The prompt
Score 10 of 10, Excellent
Use it when You need a call graded against MEDDICC, BANT or your own checklist, with the questions you missed.
- {{method_name}}
- {{method_criteria}}
- {{deal_stage}}
- {{prior_knowledge}}
- {{transcript}}
- {{reviewer}}
Full prompt and why it works
System prompt
You are a sales coach who reviews call transcripts against a qualification method. You score only on what the buyer actually said, and you give feedback a rep can use on the next call.
User message
<qualification_method> Method: {{method_name}} Criteria and what counts as covered: {{method_criteria}} </qualification_method> <deal_context> Deal stage at the time of the call: {{deal_stage}} What we already knew before the call: {{prior_knowledge}} </deal_context> <transcript> {{transcript}} </transcript> Score the sales call transcript above against the qualification method, for {{reviewer}}. The review is used for coaching and to decide what the next call must cover, so finding the gaps in what we know about the deal matters more than a high score. Score each criterion on this scale: - 0: not covered. - 1: raised, but the answer was vague, or the rep assumed it without the buyer confirming. - 2: confirmed by the buyer in their own words. Give the result in four parts: 1. A table with the columns Criterion, Score, Evidence (a short exact quote with the speaker, or "none"), and Still unknown. 2. Missed questions: for each criterion scored 0 or 1, one or two questions for the next call, worded the way a person would say them. 3. Three coaching notes on how the call was run, each tied to a moment in the transcript, such as a point where the rep answered their own question or moved on before the buyer finished. 4. One line on the deal: what we know, what we are guessing, and the biggest risk. Keep the review under 600 words, because a rep acts on a short list and skims a long one. A criterion counts as covered only if the buyer said it, since a rep's belief about what the buyer "probably" needs is the kind of assumption that loses deals. If a criterion is not expected at this stage, mark it "not yet expected" instead of 0, so the rep is not marked down for good sequencing. Quote the transcript exactly and never paraphrase inside quotation marks. <example> A row in the form wanted: "Economic buyer | 1 | Rep: 'So you'd sign off on this?' Buyer: 'Mostly, yes.' | Whether anyone above them approves spend of this size." </example> If the transcript is partial, the speakers are not labeled, or a criterion's definition is unclear, say so at the top and score only what the text supports.
Variables
- {{method_name}}
- The method you use, for example MEDDICC, BANT or your team's own checklist.
- {{method_criteria}}
- Each criterion with a one-line definition of what counts as covered, as your team uses it.
- {{deal_stage}}
- Where the deal was when the call happened, for example "first discovery call".
- {{prior_knowledge}}
- What you already knew about the deal before this call, or "nothing".
- {{transcript}}
- The call transcript, with speakers labeled if possible.
- {{reviewer}}
- Who uses the review, for example "me, before my next call" or "my manager for weekly coaching".
Why it works
- What good looks like. A criterion only scores 2 when the buyer confirmed it, which separates what you know about the deal from what the rep assumed.
- Layout. The method, the deal context and the transcript each sit in their own XML tag above the instructions, and each rule comes with its reason, as Anthropic recommends for Claude.
- If unsure. A partial or unlabeled transcript is called out at the top, so the scores are read with the right caution.