A look at PredPol, a predictive policing startup which began as a LAPD/UCLA research project and is now used in over 60 jurisdictions
Context & Ripple Effects
This 2016 Fusion profile captures PredPol at its commercial high-water mark: a tool born as a joint LAPD/UCLA research project now sold into more than 60 jurisdictions, presented as the success story of crime science turned startup. At this point the dominant frame is efficiency — sending patrols where crime is predicted.
The later corpus tells the other half of the arc. A 2021 investigation found PredPol perpetuates bias, directing police toward poor, Black, and Latino neighborhoods, while documents showed the LAPD's follow-on effort closely resembled PredPol and Operation Laser after they were shut down amid public outcry. Meanwhile rivals took different shapes: Palantir ran a secret person-based pilot in New Orleans, the NYPD built its own in-house Patternizr, and Fusus scaled camera-network fusion to 60+ cities.
First-order effects
- Police departments in over 60 jurisdictions are buying place-based patrol predictions from a vendor whose credibility rests on its academic pedigree — procurement decisions made before independent audits of the tool exist.
- LAPD and UCLA gain a template for commercializing publicly funded criminology research, with the university lending legitimacy the marketing alone could not buy.
Second-order effects
- Competitors differentiate around PredPol's weaknesses: Palantir shifts from predicting places to flagging individuals in New Orleans, Benchmark Analytics turns the same data-driven logic inward on officer misconduct, and NYPD brings development in-house with Patternizr to avoid vendor scrutiny.
- Public backlash becomes a product risk — when PredPol and Operation Laser are shut down amid outcry, agencies and vendors learn that rebranding and secrecy (as with Palantir's undisclosed deployment) are cheaper than transparency.
Third-order effects
- If the pattern holds, predictive policing survives not as any single product but as a rotating cast of successors — Fusus merging private cameras with prediction is the next generation — while investigative journalism, not regulation, functions as the primary accountability mechanism.
- The deeper structural shift is that police departments become data-infrastructure buyers, locking in surveillance vendors whose systems outlive the controversies that kill any one brand.
The trend: Predictive policing is proving durable through a shutdown-and-rebrand cycle, with each controversy spawning successor tools faster than oversight can constrain them.