A look at Benchmark Analytics, which wants to help identify “bad apples” in US police departments using data and faces significant challenges, like data quality
Joshua Brustein / Bloomberg : Tweets: @joshuabrustein , @iwillleavenow , and @fleurej Tweets: Joshua Brustein / @joshuabrustein : Calls for police reform in the aftermath of the George Floyd killing have brought new attention to software that analyzes police data to predict which officers could commit misconduct, a kind of predictive policing in reverse https://www.bloomberg.com/... Calli Schroeder / @iwillleavenow : Why do the hard work of dismantling a systemic problem when you COULD slap an ineffective tech bandaid on it?* *This is sarcasm - this is a very bad idea. https://www.bloomberg.com/... Fleur Johns / @fleurej : Where does one even begin with all the ways in which this is not it. Cities Turn to Software to Predict When Police Will Go Rogue https://www.bloomberg.com/...
Context & Ripple Effects
Predictive policing software has spent years pointing outward — at neighborhoods and patrols, as the algorithms now shaping police patrols and probation rules show. Benchmark Analytics points the same machinery inward, scoring officers for misconduct risk just as post-George Floyd reform pressure gives departments a reason to buy it.
The pitch lands in a field already scarred: LAPD's PredPol and Operation Laser programs were shut down amid public outcry, and critics argue biased training data makes these systems unsalvageable. Benchmark's specific vulnerability is upstream — the misconduct data it depends on is the departments' own, and it is notoriously incomplete.
First-order effects
- Police departments evaluating Benchmark must first audit their own complaint and disciplinary records, since the vendor's output is only as good as data the department itself controls.
- Benchmark inherits the credibility burden of the predictive-policing backlash: critics like Calli Schroeder frame officer-scoring tools as tech fixes that avoid dismantling systemic problems.
Second-order effects
- The reliability questions defense lawyers have raised about Cybercheck's use in thousands of US cases preview the courtroom exposure for Benchmark: officer-risk scores could face discovery and cross-examination if they influence discipline or termination.
- Vendors of outward-facing predictive policing (Fusus-style camera-and-prediction platforms) may face the same scrutiny extended to their inputs, as the critique of biased data applies across the category.
Third-order effects
- If the PredPol and Operation Laser shutdowns are the template, officer-scoring software faces a lifecycle where public outcry, not measured accuracy, determines survival — pushing vendors toward transparency and auditability as a market requirement.
- The deeper shift is that police accountability increasingly runs through data infrastructure: whether departments can be held to their own records becomes the battleground between reform-by-software and structural reform.
The trend: Predictive policing is turning inward from places to people, and the same data-quality and bias critiques that felled neighborhood-targeting tools are now aimed at software that scores officers themselves.