Pricing a policy blind is just writing a cheque you haven’t read yet.
That was essentially what Thornfield Insurance was doing. Their commercial liability team was underwriting SME policies based on whatever the applicant wrote on the form — turnover, employee count, industry. Honest answers from honest companies. But companies aren’t always honest, and even honest ones change.
The problem surfaced when their loss ratio on a specific SME cohort crept 11 points above target over 18 months. Digging into the claims, the pattern was ugly: companies that had grown significantly since application, changed their primary activity, or quietly accumulated late filings — all risk signals that never made it into the original pricing.
Their underwriters were essentially pricing yesterday’s company at today’s rates.
Thornfield integrated Borsch.ai’s data enrichment into their application workflow. Every policy submission now gets cross-referenced against live Companies House data — actual filed turnover, SIC codes, director history, filing behaviour, and any changes since the last renewal.
The first thing they found: 23% of renewal applications had material discrepancies between declared and filed financials. Some were innocent rounding. Some were not.
More importantly, they could now price dynamically. A cleaning company that had quietly pivoted into construction contracting? Different risk profile. A director with two previous dissolved companies? That changes the conversation.
Six months post-integration, their SME loss ratio dropped 8 points. Application review time per policy stayed roughly the same — but the decisions got significantly smarter. They also caught four cases warranting referral for potential misrepresentation.
Better data didn’t slow down underwriting. It just stopped it from being a guessing game.
See what’s behind your next application: https://borsch.ai