Flock Is Offering Buyouts. The Story Is Churn, Not Layoffs.

Flock Safety opened applications for a voluntary severance program on Friday. Employees have until October 2 to decide whether to take it.

Most coverage framed it as "buyouts to avoid layoffs," which is how the company would like it read. The number underneath is what matters. One advocacy group counted 93 city and county governments that cut ties with Flock in August alone. Across 2026, roughly three times as many local governments have dropped the company as in the previous five years combined.

This is not a cost story. It is a churn story wearing a cost story's clothes.

Flock builds automated license plate readers and sells them to police departments and city governments. It still operates well over 100,000 cameras nationwide, so this is not a company falling over. But the shape of what is happening to it is worth studying, because the same mechanics apply to anyone selling into a buyer who has to justify the purchase in public.

Three failure modes, arriving at once.

Accuracy. The Los Angeles Police Department's Inspector General released an audit on July 11, 2026 examining two months of stolen-vehicle alerts and found 161 vehicles incorrectly flagged, a 32.3 percent false-positive rate. The inspector general warned of "the risk of unnecessary enforcement actions, including vehicle stops and wrongful detentions." LAPD let its three-year contract expire three days later.

Data governance. In August 2025 an Illinois Secretary of State audit found Flock had given US Customs and Border Protection access to Illinois camera data through a pilot, contrary to a 2023 state law. Flock's CEO acknowledged the company "didn't create distinct permissions and protocols" to ensure local compliance. In Dayton, Ohio, police found more than 7,000 immigration-related searches by outside entities against city policy; workers covered all 72 cameras with trash bags.

Vendor conduct. Cambridge, Massachusetts terminated in December 2025 after finding two cameras installed without the city's knowledge, calling it a material breach of trust. Eugene, Oregon opened an inquiry when at least one camera kept logging plates weeks after officials ordered all 57 switched off.

Any one of these is survivable. Together they produced a narrative a council member could summarize in thirty seconds at a public meeting, and that is what actually moved the contracts.

Alt text: An empty city council chamber where vendor contracts are voted on in public Caption: When your buyer is a public body, renewal is a vote, and the public gets a turn at the microphone.

Because the renewal decision happens in a room anyone can walk into.

In a normal B2B deal, churn is a conversation between you and an economic buyer. If the product underperforms you negotiate, you discount, you bring in an executive sponsor. The failure is private and so is the fix.

When the buyer is a city, the renewal is an agenda item. Residents show up. Local reporters write it up. A council member who voted for the contract two years ago has to defend it, and the easiest defense is not having to make one. Harrisonburg, Virginia ended its contract on a unanimous vote after residents argued the readers infringed Fourth Amendment rights. Sedona ended unanimously too, after Flock acknowledged CBP had accessed local data.

The second difference is copying. Cities watch each other. Newsweek's mapping showed communities in 23 states rejected, canceled or deactivated the cameras in 2026 alone. That is not 23 independent decisions. That is one decision, propagating.

If you sell to school districts, municipalities, hospitals, universities, or any buyer with a board that meets in public, your churn is correlated in a way your dashboard will not show until it is already happening.

Most early-stage companies track logo churn and revenue churn. Neither tells you whether your customers can coordinate.

Try this instead. For your top 20 accounts, two columns. Is the renewal decision made in public, or by a person in private? And does this customer belong to an association, conference circuit, or regional consortium where they talk to each other about vendors?

Count the accounts where both answers are yes. That is your correlated churn exposure: the share of revenue that can leave in a cluster rather than one at a time. A company with 40 percent of revenue in accounts that talk to each other and decide in public does not have 40 small risks. It has one large one, and that changes how much runway you need.

If your model uses one blended churn rate, it is hiding this. Splitting the projection into a coordinated segment and an independent segment, each with its own rate, takes twenty minutes in a spreadsheet, in Causal, or in a planning tool like Foundra that keeps revenue assumptions next to the customer notes they came from. Model the segments separately, because they behave differently.

The 32.3 percent false-positive rate is the most-cited number in this story, and it is worth understanding why.

It was not produced by a journalist or an activist. It came from the customer's own inspector general, examining the customer's own data. That is the most dangerous kind of number, because it is unarguable. Flock could not dispute methodology, sourcing, or motive.

The uncomfortable question for your company: if your largest customer audited your core output next month, what would it find?

Most founders do not know. They know internal metrics, measured on data they control with a definition they chose. The gap between "94 percent accurate on our eval set" and "your product flagged 161 cars wrong in two months" is the whole distance between a healthy account and a public cancellation.

If you sell anything where wrong answers have consequences, publish your own error rate before somebody else does. You will lose a few deals to a competitor claiming a better number. You will also never be the subject of the sentence "an internal audit found."

Cleveland let its $250,000-a-year Flock contract lapse on June 29, 2026. Two weeks later the council approved a six-month, $125,000 extension.

It came with conditions: no data sharing with third parties, a public transparency portal, quarterly reporting, and access cut off for the regional fusion center. Half the money, half the term, and a list of constraints. The contract survived. The margin did not.

The lesson is timing. Every one of those conditions is something Flock could have offered two years earlier as a feature. Offered early, a transparency portal is a differentiator you charge for. Offered under duress during a renewal fight, it is a concession you pay for. Same artifact, opposite economics. The governance features your regulated customers will eventually demand are knowable now. Build them while they are still a sales advantage.

When a company in your market announces voluntary severance, most founders read it as weakness and move on. More signal is available.

A short decision window, October 2 here, usually means the company needs a headcount number before a board meeting or a fundraise and would rather not run a formal reduction that triggers notice requirements and press. It also means the people leaving will be the ones with options. Plan your recruiting accordingly.

More useful: it says the churn is expected to continue. Companies do not resize for a bad quarter, they resize for a revised forecast.

If you compete with them, the opening is not price. It is what caused the cancellations: audited accuracy, verifiable data boundaries, and a customer who can prove compliance to their own oversight body without calling you. Sell the audit, not the feature.

• Flock opened voluntary severance on Friday with an October 2 deadline, after roughly 93 local governments dropped it in August alone.
• Three causes clustered: a 32.3 percent false-positive rate found by LAPD's own inspector general, federal access to local data in violation of state law, and cameras installed or left running without authorization.
• Public-sector renewals happen in public, and public buyers copy each other. That makes churn correlated rather than independent.
• Model coordinated and independent customer segments with separate churn rates. A single blended rate hides the risk entirely.
• Publish your own error rate before a customer's auditor publishes it for you.
• Governance concessions extracted during a renewal fight, like Cleveland's transparency portal, are worth real money when offered two years earlier as a feature.

Is Flock Safety going out of business?

Nothing in the public record suggests that. It still operates well over 100,000 cameras nationwide. Voluntary severance signals a revised forecast, not insolvency.

How many cities have actually canceled?

Counts vary by methodology. One advocacy tally reached 93 governments cutting ties in August 2026 alone. The Washington Examiner put total departures since 2021 at roughly 98; Futurism counted at least 54 cities since the start of 2026. The direction is consistent across every count.

Does this mean selling to government is a bad idea for startups?

No. Public contracts are long, sticky, and often expand. The risk profile is different, not worse: slower to win, harder to lose on price, much easier to lose in a cluster when a narrative forms. Price that into your runway rather than avoiding the segment.

What should I do if a customer audit finds a problem with my product?

Publish your own number first, with your own methodology, before the audit happens. If it is already underway, cooperate fully, quantify the issue, and bring a remediation plan with dates. The failure mode is not the error rate, it is the vendor not knowing.

Are early concessions always worth making?

Governance ones usually are. Transparency reporting, data boundary guarantees, and audit access cost engineering time and almost nothing in margin. Offered early they are a differentiator. Offered late they are the price of keeping the account, usually alongside a discount.

1. Flock Offers Employees Buyouts as Customers Flee, WIRED, September 2026 (via Slashdot summary)
2. Cities that have dropped Flock Safety: cited cancellation tracker
3. Map Shows Cities That Have Rejected, Deactivated Flock Cameras, Newsweek, July 28, 2026
4. LAPD cancels Flock contract after cameras falsely flagged vehicles as stolen, Straight Arrow News, July 14, 2026
5. Why some cities are canceling Flock license plate reader contracts, NPR, February 17, 2026
6. Cleveland to keep Flock cameras, pause new data center projects, Signal Cleveland, July 16, 2026