When Selling Is the Good Ending: Survivors That Sold vs Sank

Not every ending is a shutdown

When a company stops existing under its own name, there are two very different ways it can happen. It can run out of money and options, laying off staff, filing for bankruptcy, and leaving customers stranded. Or it can sell: an acquisition that pays back investors, keeps the product alive, and gives employees a landing spot. Both endings remove a company from the map. Only one of them counts as a good outcome for the people who bet on it.

This distinction matters a lot when you study company mortality, because a durability dataset that treats every disappearance as a death will misread the record. WillItSurvive assessed 10,525 companies in July 2026, and one of the sharpest lessons from that work came from the blind backtest, where the line between sold and sank turned out to be blurrier than expected.

The blind backtest, briefly

The backtest took 116 dead companies and judged each on facts knowable 18 months before its terminal event, with identity masked and context time-boxed. Each dead company was paired where possible with a surviving peer from the same category, judged at the same cutoff date with the same masking, and the survivor arm was run twice with independent replicates.

The survivor arm looked the way healthy companies should: in replicate one, only 17% of the 58 survivors drew negative verdicts and 24% drew positive ones. Two independent replicates of the July 2026 survivor-arm backtest agreed on 49 of 53 companies (92%).

In the unrecognized clean dead subset of 17 masked companies in the July 2026 backtest, 88% drew negative verdicts and zero drew safe calls. The clean survivor subset was small (5 companies), so recognition-independence is claimed only for the dead arm.

One rule governs how any of this is published: the site never declares an outcome for a living company. Forward-looking material appears only as dated watch-items readers can verify themselves.

The ten survivors that weren't

Here is where acquisition-versus-collapse gets interesting. In the survivor arm of the July 2026 blind backtest, 10 of 58 seeded survivors were flagged negative by the blind judges, and all 10, including Chegg, Fisker, Nikola, Fitbit, and Getir, turned out to be genuinely failed or gutted companies mis-selected as healthy peers. Both replicates flagged the same ten. The full list in the July 2026 readout: Chegg (cutoff September 2024), Course Hero (January 2025), Fisker Inc. (July 2023), Fitbit (December 2015), Getir (June 2022), Flink (June 2021), Nikola Corporation (December 2021), Rabbit Inc. (August 2023), Sun Basket (July 2017), and Unacademy (January 2023).

Fitbit is the clearest case of the acquisition question. In the backtest pairing (July 2026 dataset), Fitbit was seeded as the surviving twin of Jawbone at a December 2015 cutoff. Jawbone, as is widely known, liquidated in 2017. Fitbit did not liquidate; it was acquired by Google, a deal completed in 2021. So was Fitbit a survivor? The blind judges, seeing only pre-2016 facts, said no on both replicates. The readout's own conclusion was that the seeding was wrong: Fitbit at that cutoff was not a healthy peer but a company under real pressure whose eventual acquisition was closer to a rescue than a triumph. The stress was visible in the fundamentals.

The same logic runs through the other flagged twins. Chegg and Course Hero were seeded as each other's surviving twins in two separate pairs at 2024 and 2025 cutoffs, and both were flagged negative in both replicates, a symmetry the readout treats as evidence that the whole online study-help category was under structural pressure by then rather than one company beating another.

Sold, sank, or something in between

The pairing file shows how often an acquisition or absorption sits somewhere between the two clean endings. Some paired survivors in the July 2026 assessment are companies whose independence ended even though the business continued: One Medical, seeded as the surviving twin of Forward's CarePod clinics at a May 2023 cutoff, was by then already an Amazon acquisition, a widely known 2023 deal. That is a survivor by any customer's definition and a disappearance by a corporate-registry definition.

Contrast the dead arm. Quibi (paired against Netflix, April 2019 cutoff) shut down and returned capital, with no buyer for the going concern. Argo AI (paired against Waymo, April 2021 cutoff) was wound down by its backers in 2022, as widely reported. Yellow Corporation (paired against Old Dominion Freight Line, February 2022 cutoff) went through bankruptcy liquidation in 2023. Katerra, Juicero, Powa Technologies, Homejoy: the dead arm is a catalog of endings where nobody wanted the whole company at any price. That absence of a buyer is itself a signal. Acquirers pay for durable assets, and the companies in the dead arm mostly had none left by the end.

The supported claim from all of this is deliberately narrow, and the publishing rule states it exactly: given only pre-collapse facts, the assessment separated the doomed from matched survivors, and it never once said safe-to-join on a company that was dead within 18 months. It is not a claim that anyone predicted specific deaths, and it says nothing about who gets acquired versus who liquidates.

What this means for reading assessments

For readers of the main dataset, where roughly 10,900 public records carry verdicts, standing labels, and answer cards (about 10,800 companies have one, per the July 2026 assessment), the sold-versus-sank distinction is a reason to read watch-items carefully. A negative-leaning assessment is not a statement that a company ends badly. Fitbit-style outcomes exist: a stressed company can find a buyer, and its users can end up fine. What the backtest suggests is that the stress itself is usually visible early, whichever exit follows. The 1,571 contested records in the July 2026 assessment that carry an independent Opus second opinion exist precisely because ambiguous cases like these deserve two looks.

Questions this data answers

Does WillItSurvive count acquired companies as failures? No, not automatically. The blind backtest's dead arm of 116 companies (July 2026) was built around terminal events, and the seeding review treated acquisitions case by case. Fitbit's Google acquisition, for example, led the readout to conclude Fitbit had been mis-seeded as a healthy peer rather than to count the sale itself as a death.

How accurate was the blind backtest on dead companies? On the 17-company clean subset the recognition probe could not identify, 88% still drew negative verdicts with zero safe calls.

What happened with the ten survivors flagged negative? All ten seeded survivors flagged negative in the July 2026 backtest, including Chegg, Fisker, Nikola, Fitbit, and Getir, turned out to be genuinely failed or gutted companies mis-selected as healthy peers. Both replicates flagged the same ten, and Fisker was caught despite the recognition probe failing to identify it.

Can an assessment tell whether a stressed company gets acquired or liquidates? No, and the site does not try. The backtest evidence supports separating stressed companies from healthy matched peers using dated facts; it says nothing about which exit follows. That is why forward-looking material is published only as dated watch-items for readers to verify.

Why were Chegg and Course Hero both flagged? They appear as each other's seeded survivors in two separate pairs (cutoffs September 2024 and January 2025) in the July 2026 pairing file, and both drew negative verdicts in both replicates. The readout reads this as category-wide structural pressure in online study help rather than one company outlasting the other.

Assessments reflect evidence available July 2026. Companies change; treat watch-items as things to verify, not conclusions.