How the score is computed
The maths is deliberately simple so the number is defensible. Here is the whole algorithm, start to finish.
The steps
- Collect every signal on the lead from the last 365 days. Anything older drops out of the window.
- Age each signal. Most signals lose weight as they get older, so a favorite from this morning counts for more than one from March.
- Multiply each signal's weight by how much it has aged, then add them all up.
- Clamp the total to the 0 to 100 range. A lead with a huge pile of evidence tops out at 100.
- Map the number to its band using the fixed cutoffs.
The REHL AI layer
After the weighted sum, REHL AI reviews the recent signal context and can nudge the number by a small bounded amount, then writes a short rationale into the factors list. The bound is the point: a base score of 40 can land a few points either side, never somewhere else entirely. If the AI call fails for any reason, the plain weighted sum stands. Aggregation is always the source of truth.
Every stored score is stamped with the model version that produced it, so a score from an older version of the model stays interpretable when the weights are retuned.
Where the number lives
The signals are the source of truth; the score on the lead is a snapshot of the last calculation. That matters because the snapshot only refreshes when something triggers a recalculation, which is the subject of the next article.
Need more help? Contact support.