A behaviorally anchored rating scale (BARS) is a scoring scale where each level is defined by a specific, observable behavior, so reviewers match what they saw to a description. The method dates to Smith and Kendall in 1963, and HR teams still use it for performance reviews.
Anchoring a call scorecard
Building one takes three steps. Collect real call moments for each skill, from strong calls and weak ones. Sort them into levels with your best managers, and write each level as what the rep did and how the buyer responded. Then test the anchors: if two reviewers scoring the same call land on different levels, rewrite the anchor they disagreed on.
- 1No question about costNo question about the cost or effect of the problem
- 2Asked, accepted a vague answerAsked about cost; accepted "it's a lot"
- 3Got a figure from the buyerOr a named consequence
- 4Tied to a business goalA goal the buyer owns
- 5Buyer states the cost of inactionIn their own numbers; agreed to fix it this year
It suits a sales call scorecard because the behavior is in the transcript. Each skill, such as impact or stakeholder mapping, gets its own ladder. A reviewer can then climb the ladder by asking "is this at least level 2? 3?" and stop at the first no, which is how our final scoring method works.
The wording of the anchors decides how well it works. In our research program (August to October 2026, more than 5,000 tests and more than 250,000 scoring decisions, with blind marking by PeerLab's trained assessors), giving our scorer the reasons trained assessors gave for each level lifted its agreement with human scores from 0.62 to 0.72, the largest single gain in the study. Textbook examples of good calls added nothing (field guide). More in the AI call scoring guide.
A five-level ladder for impact
An anchored ladder for impact on a discovery call (illustrative wording):
| Level | Anchor |
|---|---|
| 1 | No question about the cost or effect of the problem |
| 2 | Asked about cost; accepted a vague answer ("it's a lot") |
| 3 | Got a figure or a named consequence from the buyer |
| 4 | Tied the figure to a business goal the buyer owns |
| 5 | The buyer stated the cost of doing nothing in their own numbers and agreed it needs fixing this year |
A manager hearing "the delays cost us real money" with no follow-up scores it a 2 and coaches one question for the next call on that deal: "What did those delays cost you last quarter?" To see one of your calls scored on anchored levels, talk to us.
Related terms
- Call scoring: rating a call against a scorecard, skill by skill
- Coaching scorecard: the rubric a call is scored against
- Rank agreement: how closely a scorer's order matches the experts'
FAQ
Why use behaviorally anchored scales for call scoring?
Reviewers agree more when each level describes what happened on the call. HR University notes BARS can reduce leniency bias against plain rating scales. In our tests on sales calls, giving our scorer expert reasons for each level was the largest single gain; a general-purpose model given the same reasons did not improve.
How long does it take to build a BARS scorecard?
HR University puts a full BARS for one role at 4 to 8 weeks. A call scorecard is narrower: start from your best managers' reasons for scores they gave on real calls, and refine the anchors in calibration sessions.
Can AI score calls on a behaviorally anchored scale?
Yes, AI can score calls on a behaviorally anchored scale when each level carries the reasons assessors gave for it. Our final method asks a model yes/no questions up each skill's level ladder, with assessors' reasons attached; averaged with a second model's blind whole-call score, it reached 0.89 rank agreement with experts. Can AI score a sales call? covers the test.