Finding risks that spreadsheets miss
A large UK outsourcing company needed to know whether their financial forecast for a welfare-to-work programme could be trusted. Job seekers placed in work don’t always remain employed. Some cycle in and out of work many times over. The rare cases were uncommon, but disproportionately costly to resolve, and nobody could say whether the forecast was accounting for them properly.
The company’s existing model was built in Excel, and only had enough cells to represent each claimant taking up to three jobs. That was enough to see the shape of the problem, but not to answer the question that mattered: were the rare, expensive cases - claimants cycling through six, seven, eight job placements - common enough to threaten the financial viability of the contract, or rare enough to ignore?
We built an agent-based simulation that modelled every claimant individually, each one free to cycle in and out of work as many times as the data implied, instead of being capped and bucketed like the spreadsheet. We cross-validated it against the original Excel model to check it reproduced the parts both models agreed on, before trusting it on the parts only the simulation could imagine.
The simulation showed that the small share of claimants cycling through far more than three placements accounted for a disproportionate share of cost exposure. This was exposure the original model had no way to represent at all. That gave the finance team a defensible answer to a question they’d previously had to guess at: how much risk was really sitting in the programme’s long tail.
Confidence in your numbers
Find out how models can help you find answers that aren’t in the data.