Five Failure Patterns That Repeat Across the Graveyard
What 116 dead companies in a July 2026 assessment of 10,525 companies reveal: five failure patterns that repeat, with names and dates.
Of the 10,525 companies WillItSurvive assessed in July 2026, 116 belong to the graveyard: companies whose terminal event has already happened. They were used in a blind backtest, judged only on facts knowable 18 months before the end, with identities masked and matched surviving peers judged at the same cutoff dates. That separation matters here, because when you read those 116 case files side by side, the same handful of patterns keeps recurring. We never declare an outcome for any living company. But the graveyard is settled fact, and it teaches.
A note on the comparison group first, because it anchors everything below. In the July 2026 backtest, matched survivors judged at the same cutoff dates drew negative verdicts only 17% of the time (10 of 58), versus 90% of the dead arm. And even the misses were instructive: all 10 "survivor" twins flagged negative by the blind judges in the July 2026 backtest, including Chegg, Fisker, Nikola, and Getir, turned out to be genuinely failed or gutted companies mis-selected as healthy peers. The seeding was wrong, not the judgment.
Here are the five patterns.
1. The subsidized land grab in a commodity business
Instant grocery delivery contributed a whole wing to the graveyard: Gorillas (cutoff 2021-06), JOKR (2022-06), Buyk (2020-09), Getir's US operations (2022-10), plus meal-delivery predecessors Munchery (2017-07) and Sprig (2015-11). The common shape: venture money subsidizing a below-cost consumer price in a business with thin structural margins and no switching costs. When capital markets tightened, the price could not rise fast enough to cover the model. The blind judges, seeing only pre-collapse facts, flagged this shape repeatedly. Notably, the pattern also caught nominal "survivors": Getir (cutoff 2022-06) and Flink (cutoff 2021-06) were seeded as healthy peers, drew not called safe in both replicates, and later proved to be genuinely troubled.
2. Hardware with no moat and one product
Juicero (2016-03), Humane's AI Pin (2023-08), Jawbone (2015-12), Pearl Automation (2015-12), and Cue Health (2022-11) share a structure: a single hardware product, high burn, and a value proposition a cheaper incumbent or a smartphone could absorb. Their matched survivors were companies with distribution and portfolio depth, such as Breville, Fitbit at the time, Mobileye, and Abbott Laboratories. The instructive twist is that two of the "survivor" seeds in this category, Fitbit (cutoff 2015-12) and Rabbit Inc. (cutoff 2023-08), were themselves flagged negative in both replicates and turned out to be mis-selected as healthy. Single-product hardware is a pattern the assessment treats with consistent skepticism regardless of which side of the pairing it sits on.
3. Capital-intensive moonshots racing a cash clock
EV and autonomy startups appear over and over: Fisker (cutoff 2022-12), Lordstown Motors (2021-12), Canoo (2023-07), Argo AI (2021-04), Drive.ai (2017-12), and Local Motors (2020-07). The pattern is a business that requires billions before first sustainable revenue, funded by equity markets whose appetite can close abruptly. When the assessment paired these companies with peers, the seeding itself demonstrated how narrow the survivor pool was: Fisker was proposed as the "healthy" twin for Canoo, and Nikola (cutoff 2021-12) as the twin for Lordstown, and both were flagged negative in both blind replicates before their own troubles became public record. In this category, finding a genuine survivor to pair against a dead company was itself difficult.
4. Legacy retail losing the middle
The largest single cluster by count is mid-market retail: Aeropostale (cutoff 2014-11), Bon-Ton Stores (2016-08), Forever 21 (2023-09), J.C. Penney (2018-11), Nine West (2016-10), HHGregg (2015-09), Century 21 Stores (2019-03), Stein Mart (2024-03), Zulily (2022-06), and Rite Aid (2023-11). The repeated shape: heavy fixed store costs, debt loads from earlier eras or buyouts, and a customer base pulled away simultaneously by discounters below and e-commerce beside. The matched survivors that held up in the backtest, including Ross Stores, Burlington, Best Buy, and CVS Health, tended to own either a cost structure built for discounting or a scale position the failed peer lacked. Also in the freight-adjacent version of this pattern: Yellow Corporation (cutoff 2022-02), paired against Old Dominion Freight Line.
5. Substitution shock: the product a platform or model replaces
The newest pattern in the graveyard is dependence on a workflow that a larger platform or a new technology absorbs. Kite, the AI code-autocomplete startup (cutoff 2021-05), faced GitHub Copilot-class competition. InVision (2022-07) was paired against Figma. Chegg (cutoff 2025-01) and Course Hero (2024-09) appear on both sides of a pairing, each seeded once as the other's survivor, and both drew not called safe in both replicates, an outcome the seeding-error review later confirmed as correct reads of a category under AI substitution pressure. Udacity (2022-09), Byju's (2023-01), Quibi (2019-04), Rdio (2014-05), and Hipmunk (2018-07) round out the group: products whose core job was being done better, cheaper, or free by something adjacent. Unacademy, seeded as Byju's survivor at the 2023-01 cutoff, was also flagged negative in both replicates and later validated as a mis-selected peer.
Why patterns, not predictions
WillItSurvive publishes no forecasts. The five patterns above are descriptions of what already happened, and things to check when you evaluate a living company yourself.
Questions this data answers
How was the graveyard backtest run? Dead companies were judged on facts knowable 18 months before their terminal event, with identities masked and context time-boxed, by the same production judges used for the live assessment. Matched surviving peers were judged at the same cutoff dates under the same masking, with two independent replicates on the survivor arm (assessed July 2026).
An identifiability probe tested this. In the unrecognized clean subset of the dead arm (n=17), 88% still drew negative verdicts with zero safe calls (July 2026 backtest), so recognition-independence is claimed for the dead arm. The clean survivor subset is small (n=5), so no equivalent claim is made there.
All 10 such cases in the July 2026 backtest, including Chegg, Fisker, Nikola, and Getir, turned out to be genuinely failed or gutted companies mis-selected as healthy peers during pairing. Both replicates flagged all 10.
Which failure pattern appears most often in the 116-company graveyard? By count, mid-market retail under simultaneous discount and e-commerce pressure is the largest cluster, with at least ten named cases from Aeropostale (cutoff 2014-11) to Rite Aid (cutoff 2023-11). Subsidized-delivery and capital-intensive vehicle startups form the next largest groups.
Does WillItSurvive forecast which companies are next? No.
Assessments reflect evidence available July 2026. Companies change; treat watch-items as things to verify, not conclusions.
Figures are as of each company's assessment date unless dated otherwise.