What is a risk score?¶

A risk score is a platform-generated value that ranks how strongly selected signals match a risk model. Unless the product explicitly defines it that way, the number is not a measured probability of compromise or wrongdoing.

Scores combine model choices and uncertain inputs¶

Signals may include device familiarity, IP reputation, travel estimates, authentication strength, threat intelligence and account behaviour. Weighting and thresholds reflect the vendor or organisation's intended use.

Several weak or inferred signals can produce a high score, while trusted credentials may let hostile activity score low. The same number can mean something entirely different across products or model versions.

Decompose the score before relying on it¶

Obtain its purpose, scale, contributing features, weights where available, calculation time and version. Preserve component source records and test important inputs independently - especially inferred location or identity data.

Use the score to prioritise enquiry and describe it as the platform's assessment. Conclusions should rest on the contributing evidence, not on converting a label or value into unsupported certainty.