Seeing what is coming, so a community can act first.
We build predictive models for community organizations: where and when violence is likely to rise, which blocks need outreach this week, how a season or a heat wave changes the risk. The models are built from public data, read by the people who know the streets, and used to move resources, never to label people.
Six principles
Public data, local reading
Dispatch feeds, crime reports, weather, property and blight data, census and health indicators. Every model is read with outreach workers before anyone acts on it.
Places and times, not people
We score places, hours, and seasons. We do not build lists of people. That line does not move.
Seasons and context
Cleveland has a rhythm: summer, school dismissal, holidays, heat. The models carry a seasonal baseline so a spike is read against what is normal for the week.
Results, not formulas
Partners see tiers, hot spots, forecasts, and return-on-investment figures. The weights and formulas are proprietary and stay in the research workspace.
Honest about error
Every forecast carries its uncertainty. A model that cannot say when it is unsure should not be trusted.
Built to be kept
A pipeline that runs itself each morning, a dashboard the team actually opens, and a review sheet for the weekly meeting.
Predictive analytics for community organizations
Six sessions for program directors, outreach leads, and the person who ends up holding the spreadsheet. No coding required; the goal is to read a model, question it, and use it. In development.
What a prediction is
Baselines, rates, and why last year is not a forecast.
Finding the data
Open data portals, dispatch feeds, census, health and property data, and what each one leaves out.
Seasons and hot spots
Reading a map and a calendar together. Where, when, and how sure.
Risk without labels
Scoring places and times, protecting people, and the ethics of acting on a forecast.
From forecast to field
The weekly review: turning tiers into a deployment plan and a follow-up.
Measuring the return
What an intervention saved, in harm avoided and dollars, and how to say it to a funder.
Where this work runs today
Partner organizations run their own dashboards built with us. Each site belongs to its partner.
Renounce Denounce Gang Intervention Program
Daily crime intelligence, seasonal hot-spot tiers, a weekly field brief, and a community meetings calendar for Cleveland's East Side.
Visit the partner siteResearch and analytics, in the portfolio
The composite risk system, conflict mapping, and the return and harm model, described without the formulas.
PortfolioResearch workspace
Methods, instruments, and drafts for approved collaborators only.
Restricted