The Disclosure and Barring Service (DBS) processes 4.2 million applications a year.
A significant proportion of these are sent to the police forces as referrals. To ensure adequate staffing levels, DBS wanted to enhance its predictions of referral applications. The BJSS Data Science team and DBS developed a time-series forecasting model for more accurate predictions of referral volumes.
BJSS hosted an intensive Data Science engagement to assess different forecasting models. The successful outcome of this activity informed the development of a PoC forecasting model which uses Facebook Prophet – Open Source software designed for forecasting time series data.
The PoC enabled the production of an 18-month referral forecasts and demonstrated a 50% lower error rate than the legacy forecasting model, ensuring that staffing levels correlate more closely with demand.