Most health scores start at the first login and lean on usage. With AI products, usage and value have drifted apart. This demo scores 20 fictional accounts from the first discovery call onward, so a risk can be spotted when it first appears, not at renewal.
Retrospectives usually happen after an implementation goes wrong. Here, the accounts that went best and the ones that struggled are both compared against what was recorded in discovery and at kickoff, before anyone knew how they would turn out.
The bigger the gap between the two groups on a score, the more weight it earns.