A call scored high and we still lost the job. What does that tell me?

Scores measure handling; outcomes measure handling plus everything else

A rubric can only evaluate what is inside the call. Whether the customer could wait until Thursday, whether your trip charge policy is competitive, whether someone else answered first, and whether the customer had already committed before dialing are all outside it.

So a single high-scoring loss is not evidence of anything. Ten of them, examined together, are one of the most useful diagnostics available, because losses with excellent handling are where structural problems in the business become visible.

The ten-call review, done properly

Filter to bookable calls, above your handling threshold, that did not book. Sort by recency, take ten, and listen to the final ninety seconds of each. That is where the reason lives, and it saves an enormous amount of time compared to listening to whole calls.

Write down the stated reason and the unstated one separately. The stated reason is what the customer said. The unstated one is what you heard in the pause before they said it.

Structural losers no rubric will ever catch

• The soonest appointment was past the customer's tolerance. The most common one, and it is a capacity or scheduling problem, not a phone problem.
• A policy the customer would not accept. Diagnostic fee, minimum charge, deposit, or a payment method you do not take.
• Service or equipment you do not cover. Track the volume; it is telling you where your advertising is aimed wrong.
• The customer was already talking to someone else and called you as a second quote.

When the answer really is a missing criterion

Sometimes the pattern is in the handling after all, and it is something your rubric never asked about. Frequent examples: nobody asked when the work needed to happen, so urgency was never matched to the offered slot. Nobody named a specific time, only "sometime Tuesday." Nobody asked whether the customer had used you before, so a returning customer was treated like a stranger.

Add the criterion, rescore historical calls against the new version, and see whether it separates booked from unbooked. Rescoring history is only possible when audio and transcripts are retained and the scoring layer is configurable, which is a design decision worth making early in any evaluation program. Tie the result back to job value through revenue reporting before you decide how much the new criterion is worth.