Stranger things
The arrest count is the part everyone repeats and the part that means least. 173 collars across 24 deployments is a fine number, but the question the Met carefully declines to answer is what fraction of those bodies would have been picked up anyway through, you know, policing. A woman wanted since 2004 for assault did not become un-findable for 21 years through some impressive feat of fugitive craft. She presumably had a National Insurance number, a GP, and an address the whole time. Read one way, the LFR success story is also a confession that the force had filed her case in the drawer marked "later" and forgotten which drawer.
The same applies to the 37 breaches of conditions. These are by definition people the state already knows: electronic tags, SHPO registrations, bail addresses. Locating a man on a tag does not normally tax the investigative imagination. If the Met needed a camera on the High Street to do it, the interesting headline is about case management, not facial recognition.
"One arrest every 35 minutes" is the other line that does heavy lifting and means nothing. Spread 173 arrests across 24 operations of a few hours each and the suggestive cadence appears. Spread them across the actual six months between October and March and you get roughly one a day, which is less of a press release and more of a Tuesday.
Then there is the accuracy theatre. "470,000 walked past, one false positive" sounds magnificent until you notice most of those 470,000 were never matched against anything, because they were not on the list. The denominator that matters is matches generated, of which we are told precisely nothing. A system that flags one innocent passer-by per million and one per fifty matches are very different propositions, and only the latter tells you whether being stopped on the strength of an LFR ping is a coin toss or a near-certainty.
"Crime in this area is down by more than 10 percent" arrives without a control group, a baseline period, a displacement analysis, or anything else a GCSE statistics candidate would ask for before signing the press release. National crime trends, seasonal variation, and high-street footfall after the cameras went up are all left as exercises for the reader.
Step back from the individual numbers and the piece reads, to anyone hunting for the awkward figures, as the shape of a conclusion already reached with the evaluation written to fit. Every metric on offer points the same way; none of the obvious follow-ups are entertained; the comparisons that would make the headline numbers meaningful are unaccountably absent. The overall tone is that of an organisation strikingly uncurious about its own results, apparently satisfied with knowing that 173 people were arrested and apparently uninterested in knowing how many would have been arrested without the cameras, or whether any of the alerts that did not result in an arrest should have been alerts in the first place. That posture matters, because the case for continuing to install permanent surveillance infrastructure rests on questions the operator declines to ask.