We report a vehicle-fleet experiment with an economic incentive for keeping within speed limits using a speed-alert device. A traffic insurance scheme was simulated for two months with a monthly bonus that was reduced by a non-linear speeding penalty. Participants were randomly assigned into four treatment and two control groups. A third control group consisted of drivers who had the device and were monitored, but did not participate. We found that participating drivers reduced severe speeding during the first month, but in the second, after having received feedback reports with an account of earned payments, only those given a penalty changed their behaviour.