These systems take historical crime incident data, sometimes combined with socio-economic statistics, and output predictions of where crime is likely in a given period, usually as small map cells with recommended patrol times. Officers are directed to those areas. The forecast is about places, not people, which is the central distinction from individual risk-scoring tools.

The canonical commercial example is PredPol, renamed Geolitica in 2021. The Los Angeles Police Department was its earliest adopter and co-developer, dropping it in 2020 citing budget, after an LAPD inspector general report found it unclear whether the software affected crime trends. Kent Police was the first UK force to test it, and cancelled the contract after saying it was challenging to show the programme reduced crime. Geolitica has ceased operations; SoundThinking acquired specific intellectual property and hired the engineering team, folding the capability into its own product line.

The Dutch Crime Anticipation System is the clearest European example and unusually well documented. Built in-house by a Dutch National Police data scientist with no external vendor, it divided cities into grid cells and combined police crime records with national statistics on income, benefit receipt, household composition and proximity to known offenders, outputting hotspots and hot times in four-hour blocks. Officers were sent to drive through flagged areas, often for only a few minutes. The model was closed, so officers could not see why a cell had been flagged. An indicator counting non-Western immigrants was removed in 2017 after being found to add no predictive value. Piloted from 2015 and rolled out from 2017 across more than half of the then 167 police departments, it was halted in January 2026 after an internal report concluded its predictive capability was extremely limited and the force could never demonstrate its effect.

The recurring finding across both is worth stating directly: in the two best-documented cases, the operating agency itself concluded it could not show the software worked. A Gizmodo and Markup analysis of millions of Geolitica predictions found on an unsecured server reported the software disproportionately and often relentlessly targeted low-income communities of colour.