Relevance scoring

Relevance scoring is the practice of giving each incoming item a score that estimates how closely it matches an information need, so that people read the most pertinent items first and the rest can be filtered. The idea comes from information retrieval, where systems rank documents against a query; Introduction to Information Retrieval (Manning, Raghavan and Schütze, Cambridge University Press, 2008) gives a chapter to scoring, term weighting and the vector space model.

In regulatory monitoring the "query" is a company profile: its entities, licenses, products, markets and the topics it has chosen to follow. Keyword matching alone can score a planning consultation that mentions payments as highly as a payments rule, so current tools score each development against the profile and explain the result in words. The written explanation tells a reviewer more than the number does. In Regology's February 2026 survey of 204 compliance, legal and risk professionals, 85.3% said they monitor regulatory updates, but only 30.9% said the alerts they receive are always relevant (Regology, 2026).

Two cautions keep scoring honest. First, relevant is not the same as applicable: a score says a development falls inside your monitoring perimeter, while applicability to a specific entity is a judgment a person records in a regulatory change impact assessment. Second, anything filtered out should stay on record with its score and reason, so a missed rule can be traced to a threshold or a profile gap and fixed. RegWatch's triage, for instance, scores each finding from 0 to 100 against the company profile and writes a plain-language "Why this matters" for the items it accepts, while suppressed findings stay on record. Whatever the tool, calibrate thresholds on your own watchlist in a parallel run, and treat any unsourced accuracy figure with suspicion. The line between relevant and applicable is drawn in more detail in regulatory intelligence vs horizon scanning.