Predicting Insurance Risk
The risk was not in the demographics. It was in how departments actually operated, and the data made that visible.
Police and fire departments are expensive to insure. When a city self-insures its own public safety forces, which most of them do one way or another, it's taking on risk that's genuinely hard to measure. A bad year can cost a city tens of millions in liability. A single incident can reset the whole actuarial table. And for a long time, nobody selling insurance to municipalities had a real way to price that risk based on how individual departments were actually operating. They priced based on crude demographics — city size, population density, crime rate. Which is like pricing life insurance based only on the country you were born in.
While I was running another product at the same company, I realized we had something nobody else in the insurance market had. We had a broad operational picture of public safety risk. We could see which departments had elevated risk profiles and which ones didn't. Not based on demographics. Based on how the officers were actually acting, day to day, over years of data.
That was a whole new product. I pitched it to leadership. They backed it. I built the thing from the concept up, working initially with a team of two other people, and extended it to cover fire departments too. The goal was to give insurance carriers the real risk picture on the departments they were writing policies for, so they could either price the risk correctly, help departments fix what needed fixing, or decline to cover the ones that couldn't be fixed. All three options made the system work better than it was working before.
Shipping a new product line inside an existing company is, in my experience, harder than building one from a garage. You're fighting for attention, engineering time, and strategic air cover against everything else the company is trying to do. But it worked. The product shipped. It started generating its own revenue. And by the time the business was sold, it had become a real part of what made the whole thing valuable. Zero to one, inside a company that already had a one.