How to
A deal score is accurate when the deals you won scored clearly higher than the deals you lost, using scores calculated before those deals ended. Most teams never check, because the score was switched on and then trusted. The check needs your closed deals, the score each one carried, and one comparison.
Take every deal you have closed. Find the score each one had shortly before it ended. Average the scores of the ones you won, average the scores of the ones you lost, and compare the two.
If the averages are close, the score is not telling you anything. If the won deals sit well above the lost ones, it is doing real work. Everything else is detail on that one comparison.
A score read after a deal closes is not a prediction. Won deals collect activity right up to the end, so a score calculated at the close rates them highly almost by definition.
The score has to be the one the deal carried before the outcome was known. If your CRM only stores the current value, recompute it from the activity that existed at an earlier date. A point a little before the close is a reasonable choice.
You get two numbers and a gap. Wins averaged one figure, losses averaged another, across a stated number of deals. The gap matters more than either number.
There is no universal size that counts as good. What matters is that the gap is large compared with how spread out the scores are inside each group, and that it holds when you split the deals into older and newer halves.
A small gap means the weights are wrong for your business, not that scoring cannot work. The signals your buyers give before they sign are usually specific to how you sell, and a generic score weights them generically.
The fix is to find which signals separated your wins from your losses and weight those. That is what a back test reports, and it is the part most tools skip.
Questions
Compare the average score of your won deals with the average score of your lost deals, using scores calculated before those deals closed. A score is doing real work when wins sit well above losses. When the averages are close, the score is not separating anything.
It was probably calculated after the outcome was known. Won deals collect activity up to the close, so a score read at the end rates them highly by construction. The honest test uses the score a deal carried before the outcome was known.
Use every closed deal you have. Small samples can be checked, but a pattern on a few dozen deals is a hint rather than a finding, and a larger book makes the result harder to argue with.
A HubSpot score can be tested the same way. Export your closed deals with the score each one carried before the close and run the comparison. If only the current value was kept, recompute the score from activity at an earlier date instead.
It arrives in Slack
Every morning, each rep gets one message naming the deals on their own book that moved overnight, biggest mover first, with the reason beside each number and the deals that have gone quiet underneath.
$99 a month, cancel any time.. Everyone on the team. No per seat charge. Connect a CRM and it runs. No implementation fee. 14 day trial. No card.
Related reading
A back test scores deals you have already closed as they stood before they ended, then checks the scores against what happened. The steps.
Most teams trust signals they have never tested. Reply speed, email opens and meeting count can all carry no information. How to find out which.
Kaypo Hindsight is a free, read only back test. It reads your closed deals and reports what separated the ones you won from the ones you lost.
Every pipeline has deals that are alive and deals being carried. The difference shows in what the buyer did, not the stage somebody dragged the card to.
HubSpot ships a predictive deal score with Sales Hub Professional, and its own documentation says not to rely on it. Why that matters.
Reviewed September 20, 2026 against the product as it behaves today.