Help
Predictive features are those traits that show up unusually frequently (or infrequently) in the applications of past hires who achieved the desired outcome (e.g. staying in seat for at least 6 months). These features are not filters; in fact, they are independent elements that impact on probability. This means that a candidate possessing a positive predictor makes them more likely to achieve the desired outcome. It is very unlikely that a candidate will have none of the negative predictors. For this reason, the color predictions are based on the underlying weightings for each trait.
We cannot share these weightings with clients; but they are a critical part of making the algorithm effective. They cannot (and should not) be used as standalone traits to make decisions about who to take through to the next part of the process. However, some of the traits can guide sourcing decisions (e.g. when a specific job board is listed as a negative predictor, clients can adjust how they source in line with this).