6.2 MetricsCDPH would deploy our models to rank children (or buildings) according to their risk for getting (or causing) lead poisoning. Due to limited resources, CDPH can only investigate a subset of cases. Therefore we measure the performance of a model by computing its precision in the examples predicted to be most at risk by that model. In this way we can estimate how many cases of future lead poisoning would be found, and so potentially ameliorated or avoided, if CDPH investigated a given number of cases.