental, plus a small fee. The discs arrived in the mail. You watched them and sent them back. It was not an obvious disruption to anything.

Eighteen months later, in September 1999, the company switched models. For a flat monthly fee — initially around sixteen dollars — subscribers could hold up to three DVDs at a time, return them whenever they liked, and rent as many as they wanted. No due dates. No late fees.

At that moment, Netflix had a few hundred thousand subscribers. Blockbuster had roughly nine thousand stores, thirty million active customer accounts, and enough revenue from late fees alone to fund most of what Netflix was spending to exist. The competitive picture was not ambiguous.

What happened over the next eight years is usually told as a story about vision: Reed Hastings saw streaming coming, built a technology company in the guise of a DVD rental service, and outmaneuvered a complacent incumbent. That framing is satisfying, and it is largely wrong. Not because Hastings lacked foresight — he didn’t — but because the decisions that built Netflix’s structural position during the DVD era were not primarily acts of foresight. They were operational responses to immediate problems that accumulated structural value before anyone, including Netflix, fully recognized what was building. The streaming transition did not create that position. It inherited it.

The System

Netflix’s subscription model had a specific structural property that the per-rental model did not: it changed what the company needed to be good at.

In a per-rental business, revenue follows transactions. The better the selection of new releases, the more transactions. The better the store locations, the more transactions. Late fees were a secondary revenue line that required no additional capability — customers simply paid them when they kept discs too long, and the business collected.

In a subscription business, revenue follows retention. Every month a subscriber stays is a month of revenue that requires no additional transaction. Every month a subscriber cancels is revenue lost that requires an acquisition to replace. The operative variable is not how many discs get rented — it is how many subscribers stay. And subscribers stay when the product removes friction.

This shifted Netflix’s capital allocation logic entirely. Building more distribution centers was not a growth move — it was a retention move, because faster delivery meant subscribers got more value from their subscription and were less likely to cancel. Building a recommendation engine was not a technology indulgence — it was a retention move, because subscribers who found titles they actually wanted to watch used their subscription more and were less likely to cancel. Expanding the catalog was not a selection strategy — it was a retention move, because subscribers who could always find something worth watching did not conclude that Netflix had nothing left to offer them.

Each of these investments compounded the others. A subscriber who received fast delivery, had reliable recommendations, and could draw on a deep catalog was a subscriber with few reasons to leave. Each subscriber who stayed added another month of behavioral data to the recommendation engine. Each investment in delivery speed reduced churn, which reduced the cost of revenue.

The model required a growing subscriber base to function at the unit economics it needed. That too was accumulating: 857,000 subscribers at year-end 2002, 1.5 million by the end of 2003, 2.6 million by the end of 2004.

In 2000, Netflix approached Blockbuster about an acquisition. Blockbuster declined. Netflix was burning cash at the time, the subscriber base was small, and the outcome was not a foregone conclusion.

The structural value was not in the subscriber count. It was in the relationship.

The Shift

In August 2004, Blockbuster launched Blockbuster Online. It was a direct response to Netflix — subscription model, DVD-by-mail, flat monthly fee. Within a year, Blockbuster Online had grown to more than 750,000 subscribers. It added an in-store exchange privilege that Netflix could not match: subscribers could return DVDs to a physical store and receive a new rental immediately, without waiting for the mail cycle.

This was a genuine competitive threat, and Blockbuster’s leadership recognized it as such. John Antioco, Blockbuster’s chairman and chief executive officer, described the company’s strategy in its Q4 2004 earnings call as an effort to “build on the initial success of Blockbuster Online” and address what he acknowledged was a declining in-store rental industry.

The problem was not the strategy. The problem was what executing the strategy required Blockbuster to do to itself.

Blockbuster’s revenue model ran on late fees. At the Q4 2004 earnings call, Antioco acknowledged that late fees — which Blockbuster called “extended viewing fees” — had contributed approximately $250 to $300 million to the company’s operating income in 2004. He described the decision to eliminate them as “a lot of cash flow to offset and obviously a very serious decision.”

In January 2005, Blockbuster launched its No Late Fees program nationally. In the first quarter of 2005 alone, the elimination of extended viewing fee revenues reduced rental revenues by approximately $145 million. Blockbuster partially offset this through growth in base rental transactions — the No Late Fees program genuinely drove traffic — but the revenue model had been structurally altered in a way that required sustained volume growth to compensate for a lost revenue line that had required no volume at all.

Netflix reduced its monthly subscription price from $19.99 to $17.99 in October 2004, the same quarter Blockbuster Online launched. The price reduction cost Netflix revenue per subscriber. But Netflix was competing on a retention model where acquisition cost was the relevant variable, not operating income from a passive fee.

The Response

Blockbuster Online was not a halfhearted effort. Over the course of 2004 and 2005, Blockbuster invested approximately $140 million in capital expenditures and $120 million in operating costs to build the online subscription business and related initiatives. By the third quarter of 2005, Blockbuster Online had reached one million subscribers.

Netflix, at the same point, had approximately four million subscribers and was building toward profitability on a subscriber base it had been retaining for five years.

Blockbuster Online’s challenge was not execution — it was starting position. Building a distribution center network competitive with Netflix’s required years of iteration that Netflix had already completed. Building a recommendation system required behavioral data at the scale that Netflix had been accumulating since 2000. The in-store exchange privilege was a genuine differentiator, but it was also structurally tied to the retail cost structure that made the revenue sacrifices unsustainable.

The No Late Fees policy reversal later in 2005 captured the structural problem cleanly. Blockbuster had announced the policy as a permanent transformation. It reversed it — reinstating fees for some rentals — because the revenue impact was too large to sustain while also funding Blockbuster Online’s growth, managing the debt load from the 2004 Viacom spinoff, and maintaining the retail network.

The strategic response was not a solution to what Blockbuster had become. It was a bridge. When the bridge became too expensive to maintain, what remained was the same structural problem — a declining in-store rental business, a subscription competitor that had been accumulating advantages for five years, and a capital position that could not sustain both.

The Constraint Layer

Blockbuster Online faced five compounding disadvantages by 2005. They did not arrive in sequence. They shared the same resource pool.

The first was the distribution network gap. Netflix had been optimizing its network of distribution centers for one-day delivery since the late 1990s. By 2005, it operated more than forty facilities positioned to deliver to most of the U.S. subscriber base within a single business day. Blockbuster Online was building from zero in 2004. Matching Netflix’s delivery speed required years of network investment that would not generate returns until the network was built — investment funded from a balance sheet already weakened by the late-fee policy and the Viacom spinoff.

The second was the recommendation engine gap. Netflix launched its Cinematch recommendation algorithm in 2000. By 2006, the system had six years of behavioral data from millions of subscriber interactions. In October 2006, Netflix publicly announced the Netflix Prize — a $1 million competition to improve Cinematch’s accuracy by ten percent — a move that simultaneously drove product improvement and demonstrated the scale of the behavioral dataset to anyone paying attention. Blockbuster Online launched with a smaller subscriber base and a recommendation system with years less training data. The gap was not static. It widened with each additional month of Netflix operation.

The third was the catalog utilization structure. Netflix’s recommendation engine, by directing subscribers toward catalog titles, reduced the pressure on new-release inventory — historically the most expensive and competitive part of the rental model. Subscribers who regularly watched catalog titles needed fewer of the same new releases that every competitor was also trying to stock. This reduced costs while improving subscriber satisfaction. Blockbuster’s store-based model remained dependent on new-release traffic.

The fourth was the capital position. Blockbuster entered the critical competitive window — 2004 through 2006 — with debt from the Viacom spinoff, declining in-store revenues, the revenue cost of the No Late Fees policy, and escalating Blockbuster Online investment requirements. Netflix, by contrast, reached profitability in 2003 on a subscriber base of approximately 1.5 million — before Blockbuster Online launched — and was funding its growth from a clean balance sheet with steady recurring revenue. The capital asymmetry meant that every competitive move Blockbuster made cost more than the equivalent move by Netflix.

The fifth constraint was organizational. Blockbuster’s retail network — approximately nine thousand stores at its peak — was simultaneously its most valuable asset and its primary competitive disadvantage. The stores generated revenue, employed tens of thousands of people, supported a real estate and supply chain infrastructure, and defined the brand. They also generated the cost structure that made the revenue sacrifices required to compete with Netflix operationally unsustainable. Blockbuster could not become Netflix while remaining Blockbuster. Each constraint pulled from the same pool of capital and organizational attention that every other constraint also required.

The Compression

In January 2007, Netflix announced that it would offer streaming video as a feature for existing subscribers — at no additional charge — using the infrastructure already in place.

The streaming service launched not into an empty market but into an existing base of 6.3 million monthly subscribers, each with a credit card on file and a billing relationship with Netflix. The recommendation engine that would surface streaming titles had been trained on seven years of subscriber behavior. The brand identity associated with no-friction access to film had been built over eight years. The direct consumer relationship — unmediated by any physical location — had existed since the first subscription billing in 1999.

None of this was built for streaming. It was built to retain DVD subscribers. But the infrastructure requirements for a streaming launch — a direct billing relationship, a recommendation system, a brand identity around frictionless access, a subscriber base already paying monthly — were identical to the infrastructure Netflix had built for the DVD subscription model. The streaming transition was not a new beginning. It was a transfer.

Blockbuster entered bankruptcy in 2010, its capital exhausted and its strategic window closed. The DVD era was not the prelude to the interesting part. It was where the position was built.

What Netflix Left Behind

The standard framings of this story locate the decisive moment in different places — Reed Hastings’ foresight, Blockbuster’s complacency, the inevitability of the internet disrupting physical retail. Each of these frames carries some truth and misses the structural mechanism.

Hastings had foresight. But the decisions that built Netflix’s structural position were not primarily acts of foresight. The subscription model was an operational experiment. The no-late-fee policy was a customer retention mechanism. Cinematch was a solution to an inventory utilization problem. The distribution network was a response to subscriber complaints about delivery wait times. None of these were announced as moat-building. They became a moat because each one reinforced the others and because the model that made them necessary also made them valuable.

Blockbuster was not complacent. Blockbuster Online was a genuine, well-funded competitive response that reached one million subscribers in under a year. The problem was not effort — it was that the required response involved simultaneously dismantling a revenue model the company depended on, building capabilities from zero that Netflix had been building for five years, and sustaining losses on a balance sheet that could not absorb them indefinitely. That is not complacency. That is a structurally constrained position.

The internet did disrupt physical retail. But the streaming transition built on a foundation that had nothing to do with the internet. The subscriber relationship, the billing infrastructure, the recommendation engine, the brand identity — all of it came from eight years of DVD subscription operations.

The structural advantage Netflix carried into the streaming era was not technological. It was relational. It was the accumulated product of eight years of monthly billing, behavioral data collection, and retention-driven investment.

Discussion Question

Netflix’s structural advantages during the DVD era accumulated from decisions that were made for operational reasons — retention, cost management, inventory utilization — rather than as deliberate moat-building. The subscription model, the no-late-fee policy, the recommendation engine, the distribution network: each was a response to an immediate problem that happened to compound the others.

The question worth sitting with: if the structural advantage was not the product of deliberate strategy, what does that suggest about how durable advantages actually form — and how much of what we call strategic foresight is the retrospective framing of decisions that worked out?

If you have a view on this, I’d like to hear it. Leave a comment below.

The podcast version is available wherever you listen to podcasts or at deliberatedriftpodcast.com

The decisions that built Netflix's structural position during the DVD era — the subscription model, the no-late-fee policy, the Cinematch recommendation engine, the distribution center network — were not acts of foresight. They were operational responses to immediate problems. Each one compounded the others quietly across seven years.

By the time Blockbuster Online launched in 2004, it was not competing with a startup with a good idea. It was competing with a system that had been iterating for five years and had structurally different economics. By the time Netflix launched streaming in January 2007, it launched into 6.3 million existing subscribers, each with a credit card on file and a trained relationship with the service.

The streaming transition was not a new beginning. It was a transfer.

This episode covers the DVD era from 1998 to 2007 — the period when Netflix's structural position was built, before anyone, including Netflix, fully recognized what was accumulating.

In this episode:

• Why the subscription model changed what Netflix needed to be good at

• How Blockbuster's late-fee revenue model made the required competitive response structurally unsustainable

• The five compounding disadvantages Blockbuster Online faced that shared the same resource pool

• Why the streaming launch in 2007 was a transfer of an existing structural position, not a new beginning

Sources: Netflix 10-K filings FY2002–2007, Blockbuster 10-K filings FY2004–2006, Blockbuster Q4 2004 earnings call, Netflix Prize competition documentation, AP Wire contemporaneous coverage 2002–2007.

Full article and transcript at deliberatedrift.com.

In late 2009, Domino’s Pizza did something unusual. The company ran a national advertising campaign built around footage of its own executives reading customer complaints aloud. The pizza, customers said, tasted like cardboard. The sauce tasted like ketchup. The crust tasted like nothing. Domino’s did not dispute this. It apologized, announced a complete reformulation of its core product, and invited the country to watch.

The campaign worked. First-quarter same-store sales in 2010 increased 14.3% — the largest quarterly jump in the company’s history at the time. Business press covered it as one of the more remarkable brand turnarounds in recent memory. The story was clean and legible: admit failure, fix the product, rebuild trust, recover.

The recipe change was necessary. Without it, Domino’s would have continued losing ground in consumer perception to Pizza Hut and Papa John’s, and the digital platform it was simultaneously building would have been selling a product nobody wanted. The recipe mattered. But it was not the structural event.

The structural event was a decade of capital allocation decisions — beginning around the same time as the recipe change and continuing through 2018 — that built proprietary delivery infrastructure, a digital ordering platform, and a franchisee economics model optimized around owned last-mile delivery. Those decisions looked like operational overhead while they were being made. Some looked like marketing gimmicks. One — the refusal to partner with third-party delivery platforms at the moment every other restaurant chain was signing with them — looked like stubbornness or overconfidence.

By the time the delivery economy arrived at scale, Domino’s had been building for it for a decade. The gap between Domino’s and its competitors was not a product gap. It was not a brand gap. It was an infrastructure gap that required rebuilding supply chain, technology, data, and franchisee economics simultaneously to close — and no competitor has closed it.

What follows is not the turnaround story. It is the story beneath it.

The System

Domino’s was founded in 1960 by Tom Monaghan in Ypsilanti, Michigan, on a simple and durable premise: pizza delivered fast to your door. The 30-minute guarantee — introduced in the 1970s and discontinued in 1993 following liability litigation — became the brand’s identity for three decades. Domino’s was not trying to make the best pizza. It was trying to make a reliable pizza arrive quickly.

That distinction matters for understanding what the company actually built during its growth period. The operational infrastructure required to fulfill a 30-minute delivery guarantee — driver networks, kitchen throughput systems, routing optimization, delivery radius management — was not incidental to the brand. It was the brand. When the guarantee was discontinued, the infrastructure remained. Domino’s entered the 2000s as the largest pizza delivery company in the United States by store count, carrying a delivery system that had been refined over thirty years and a product that consumer surveys consistently ranked below its primary competitors.

The business model was straightforward. Domino’s charged royalties to franchisees — ongoing percentage-of-sales fees for use of the brand — and generated additional revenue by selling food, equipment, and supplies to franchisees through an internal supply chain it owned and operated. Dough was manufactured in Domino’s facilities and distributed by Domino’s trucks to franchise locations. The company did not merely collect fees on system sales; it had a direct financial stake in franchisee volume through supply chain revenues that grew as system throughput grew.

This structure created a specific incentive: Domino’s had more to gain from increasing the number of orders flowing through the system than from extracting margin on individual transactions. That incentive would govern capital allocation decisions for the next two decades.

In 2004, Domino’s completed its initial public offering under Bain Capital ownership. The company carried significant long-term debt from its leveraged buyout structure — a capital constraint that would, in practice, concentrate spending on the highest-return infrastructure rather than dispersing it across brand extension or product premiumization.

Patrick Doyle joined the company in 1997, became President of Domino’s US operations in 2008, and was named chief executive in 2011. The decisions that produced the structural advantage were made primarily under his leadership.

The condition the system required to function was simple: customers needed to believe that Domino’s was a reasonable option for delivery. In 2008, that belief was eroding.

The Shift

Two structural shifts were underway simultaneously in the late 2000s, and Domino’s was positioned to respond to both — though the response to neither was immediately legible from outside the company.

The first shift was the rise of digital commerce as a primary ordering interface. By the late 2000s, the question of where and how a customer ordered was becoming as commercially significant as what they ordered. E-commerce had already restructured retail, travel, and media; the restaurant category was moving later and more slowly, but the direction was not ambiguous.

The second shift was slower and would not become fully visible until 2015 to 2018: the emergence of third-party delivery platforms. UberEats, DoorDash, and Grubhub began their aggressive US market expansion in that window, offering restaurants a distribution channel for delivery without the capital investment of building an owned delivery system. For most restaurant chains — built around dine-in or counter-service models with no existing delivery infrastructure — this looked like an opportunity. For a company that had spent decades building owned delivery infrastructure, it looked like a threat dressed as a partnership.

In 2014, the three major pizza chains were running close together on digital order share — each measured as a proportion of their own total sales. Papa John’s was the first to cross 50%, reaching that threshold in late 2014, while Domino’s was at approximately 45% and Pizza Hut at approximately 40%. The metric alone did not separate them. What separated them was what each company concluded the metric meant, and what each company decided to do about the second shift as it arrived.

By February 2018, that divergence was visible in a single transaction. Yum Brands — Pizza Hut’s parent company — entered a master services agreement with Grubhub and invested $200 million in Grubhub stock. Pizza Hut would use Grubhub’s platform, logistics, and last-mile delivery infrastructure for its US delivery channel. The decision was rational given Pizza Hut’s position: it had not built owned delivery infrastructure, and Grubhub offered a path to delivery volume without the capital cost of building it. Domino’s, which had been building that infrastructure for eight years, declined to partner with any third-party delivery platform during the same period. The two largest pizza companies in the world, at roughly similar digital order shares four years earlier, had arrived at structurally opposite positions on who would own the last-mile delivery relationship.

In late 2009, Christopher McGlothlin, then Domino’s chief information officer, announced that the company was pulling its technology development in-house. At the time, Domino’s had an internal technology staff of approximately 150. McGlothlin was direct about the reasoning: the company had concluded that the technology was too important to continue outsourcing. He was not describing a digital transformation initiative. He was describing a decision about where competitive advantage would be located.

That decision — made at the same moment as the recipe reformulation, and receiving almost none of the same coverage — was the structural inflection point.

The Response

What Domino’s did between 2010 and 2018 looked, from the outside, like a sequence of mostly unremarkable operational decisions interrupted occasionally by marketing novelties.

The digital ordering platform launched in 2010 and was built and maintained internally. The engineering team grew from the approximately 150 at insourcing to a dedicated technology facility — the Innovation Garage, opened in August 2015 — housing 150 technology team members in a 33,000 square foot building designed specifically for developing and testing ordering and logistics technology. By 2016, Patrick Doyle was describing Domino’s publicly as “a technology company that sells pizza.” This was not a rebranding exercise. It changed what the company hired for, what it invested in, and how it evaluated competitive threats.

The Domino’s Anyware program, developed from 2012 through 2015, extended the ordering interface across more than fifteen platforms: Twitter, text message, smart television, smartwatch, Ford Sync, Amazon Echo, Google Home, and others. Each integration was built and maintained internally. In the trade press, these launches were covered primarily as novelties — ordering pizza by tweeting a pizza emoji generated considerable coverage and considerable skepticism about whether anyone would actually use it. The coverage was not wrong about the novelty. It missed the accumulation. Each integration was adding ordering surface area and, behind it, data. By 2014, Domino’s had launched Pizza Tracker — real-time order tracking with named driver identification and stage-by-stage updates. It was the first of its kind in the restaurant industry. It looked like a customer experience feature. It was also a logistics management system.

By 2015, digital orders had crossed a threshold Domino’s would record in its own filings:

“In 2015, approximately 50% of our U.S. sales came via digital platforms.”

The following year, the figure crossed “more than half” of US sales again — the language the company used in both its fiscal 2016 and fiscal 2017 annual reports — while global digital sales reached approximately $5.6 billion in fiscal 2016.

In 2016 and 2017, as UberEats, DoorDash, and Grubhub were signing restaurant partners across the industry, Domino’s explicitly declined to join third-party delivery platforms. The stated rationale centered on unit economics and data ownership — Domino’s did not want to pay commission structures on orders it was already fulfilling through its own infrastructure, and it did not want to cede the customer data those transactions would generate. From inside a company that had been building owned delivery infrastructure for decades and had just crossed 50% digital order share, the decision was consistent. From outside, it looked like a company leaving distribution on the table.

The Constraint Layer

To understand what Domino’s had built by 2018, it helps to understand what a competitor would have needed to build to replicate it — and why replication was not simply a capital question.

The platform. Domino’s proprietary digital ordering platform was not a licensed technology stack. It was built internally, maintained internally, and had accumulated data on delivery customer behavior across millions of transactions over nearly a decade. By fiscal year 2015, Domino’s was describing itself as among the largest e-commerce retailers in the United States in terms of annual transactions. A competitor seeking to build a comparable platform would be starting without that data history, without the internal engineering capability to build and iterate on the platform, and without the ordering surface area that Domino’s had spent years extending. A licensed platform from a vendor could replicate the interface. It could not replicate the data, the institutional engineering knowledge, or the integration depth across fifteen-plus ordering surfaces.

The supply chain. Domino’s owned and operated its supply chain — dough manufacturing facilities, distribution centers, delivery to franchise locations. By fiscal year 2019, supply chain revenues were $2.23 billion, making it the largest single revenue segment in the company’s financials, exceeding the Stores segment at $1.27 billion. The supply chain gave Domino’s cost visibility and quality control that competitors operating through third-party food distribution lacked. It also generated revenue that funded further infrastructure investment. A competitor replicating this position would need to build manufacturing and distribution infrastructure from scratch, at cost, with no revenue offset during the build period.

The franchisee economics. Domino’s franchisees operated on unit economics that had been optimized around owned delivery before the delivery economy arrived. That optimization ran through the supply chain: because Domino’s manufactured and distributed food to its own franchise locations, improvements in supply chain efficiency translated directly into franchisee cost structure — lower food costs, consistent quality, predictable delivery schedules. Corporate investment in supply chain infrastructure was not separable from franchisee profitability. They moved together. Average unit volumes were, by Domino’s chief executive Russell Weiner’s own characterization on an earnings call, better than any primary competitor at nearly $1.4 million annually. When third-party platforms began approaching restaurant franchisees across the industry in 2016 and 2017, Domino’s franchisees had a specific financial reason to resist: their existing system was already working. A competitor seeking to build comparable franchisee alignment would need to simultaneously improve unit economics and persuade franchisees to absorb the capital cost of building owned delivery infrastructure — while those franchisees were being offered what looked like an easier path through platform partnerships.

The data position. Every order placed through Domino’s proprietary platform generated data that Domino’s owned: customer identity, order history, delivery address, delivery time, driver routing, reorder patterns. At 50% digital order share in 2015, rising to more than half across both US and global sales by 2016 and 2017, Domino’s was accumulating this data at a rate no competitor in the pizza category could match. The data informed menu decisions, promotional targeting, driver routing optimization, and delivery radius management. It also made each subsequent order easier to fulfill, which reinforced digital order share, which generated more data. The loop was self-reinforcing, and it had been running for years before competitors recognized it as a competitive variable.

The identity constraint on competitors. The organizational identity Domino’s had built — explicitly a technology company, not a pizza company — was not merely a reputational claim. It determined what the company hired for, what it measured, and what it treated as competitively significant. A restaurant chain that had built its identity around food quality, dining experience, or brand prestige would face an organizational identity problem in attempting to replicate Domino’s position: it would need to become a different kind of company, not merely make a capital investment. Papa John’s, Pizza Hut, and the broader casual dining category entering delivery were not facing a technology gap alone. They were facing a gap in what they believed they were competing for.

None of these five dimensions closed independently. The platform was valuable because the data was deep. The data was deep because digital order share was high. Digital order share was high because the platform was good and the ordering surface was wide. The franchisee economics worked because the owned delivery infrastructure was already built. The owned delivery infrastructure was sustainable because supply chain revenues funded it. Each element supported the others. A competitor entering any single dimension would find the others already defended.

The Compression

The compression, in the Domino’s case, runs in reverse. It did not narrow options — it foreclosed competitor options as it accumulated.

In 2018, Domino’s surpassed Pizza Hut globally to become the largest pizza company in the world by sales. International store count exceeded US store count for the first time. The company had operated through a period of consistent same-store sales growth, rising average unit volumes, and expanding digital order share — all while carrying significant long-term debt from its recapitalization history and returning cash to shareholders through share repurchases. The capital allocation pattern was consistent with a management team that believed the infrastructure was already built and the returns were arriving.

Then, in 2020, the environment Domino’s had been building for arrived all at once.

COVID-19 accelerated delivery adoption across the restaurant industry by several years. Consumers who had previously dined in or carried out shifted to delivery under lockdown conditions. Third-party delivery platforms — UberEats, DoorDash, Grubhub — experienced volume increases that transformed them from startup-phase experiments into essential infrastructure for restaurants that had no owned delivery capability. Restaurants across every category signed platform agreements under conditions of urgency, absorbing commission structures that DoorDash’s own earnings calls described as typically 20 to 30 percent of an order — a cost structure that, for restaurants operating on single-digit margins, consumed most or all of the profit on each delivery transaction.

Domino’s operated through its existing infrastructure — no platform dependency to build, no commission structure to absorb, no urgent logistics problem to solve. Where competitors were encountering the delivery economy as a crisis, Domino’s was encountering it as confirmation. Restaurants that had signed with third-party platforms reported margin compression that Domino’s franchisees were not carrying. Third-party platforms reported significant per-order losses as they competed for market share, raising the question of what restaurant economics would look like when platform pricing reached its equilibrium. Domino’s franchisees were already there — and doing so with unit economics the platform model was structurally unable to match.

In 2023, Domino’s announced a limited partnership with Uber Eats — structured as carryout order visibility rather than delivery outsourcing, framed internally as incremental customer acquisition. Domino’s was using a third-party platform as a customer discovery channel while retaining the delivery relationship, the transaction data, and the per-order economics on the fulfillment side. It was the opposite of the arrangement most restaurant chains had accepted under urgency in 2020.

By fiscal year 2024, more than 85% of Domino’s US retail sales were generated through digital channels. The platform that started as an internal technology team of 150 in 2009 had become the primary surface through which the business ran.

What Domino’s Built

The standard framings of the Domino’s story do not survive contact with the structural record.

It was a turnaround story. The recipe change produced a real and necessary improvement in consumer perception. Same-store sales in Q1 2010 increased 14.3% — a genuine inflection point in consumer response. But a turnaround implies a return to a prior state. Domino’s did not return to anything. It built a position that did not exist before the decade of investment that followed the recipe change. The turnaround framing accounts for 2010. It does not account for 2020.

It was a technology story. The digital platform and the Anyware integrations are real and significant. But describing Domino’s as a technology company mislocates the advantage. The technology was valuable because it was embedded in owned delivery infrastructure, franchisee economics, and supply chain integration that no technology investment alone could replicate. A technology company can be disrupted by a better technology. Domino’s position required disrupting the supply chain, the franchisee network, the data history, and the delivery infrastructure simultaneously.

It was a management story. Patrick Doyle’s leadership was consequential — the conviction that the delivery relationship was the competitive asset, held consistently over a decade, governed the decisions that accumulated the advantage. But the advantage is structural, not personal. It persists beyond any individual’s tenure because it is embedded in the economics of 6,000-plus franchise locations and the data history of hundreds of millions of digital transactions.

What Domino’s actually built was a closed loop. The customer ordered digitally, generating data Domino’s owned. The order was fulfilled by a driver working within a system the franchisee owned and had optimized. The food arrived from a supply chain Domino’s manufactured and distributed. The franchisee’s unit economics made the system worth defending. The data made the next order easier. The loop ran, largely unobserved, for a decade before the environment arrived that made it legible.

The delivery economy did not create Domino’s advantage. It revealed it.

Discussion Question

Domino’s built its structural position during a window when the decisions looked like operational overhead, marketing experiments, or stubbornness — before the delivery economy made them legible as competitive investments. Leadership held a conviction about where competitive advantage would be located in a world that had not yet arrived, and allocated capital consistently with that conviction for nearly a decade before the bet was confirmed.

What would a competitor need to believe about the future of delivery to justify the investment required to close the gap today — and is there any evidence that belief is forming?

If you have a view on this, I’d like to hear it. Leave a comment below.

The podcast version is available wherever you listen to podcasts or at deliberatedriftpodcast.com

Takeaways:

• In 2009, Domino's admitted their pizza was bad and took steps to fix it.

• The company's transformation was not just about recipe changes, but strategic decisions made over years.

• Domino's built a unique delivery infrastructure that competitors are still trying to understand.

• By 2018, Domino's became the largest pizza company globally, surpassing Pizza Hut in sales.

• Digital ordering systems developed internally gave Domino's a competitive edge in the market.

• The success of Domino's came from a combination of factors that others can't easily replicate.

Links referenced in this episode:

• deliveratdrift.com

Companies mentioned in this episode:

• Domino's Pizza

• Uber Eats

• DoorDash

• Grubhub

• Papa John's

• Pizza Hut

• Yum

• Bain Capital

The conventional explanation is that Fisker ran out of money in a cooling EV market. That explanation is true as far as it goes. It doesn't go far enough.

This episode looks at the structural problem that was present from the beginning — a single manufacturing dependency built into the original Magna deal that nothing in the contract required Magna to maintain when Fisker's payments became uncertain. The asset-light model wasn't the wrong idea. The structure of the relationship was the wrong design for the risk Fisker was carrying.

Topics covered: the logic of the asset-light manufacturing model in 2020, the production ramp problems and recall cascade of 2023, the five constraints that accumulated simultaneously, and the moment the compression became irreversible — which wasn't February 2024.

Read the full analysis: deliberatedrift.com

Henrik Fisker had watched one car company carry his name into bankruptcy already. The first Fisker Automotive collapsed in 2013, taking its federal loan and its half-finished factory with it. When he started again in 2016, he built the second company around a different idea: the factory was the problem. He wasn’t going to build one.

Instead, he struck a deal with Magna International, one of the largest automotive suppliers in the world. Magna’s contract manufacturing division, Magna Steyr, operated a plant in Graz, Austria, that already built the Jaguar I-Pace — a battery electric vehicle, at scale, for a premium brand. The arrangement was straightforward: Magna would manufacture the Fisker Ocean to Fisker’s specifications, and Fisker would pay per vehicle produced. No factory. No billion-dollar construction timeline. No manufacturing hell.

In October 2020, Fisker merged with a publicly listed shell company — a structure called a special purpose acquisition company, or SPAC, which lets a private business go public quickly without the usual stock market listing process — and arrived on the NYSE with roughly one billion dollars. The model looked smart. The partner was credible. The vehicle, a premium electric SUV called the Ocean, had generated real consumer interest.

Fisker filed for bankruptcy in June 2024.

When it did, the Magna production line had been idle for four months. Talks with a large established automaker about a potential rescue had collapsed. Cash was nearly gone. The company had no factory, no alternative way to build its vehicle, and no path back to revenue.

The story that followed Fisker into bankruptcy court was about money: not enough of it, spent too fast, in a market that had cooled on electric vehicles. That story is true as far as it goes. But it doesn’t go far enough. The money ran out because the business model consumed it without generating enough revenue to sustain itself. And the business model was built that way from the start. The problem wasn’t the market, or the timing, or the capital raise. The problem was a single dependency — on Magna’s willingness to keep producing — that nothing in the original deal required Magna to maintain when Fisker ran short of cash.

The System

To understand why Fisker chose the path it did, you have to understand what the alternative looked like in 2020.

Tesla had nearly died building its Fremont factory. The company came within weeks of bankruptcy during the Model 3 production ramp in 2018 — Elon Musk was sleeping on the factory floor trying to solve assembly line problems by hand. Rivian, which had raised billions before delivering its first truck, was burning through capital so fast that analysts questioned whether it would survive long enough to reach meaningful sales volume. The pattern was consistent enough to have a name: manufacturing hell. Every electric vehicle startup that tried to build its own factory discovered that automotive manufacturing was extraordinarily difficult, extraordinarily expensive, and extraordinarily unforgiving of the mistakes that new companies inevitably make.

Fisker’s answer was to skip the factory entirely. Outsource manufacturing to a company that already knew how to do it. Keep Fisker focused on design, software, and brand — the things that create the experience a customer pays for — and pay Magna to handle everything else. This model is sometimes called asset-light, meaning the company avoids owning the heavy, expensive physical assets — the factories and machinery — that traditional manufacturers carry on their books.

Magna Steyr was not a speculative choice. It had built the I-Pace at Graz. It had built the Mercedes-Benz G-Class. It had built the BMW Z4. These were complex, premium vehicles with demanding customers and zero tolerance for quality failures. Magna knew how to build cars. Fisker wasn’t asking it to learn.

The Ocean came in four versions — the Sport at around thirty-seven thousand dollars, the Ultra and Extreme in the middle, and a limited One edition at the top — giving Fisker a range of price points without building separate platforms. Reservations came in. The thesis seemed to be holding.

There was one thing the model required to function. Magna had to keep producing. Everything else — the revenue, the brand, the survival of the company — depended on that single relationship staying intact.

Magna’s financial stake in Fisker’s success amounted to options to buy about 19.4 million Fisker shares at a penny each. It was not a partnership in any meaningful sense. Magna built vehicles. Fisker paid per vehicle. When Fisker couldn’t pay, Magna had no financial reason to keep building.

The Shift

Production began at Graz in late 2022. First deliveries went to European customers before the year ended. American deliveries started in mid-2023.

Fisker had planned to produce around 42,400 vehicles in 2023. It produced around 10,000. Deliveries fell short of even that number. By the end of the year, thousands of finished Oceans were sitting in storage — built, paid for, and unsold.

The Ocean had problems. The federal agency responsible for vehicle safety, the National Highway Traffic Safety Administration (NHTSA), opened investigations. Fisker issued multiple recalls covering door latches, brakes, and software. Reviews from automotive journalists were mixed in a specific and damaging way: the design was praised, the driving experience was often praised, but the software was unreliable and the build quality was inconsistent. That combination — a vehicle that felt promising but didn’t feel finished — is harder to recover from than a vehicle that is simply bad. A bad vehicle gives you a clear problem to solve. A half-finished one gives you a story that travels.

Each recall cost money. Each vehicle sitting in a storage lot was money already spent on production that hadn’t come back yet. The gap between when Fisker paid Magna to build a car and when a customer actually took delivery was the gap the company was slowly drowning in. Fisker paid per vehicle produced. Revenue only arrived when a vehicle was delivered. The wider that gap got, the worse the cash position became.

By the end of 2023, Fisker’s auditors had seen enough. The annual financial filing disclosed what the auditors had formally concluded: there was substantial doubt about whether the company could survive the next twelve months. This is a specific accounting term — a going concern warning — that auditors are required to issue when the numbers raise serious questions about a company’s ability to keep operating. Cash on hand was about 128 million dollars. At the rate Fisker was spending, that was months, not years.

Magna was watching all of this. It could read the same filing. And it had its own clients — Mercedes, BMW, Jaguar — whose vehicles shared the Graz facility with the Ocean. When the question shifted from ‘will Fisker pay us’ to ‘can Fisker pay us,’ the calculation changed.

The Response

Fisker tried two things.

The first was to cut prices. Discounts on the Ocean started appearing in late 2023, designed to move the vehicles piling up in storage and get cash coming back in. It worked, to a point. Vehicles moved. Cash came in. But discounting a car that was supposed to represent premium value sends a signal to the market that is very hard to unsend. Every buyer who paid full price is now looking at a cheaper version of what they bought. Every prospective buyer now knows the price is negotiable. Dealers watch residual values — what the car will be worth when customers trade it in — and those were falling. The discount strategy solved a short-term cash problem while making the long-term brand problem harder.

The second move was to find a buyer. By late 2023, Fisker’s leadership understood clearly that the company could not survive independently. What it needed was a large, established automaker — the kind of company, sometimes called an original equipment manufacturer or OEM, that has its own factories, its own supply chains, and a balance sheet that doesn’t depend on a single vehicle selling well in its first year. Talks began. Reuters reported in early 2024 that Nissan was involved. Fisker confirmed in a regulatory filing on March 22, 2024 that discussions with a large automaker had taken place and had ended without a deal.

Neither move addressed the actual problem. The discounts helped with cash but didn’t fix the Magna relationship, the recall backlog, or the production ramp. The acquisition talks, if they had succeeded, might have addressed all of it — but by the time those talks were happening, every number in Fisker’s public filings was visible to any prospective buyer. A going concern warning. A falling cash balance. Thousands of discounted vehicles in storage. A manufacturing partner that had already started asking when it was going to get paid. That is not a company anyone acquires on generous terms.

The acquisition talks were not a solution. They were a bridge. And by the time Fisker needed to cross it, the other side had already assessed what crossing would cost them.

The Constraint Layer

Five problems were running simultaneously through 2023 and into 2024. Each one made the others harder to solve.

The manufacturing dependency was the foundation of all of them. Fisker had no factory, no tooling, no way to build its own vehicles. It had chosen that deliberately. But it meant that every dollar of revenue the company would ever earn required Magna to keep producing. There was no backup. There was no partial solution. If Magna stopped, Fisker stopped.

The storage trap was the manufacturing dependency’s immediate consequence. Fisker had paid for thousands of vehicles that customers hadn’t taken yet. That money was gone — sitting in parking lots in the form of unsold cars. Moving those cars through discounts recovered some of it, but the discounts made the brand problem worse, which made it harder to sell future vehicles at the prices the business needed to survive. You couldn’t fix the cash problem without damaging the brand, and you couldn’t fix the brand without letting the cash problem get worse.

The recalls and quality issues added a third layer. Every recall meant engineering resources diverted to fixing problems on vehicles already delivered, at a moment when those same resources were needed to improve the software on vehicles still being produced. Every NHTSA investigation gave journalists and potential buyers another reason to hesitate. A reservation is only worth something when it converts to a sale. Fisker’s reservation list was not converting at the rate the business required.

The going concern warning changed the terms of every conversation Fisker was having. When a company’s auditors formally declare that its survival is in doubt, that declaration becomes public. Suppliers read it. Potential partners read it. Potential acquirers read it. Magna read it. Everyone who might have helped Fisker now had a documented reason to be cautious about how much exposure they were willing to take. The warning didn’t just reflect the company’s weakness — it deepened it.

And then there was identity. Fisker had been sold to investors, to customers, and to the market on a specific idea: that the asset-light model was smarter than building a factory. Henrik Fisker had built his public presence around that argument. To say out loud, in 2023, that the model wasn’t working would have been to repudiate the entire premise of the company at the moment the company most needed people to believe in it. So the story held. The discounts were framed as making the Ocean accessible. The production shortfalls were framed as temporary. The organizational identity had become a constraint on the honesty that might have opened different options.

Five problems. The same shrinking window. No reserves left to buy time with. No exit that didn’t require solving at least two of them simultaneously with money that was already running out.

The Compression

In February 2024, Magna stopped building the Ocean. Fisker hadn’t been paying its bills.

Magna’s chief executive Swamy Kotagiri addressed it directly on the company’s earnings call in May 2024. Production was idled. Their financial outlook assumed no further Ocean production. That assumption alone reduced Magna’s expected 2024 revenue by about 400 million dollars. In the same quarter, Magna wrote off its entire Fisker-related investment — the equipment at Graz allocated to Ocean production, and the share options it held — totaling 294 million dollars. Writing off an asset means concluding that it has no value left to recover. Magna had concluded the relationship was over.

The moment Magna stopped, Fisker’s revenue went to zero. Not reduced. Zero. There was no other product, no service income, no subscription revenue that could sustain the company while it figured out what to do next. The asset-light model had concentrated all of Fisker’s revenue into a single source — selling the Ocean — and that source required Magna. Without Magna, there was nothing.

On March 22, 2024, Fisker filed a regulatory disclosure confirming that the acquisition talks had ended. The company said it was exploring options: more financing, selling assets, restructuring its debts. These are the things a company says when it has run out of better options.

The moment the compression became irreversible wasn’t February 2024, when Magna stopped, or March 2024, when the acquisition talks collapsed. It was sometime in 2023, when Fisker’s cash burn rate and its deteriorating balance sheet crossed a threshold: the point at which any potential partner or acquirer could look at the numbers and conclude that a rescue would cost more than the company was worth. Everything after that threshold was consequence.

Fisker Inc. filed for bankruptcy on June 17, 2024, in the District of Delaware. The court process included an attempt to sell the business as a going concern — to find a buyer willing to keep it running rather than simply break it apart. No buyer came forward. The case converted to a full liquidation. The vehicles were sold. The company was wound down. The second company Henrik Fisker had built to carry his name was gone.

What Fisker Left Behind

Three explanations for what happened to Fisker circulate in the coverage. Each one is partially right. None of them gets to the actual problem.

The first is that Fisker ran out of money. This is true, but it skips the important question, which is why. Fisker raised nearly a billion dollars in its public listing and kept raising through 2021 and 2022. The capital was real. It ran out because the model consumed it without producing enough revenue to sustain itself — and the model was structured that way from the beginning. More money would have bought more time. It wouldn’t have changed the structure.

The second explanation is that Fisker got caught in the broader slowdown in electric vehicle demand. That slowdown was real, and it made everything harder. But Fisker’s structural problem existed independently of market conditions. A company with a single manufacturing partner and no leverage over that partner is vulnerable to the same dependency in a strong market — it just has more time before the vulnerability becomes visible. The market slowdown shortened the window. The window was already closing.

The third explanation is that the asset-light model failed. This is the closest, but it needs precision. The asset-light model as Fisker built it — with a single manufacturing partner, paid per vehicle, holding no meaningful stake in Fisker’s survival — created a dependency that contract terms alone couldn’t protect against. Magna was the right partner in 2020. The deal structure was not right for the risk Fisker was taking on. Those are different problems, and conflating them makes the lesson harder to see.

What Fisker actually demonstrated is something specific: that outsourcing manufacturing only works when the manufacturer has a genuine reason to absorb the difficulties that early-stage production always creates. A joint venture, a co-investment, a deal where Magna owned a real stake in Fisker’s outcome — any of those would have changed the incentives when payments started coming in late. A fee-for-service contract with a small share option attached did not.

The Fisker Ocean was a real car. It drove. People who drove it often liked it. The design was distinctive enough to attract buyers who weren’t just looking for a Tesla alternative. The company wasn’t a fraud. It was a genuine attempt to solve a genuine problem — how do you build an electric vehicle company without manufacturing hell — that ran directly into the limit of what the structure it chose could support.

The dependency was there from the start. It just took four years to matter.

Discussion Question

By late 2023, Fisker’s leadership could see clearly what was happening: the production shortfall, the vehicles piling up unsold, the cash running out, the auditor’s formal warning. They pursued the one option that might have saved the company — a deal with a larger automaker. What they couldn’t change was the position they were negotiating from. The balance sheet was public. The dependency on Magna was visible. The going concern warning was in the filing. The window for a deal on terms that would have preserved the company had already closed before the talks began.

Is asset-light manufacturing — outsourcing production rather than building a factory — a viable model for an early-stage vehicle company? Or does it only work when the manufacturer has a real financial stake in the outcome? And if Fisker had structured the Magna deal differently from the start — shared ownership, a joint venture, genuine skin in the game on both sides — would it have changed what happened in 2024?

If you have a view on this, I'd like to hear it. Leave a comment below.

The podcast version is available wherever you listen to podcasts or at deliberatedriftpodcast.com

In 2014, Spirit Airlines was the most profitable airline in the United States.

Not the largest. Not the most admired. By almost every measure of the passenger experience, it ranked near the bottom of the industry. But on operating margin, it sat at the top — ahead of Delta, ahead of Southwest, ahead of every carrier that offered more legroom, more reliability, and a reason to come back.

The model that produced that result was not complicated. Strip the product to its functional minimum. Charge the lowest possible base fare. Then charge separately for everything else — the bag, the seat selection, the itinerary change. Keep the aircraft moving. Keep the costs below what any legacy carrier could match. The gap between what Spirit charged and what a legacy ticket cost was wide enough that no amount of discomfort closed it for a price-sensitive traveler.

That gap was the business. Not the planes, not the routes, not the brand. The gap.

Bill Franke, who served as Spirit’s chairman during its most profitable years, said in 2025 that the original ultra-low-cost carrier model — where an airline competes primarily on base fare by stripping the product to its functional minimum — was gone for good in the United States. He was describing what had already happened to the industry. He was also describing what had already happened to Spirit.

On May 2, 2026, Spirit Airlines ceased operations. It was the first significant U.S. carrier to shut down in twenty-five years. Seventeen thousand direct and indirect employees lost their jobs. Millions of passengers held tickets that were immediately worthless.

The coverage that followed reached for the most visible explanation. Jet fuel prices had surged following U.S. and Israeli strikes on Iran in February 2026. A government bailout had been negotiated and then collapsed when a key creditor group rejected the terms. The airline had filed for Chapter 11 bankruptcy protection twice in less than a year.

All of that is accurate. None of it is the structural story.

Spirit did not fail because a war sent fuel prices up. It failed because the condition its model required to function — a price gap wide enough to justify every tradeoff the product asked of its passengers — had been closing for a decade. By the time the fuel shock arrived, there was nothing left to absorb it.

The System

Spirit’s ultra-low-cost model worked through a logic that was internally consistent and, for a time, genuinely difficult to compete with.

The mechanism began with the base fare. By stripping out every included service — carry-on bags, seat selection, itinerary flexibility, snacks, printed boarding passes — Spirit could price its tickets at levels no legacy carrier could match without losing money on the route. In 2011, Spirit’s base fares ran roughly forty percent below the industry average. Even after a passenger paid every applicable fee, the total price was still around thirty-five percent lower than a comparable legacy ticket.

Those fees were not incidental to the model. They were load-bearing. Non-ticket revenue per passenger grew from around five dollars in 2006 to forty-five dollars by 2011 — an increase of roughly eight hundred percent in five years.

The fees funded the operation. They also shaped it. Charging for carry-on bags reduced the weight passengers brought onto aircraft, which reduced fuel consumption and shortened turn times at the gate. The fee structure was simultaneously a revenue mechanism and an operational discipline.

The result was a flywheel. Low fares filled seats. Full aircraft spread fixed costs across more passengers. Low unit costs allowed fares to stay low. The airline’s 2011 annual filing describes this logic explicitly — the model was designed so that each component reinforced the others.

At peak, it worked. Spirit and Allegiant were the two most profitable U.S. airlines in 2014, at operating margins of 19.2 and 17.6 percent respectively. The legacy carriers — Delta, United, American — were profitable, but not at those margins. Spirit had built something the industry’s largest players could not replicate without dismantling their own cost structures.

The condition the model required was simple: the price gap had to be wide enough that passengers would accept the tradeoffs. No seat selection. No free bag. No flexibility. A Spirit ticket had to be cheap enough that the answer to every inconvenience was the price.

That condition held as long as legacy carriers couldn’t get close. It held until they could.

The Shift

Delta introduced basic economy fares in 2015. By early that year, the product was available in seventy-five markets, many of them overlapping directly with Spirit routes. United and American followed. The structure was familiar: non-refundable, no seat selection, bag fees required. The restrictions mirrored Spirit’s in almost every meaningful way.

What the legacy carriers kept was what Spirit had never had. Loyalty program miles. Hub network access. Operational reliability built on decades of infrastructure. A basic economy fare from Delta was not identical to a Spirit ticket — but it was close enough on price that the remaining differences started to matter.

Spirit’s chief executive Ben Baldanza acknowledged the pressure directly in December 2015, telling a Credit Suisse investor conference that Delta was very aggressive toward a carrier like Spirit and made it hard to grow in their markets. He argued the model was resilient enough to absorb it. He was not wrong that Spirit could absorb it. He was describing a ceiling, not a floor.

The numbers tracked the shift. Spirit’s total revenue per passenger flight segment fell from $137.19 in the first half of 2014 to $110.41 in the first half of 2018 — a decline of nearly twenty percent over four years. By 2018, Spirit’s own executives acknowledged in contemporaneous analysis that their price point was no longer always the lowest in the market.

The gap was closing. Not because Spirit’s costs were rising faster than expected — they were rising, but so were everyone else’s. The gap was closing because the legacy carriers had absorbed the lesson Spirit had taught the industry and applied it with resources Spirit could not match.

This is the shift the fuel shock coverage skipped. The external crisis in 2026 was real. But the structural condition that made Spirit vulnerable to any external crisis had been developing for a decade before a single Iranian facility was struck.

The Response

Spirit’s response to the competitive shift took two forms, and both were rational given what management could see at the time.

The operational response was to push harder on the levers the model provided. More routes. More aircraft. Higher utilization. The theory was straightforward: if the margin per passenger was compressing, volume could compensate. The airline grew its fleet aggressively through the late 2010s, reaching 194 aircraft by the end of 2022 — a thirty-four percent increase from its pre-pandemic size.

Ancillary revenue kept climbing in absolute terms, reaching $3.0 billion in 2023 against $1.1 billion in 2021. But the growth was masking the underlying problem. Non-ticket revenue per passenger was rising only marginally while fare revenue per segment was falling sharply. By the third quarter of 2023, total revenue per passenger flight segment had dropped 13.5 percent year over year, to $116.43. Fare revenue per segment was down 27.8 percent. The ancillary ceiling had been reached. There was no more room to push.

The strategic response was a merger. In February 2022, Spirit agreed to be acquired by Frontier Airlines. In July 2022, Spirit’s board switched to a higher offer from JetBlue — a $3.8 billion deal that would have created the fifth-largest U.S. carrier.

What makes this decision analytically significant is not the outcome. It is the process. Spirit’s own board, in a filing with the Securities and Exchange Commission in 2022, characterized the regulatory risk of the JetBlue deal in explicit terms. The board described the transaction as facing substantial regulatory hurdles and assessed it as not reasonably capable of being consummated given the anticipated Department of Justice opposition. The board approved it anyway.

This was not recklessness. The merger was the only structural path available. A carrier that couldn’t solve its cost gap problem organically, and couldn’t outgrow it through volume, needed scale it couldn’t generate alone. The JetBlue deal was a bridge — an attempt to transfer Spirit’s structural problem into a larger organization that might manage it differently. The board understood the regulatory risk and concluded it was worth taking because the alternative was to remain independent with a narrowing model and no structural resolution in sight.

On January 16, 2024, a federal judge blocked the merger. Spirit’s stock fell approximately forty-seven percent that day.

The bridge had been removed. What remained was the problem — unmodified, and now without a crossing.

The Constraint Layer

Five constraints were operating simultaneously by the time the merger failed. None was individually fatal. Together, they left no room.

The first was the revenue model. The price gap Spirit’s model required had narrowed to the point where it could no longer reliably justify the product’s limitations. Legacy basic economy had accomplished what no amount of Spirit efficiency could reverse — it had moved the competitive floor up to meet Spirit’s ceiling.

The second was the cost structure. Spirit’s unit costs had been among the lowest in the industry for years, and remained low in absolute terms. But the advantage was thinner than it had been. Labor costs rose from 22.1 percent of operating costs in 2022 to 28.1 percent in 2024. Operating cash flow ran negative $758 million in 2024. The cost discipline that had been the model’s foundation was holding — but the model it was holding up was no longer generating the returns that had once made the discipline worthwhile.

The third was the fleet. Spirit had taken delivery of new Airbus A320neo aircraft through the growth period, modernizing its operation. But when Pratt & Whitney issued a recall on geared turbofan, or GTF, engines in 2025, Spirit’s fleet was disproportionately affected. All seventy-nine of its GTF-powered aircraft required inspection and potential replacement. The recall process was expected to take more than two years. Planes on the ground don’t generate revenue, and Spirit had no reserve capacity to absorb the gap.

The fourth was capital. The airline ended 2022 with $1.8 billion in liquidity — a buffer that looked sufficient at the time. Three years of operating losses consumed it. The first bankruptcy filing, in November 2024, targeted a restructuring from $7.4 billion in total debt and lease commitments to $2.1 billion. The second filing, in August 2025, came before the restructuring had stabilized. By April 2026, Spirit’s attorney told the bankruptcy court that the airline’s cash was not going to last for very much longer.

The fifth was organizational identity. Spirit had built its entire operation — its cost structure, its customer base, its brand position, its internal culture — around the ultra-low-cost model. That identity was not a liability in the years when the model worked. It became one when the model stopped working, because it foreclosed the range of responses available. Spirit could not move upmarket without abandoning the cost discipline that kept it competitive. It could not build a loyalty program that competed with Delta’s without the network infrastructure Delta had spent decades constructing. It could not become something else, because everything it was had been built to be this one thing.

Five constraints. The same window. No reserves. No available exit.

The Compression

The restructuring plan Spirit filed in November 2024 was built on assumptions that were reasonable at the time. The airline projected a return to positive operating margins by the first quarter of 2026, moving from negative 27.1 percent in the first quarter of 2025 to negative 5.6 percent a year later. A net profit of $219 million was projected for 2027. The plan required fuel prices to remain stable, bookings to recover as the fleet stabilized, and creditor patience to hold while the restructuring took effect.

The fuel price assumption embedded in the restructuring support agreement was $2.24 per gallon for 2026.

On February 28, 2026, the United States and Israel conducted strikes on Iran. Jet fuel prices rose approximately seventy percent from that point. J.P. Morgan estimated that if fuel remained at $4.60 per gallon, Spirit’s projected 2026 operating margin would deteriorate from positive 0.5 percent to approximately negative twenty percent — adding roughly $360 million in costs the plan had not anticipated and the balance sheet could not absorb.

Spirit’s chief executive Dave Davis said in the wind-down statement that the sudden and sustained rise in fuel prices had left the airline with no alternative but to pursue an orderly wind-down. He was accurate. The fuel shock was the proximate cause of the closure.

But the restructuring plan had required all of its assumptions to hold simultaneously. Fuel stability. Booking recovery. Fleet availability. Creditor alignment. When one failed, the others became irrelevant. The plan was not designed to survive a shock of that magnitude because no plan calibrated to Spirit’s available resources could have been.

The fuel shock did not create Spirit’s problems. It revealed the degree to which every buffer had already been consumed.

Spirit ceased operations at 3:00 a.m. on May 2, 2026. The final flight landed at Dallas–Fort Worth shortly after midnight, arriving from Detroit.

The Ending

Three framings have dominated the coverage of Spirit’s closure, and all three are incomplete.

The first is that the Department of Justice killed Spirit by blocking the JetBlue merger. The merger’s failure was consequential — it removed the one structural path that might have resolved Spirit’s position. But the constraint accumulation that made a merger necessary had been building since 2015. A carrier whose model was structurally intact does not need a merger to survive. Spirit needed one because its model was already under pressure when the deal was announced.

The second is that the Iran war killed Spirit. The fuel shock was real, and at the margin it was decisive — it invalidated the one assumption the restructuring plan could not survive losing. But Spirit had filed for bankruptcy twice before a single Iranian facility was struck. The war accelerated a conclusion the financials were already approaching.

The third is that Spirit was simply a bad airline that deserved to fail. This framing mistakes the product for the business. Spirit was not beloved. It ranked second-worst among the ten largest U.S. carriers for passenger complaints in both 2023 and 2024. But its presence in a market suppressed fares by approximately fourteen percent on routes it served. Some routes were projected to see fare increases of fifteen to twenty percent following its closure. You did not have to fly Spirit to benefit from Spirit flying.

What the structural story reveals is something the simpler framings don’t reach. Spirit’s model worked because it created a price gap the industry couldn’t close. The industry closed it anyway, not by matching Spirit’s costs, but by matching its price points with a product that retained advantages Spirit had never built and couldn’t build. The model didn’t fail because Spirit stopped executing it. It failed because the competitive environment stopped rewarding it.

U.S. Transportation Secretary Sean Duffy said on the day of the closure that Spirit was in dire straits long before the war with Iran. He was describing the outcome. The structural story is about why the straits narrowed — and how long it took for the narrowing to become visible to everyone who hadn’t been watching the gap.

Discussion Question

The Department of Justice blocked Spirit’s merger with JetBlue in January 2024 on the grounds that eliminating Spirit as an independent competitor would harm consumers. Eighteen months later, Spirit filed for bankruptcy for the second time. Sixteen months after that, it ceased operations entirely — taking its fare-suppressing effect with it.

When a regulator intervenes to preserve competition, it is making a prediction about what the market will look like with and without the intervention. The DOJ predicted that a JetBlue-Spirit merger would reduce competition and raise fares. It may have been right about the merger. It does not appear to have been right about the alternative.

If regulators can model the competitive risks of consolidation but struggle to model the structural trajectory of the companies they’re protecting, what should the standard for intervention actually be?

If you have a view on this, I’d like to hear it. Leave a comment below.

The podcast version is available wherever you listen to podcasts or at deliberatedriftpodcast.com.

In this episode, Dawn Porthouse examines how Spirit's ultra-low-cost model depended on a price gap wide enough that passengers would accept every tradeoff — and how that gap closed, slowly and deliberately, long before the final crisis arrived.

Deliberate Drift analyzes how companies change structurally over time. Not through sudden failure, but through slow drift — decisions that appeared rational while constraints accumulated beneath the surface.

Full analysis at deliberatedrift.com