We strengthened response validation in Visibility Opportunities (intent-segmented entry-question analysis). Abnormal or unusable AI responses are now filtered more strictly so they never blend into your results, and on-screen figures were aligned with the underlying measurements.
Automatic refunds on total failure. If a run fails to complete even a single measurement — for example, during a temporary outage — the credits used are restored automatically and a failure notice is shown. Partial failures keep only the completed measurements in your results.
Since these results drive decisions, the principle is simple: if you have results you can trust them, and if you don't, you don't pay.