Does it even work?

the hit rate

• <1% of scanned cars are connected to any wrongdoing
• 0.3% of ALPR hits become a useful investigative lead

Everything else - every plate, every timestamp, every trip to your doctor - is recorded and retained anyway. The University of Michigan's Ford School put the wrongdoing figure at far under 1%. California Highway Patrol's own numbers showed roughly 0.3% of reads produced an alert at all.

This is the design, not a bug. A dragnet works by capturing everyone. The question is whether what it catches justifies what it costs.

U. Michigan Ford School · Center for Human Rights and Privacy

the marketing, audited

• Forbes, 2024

  The town Flock built its "70% crime drop" on saw burglaries go up

  San Marino, California installed Flock cameras in June 2020. Between January and May 2021, residential burglaries fell 80% - from 32 to seven. Flock seized on the window and marketed a 70% drop in all crime.

  Forbes looked at the wider picture. San Marino reported 60 residential burglaries in 2019, before the cameras. In 2023, three years after they arrived: 63. An increase. Serious offenses stayed effectively flat - 230 in 2019, 231 in 2023. The police chief, who still finds the cameras useful, acknowledged the marketing figure was not accurate.

  Forbes found the same pattern across four of Flock's most-cited jurisdictions, including Fort Worth, Dayton, and Lexington: a favourable few months, extracted and sold as a trend.

  Forbes

• Peer review

  "10% of US crime is solved with Flock" - a claim Flock studied itself

  It is Flock's flagship statistic, and it came from a paper Flock had a hand in producing. Six criminal justice academics reviewed it for Forbes. A University of Texas sociologist said the conclusion "borders on ludicrous" absent extraordinary evidence. A Columbia law professor doubted it would survive peer review at all.

  One of the academics involved later told 404 Media the underlying police data was too varied and incomplete to support meaningful statistical analysis. Flock published anyway.

  Forbes · Techdirt

• Cobb County, GA

  The police themselves wouldn't credit the cameras

  Cobb County is one of Flock's cited success stories. Its own deputy chief of police, asked whether the cameras caused the drop, said the department was not certain they made the difference. When the customer won't make the claim, and only the vendor will, that tells you what kind of claim it is.

  Independent Institute

sixteen years of data, one conclusion

• Piedmont, CA

  $576,378.80 spent. Less than 0.3% of hits produced a lead.

  Piedmont is an ALPR super-user - 39 cameras for a city of 11,000 people, ringing almost its entire border. It also kept unusually complete records: researcher Jonathan Hofer found no other department in the country with a comparable historical dataset.

  Analysing sixteen years of it, the report found no statistical evidence that the cameras deterred vehicle theft, no statistical evidence they generated investigative leads, and not even a moderate correlation between a hit and actually recovering a stolen car. Auto thefts dropped after installation, then climbed back to where they started.

  There's a detail worth knowing about the author. In 2018, Hofer and his brother were pulled over and thrown to the ground at gunpoint because a sheriff's ALPR misidentified their car as stolen. He is not against the technology on principle. He looked at the numbers.

  Independent Institute / Secure Justice

being honest about the research

The independent evidence is not uniformly damning - it's thin and mixed, which is a different problem. A 2025 evaluation of an ALPR expansion in Atlantic City found no reduction in violent crime, but did find associations with drops in shootings, vehicle thefts, and property crime. Some studies find small effects. Some find none. Rigorous evaluations remain scarce, and the technology spread far faster than anyone measured it.

That is the actual state of the field, and it's worth stating plainly rather than overclaiming in the other direction. But notice what it means:

Nobody has shown these cameras clearly work. They are being sold as if someone had. A company asking an entire country to accept permanent movement tracking does not get to treat "maybe, sometimes, in some places" as a mandate. The burden of proof belongs to the side taking something from you.

Journal of Policing evaluation, Atlantic City

what the money could buy instead

The alternative to cameras isn't nothing. It's the thing with better evidence behind it.

After READI Chicago began working in North Lawndale, the neighbourhood saw a 58% decrease in gun violence. Across US cities, researchers have found the presence of local community nonprofits statistically linked to reductions in homicide, violent crime, and property crime. That research is peer-reviewed. Nobody had to be entered into a federal database for it to work.

And there's a cost to getting this wrong beyond the money. The NAACP Legal Defense Fund has warned that predictive policing tools trained on biased data reproduce that bias while wearing the appearance of objectivity - injustice with a veneer of science. The Oak Park numbers are what that looks like in practice.

Brennan Center · American Sociological Association · NAACP LDF