How to Write Avoma Scorecards That Reps Will Actually Trust
A scorecard that scores every call 40 out of 100 gets ignored in a week. The problem is almost never the scoring engine — it is that the criteria were written on a whiteboard, not from real calls.
This is the part of Avoma setup nobody documents. Here is a process that produces a scorecard reps will look at themselves.
Why do scorecard rollouts fail?
Three reasons, in order of frequency:
- Too many criteria. Twenty items means nothing in the output is actionable.
- Criteria that cannot be heard. “Built rapport” is not in the transcript. “Asked about the current solution in the first 8 minutes” is.
- No calibration. Managers score differently, reps get contradictory feedback, everyone stops caring.
Fix all three before you enable scoring on live calls.
How many criteria should a scorecard have?
Six to ten. Five is better for a new team.
A workable starter set for a mid-market sales motion:
- Opened with a reason for the call, not small talk
- Confirmed who else is involved in the decision
- Asked about current tooling and cost
- Quantified the problem in numbers
- Discussed pricing or procurement timeline
- Agreed a specific next step with a date
Every item is a yes/no the model can detect from the transcript. That matters more than whether you like the wording.
How do you weight and calibrate the score?
Weight qualification criteria higher than delivery criteria. A rep with poor pacing and great discovery will close more deals than the reverse.
Calibration process:
- Pull ten past calls where you know the outcome.
- Score them manually — 0 or 1 per criterion.
- Compare with Avoma’s output.
- Where they disagree, rewrite the criterion, do not adjust the weight.
Two rounds of that and the scoring aligns with how your managers actually think. One round is not enough; the first pass always over-asks.
What should the roll-out look like?
Do not email a score. Show the comments.
Week one: score privately, share the topic list and clips with each rep in a 1:1. No numbers.
Week three: share scores with each rep individually. Managers only.
Week six: if the criteria are stable, make the score visible in dashboards.
Skipping straight to a leaderboard is the fastest way to get everyone to decline recording. And once reps decline, you get no data at all.
On calls in other languages the transcript quality drops, so hold the score for 30 days and review manually rather than feeding bad data into a good scorecard.
How does scoring feed the CRM without noise?
Map three fields, not fifteen. Next step, decision-makers, and one risk flag. Everything else creates junk data nobody updates.
The action items in the summary should be reviewed by the rep before sync. Auto-pushed notes damage CRM hygiene faster than no notes.
If your team also needs to publish meeting content rather than just record it, AudioPen handles the cleanup step for written output, and Blizzy covers the video version of the same recording.
Quick Summary
- Six to ten yes/no criteria beats twenty vague ones; every item must be detectable in the transcript.
- Weight qualification above delivery — discovery matters more than smoothness.
- Calibrate against ten calls with known outcomes and rewrite criteria, not weights.
- Roll out comments first, individual scores second, dashboards third.
- Map only three CRM fields; auto-pushed notes without rep review damage data quality.
