Sales rep performance review: how to judge the conversation, not the person

Quality doesn’t show up in a report. It shows up in one conversation, read to the end.

Two reps, identical numbers: fast replies, plenty of conversations, close to plan. One walks clients to payment. The other answers everything except what was asked. The report can’t tell them apart.

Metrics answer “how much.” A daily look at the chats answers “where to step in.” A quality review answers a third question: “how exactly does this person talk to clients.” It takes three things: a short scorecard, a sensible sample and a calm conversation afterward.

The scorecard: five to seven questions for a conversation

Not a one-page form. A handful of questions, each answered by pointing to a spot in the chat.

1. Was the client’s need understood. The rep asked what the person wanted, not what was easier to discuss.
2. To the point and without errors. Price, timing and terms are stated correctly. A direct question gets a direct answer.
3. Tone. Even and respectful. No pressure, no irritation, no corporate boilerplate.
4. Next step set. The conversation ended with specifics: who does what, and when.
5. Real promises. Nothing the company won’t deliver or never approved.
6. Close or agreement. The conversation reached the result possible at this stage: payment, a meeting, an agreed date.
7. Company rules. Everything in your client chat policy: client data, off-limits topics, handoffs, escalation.

Shorter is better. A criterion you can only “feel” should be rewritten or dropped.

Reply speed can stay off the scorecard: a number measures it more honestly than a reviewer’s eye, and it has its own rules. The scorecard is for what a number can’t capture.

The scale: yes, partly, no

A ten-point scale looks precise but isn’t. One reviewer gives a seven where another gives a five, and the review becomes a negotiation.

Three values work better:

• yes — the criterion is met;
• partly — met, with a loss: a step without a date, a right answer three messages late;
• no — not met.

For “partly” and “no,” the reviewer marks the exact spot in the conversation. Without it, a score is just an opinion.

Don’t add the marks up into one overall number for the person. Watch which criterion slips again and again. Promises and company rules are special: a “no” there gets its own discussion, even when everything else is fine.

The sample: how many conversations, and which

You don’t need to read everything. Three to five conversations per rep a week is usually enough. A month gives you a picture worth discussing.

Split the sample in half:

• Random. Everyday work, not extremes.
• The ones that matter. Lost deals. Long threads with no outcome. Complaints or irritation. Large amounts.

Only failures, and the picture looks darker than it is: the rep feels hunted, not reviewed. Only random ones, and you miss the most expensive moments.

Read each conversation in full. A single line pulled from the middle almost always misleads. The rep doesn’t choose what gets reviewed. New hires get a bigger sample; experienced reps can be reviewed less often, but at least once a month.

Calibration: so that “yes” means the same thing

With two or more reviewers, scores drift apart. That’s normal, but reps notice fast: “one says yes, the other says no, so it isn’t about me.”

Once a month, every reviewer independently scores the same three or four conversations, then compares. Where scores differ, sharpen the criterion and write down an example. After a few rounds, the scorecard has reference cases: this is “yes,” this is “partly.”

Have the rep score the same conversation. Where your scores differ, the review starts.

The one-on-one

A score without a talk is useless. A talk without a format becomes a reprimand. The format is simple.

1. Start with a good example. Specific, from the rep’s own chats. Good work gets noticed too.
2. Then the questionable spot, as a question. “What did the client want here?” “What would you write now?” A question lets the rep find the mistake themselves. A verdict teaches them to defend.
3. One thing to improve. Not five. One, until the next review.
4. Write it down and check it. In the next sample, look at that criterion specifically.

Fifteen to twenty minutes is enough. Regularity beats length. A review that ends with a list of mistakes teaches people to hide conversations, not to write better.

Training and probation

If one criterion slips for many people, the problem isn’t the people. Training is missing, the policy is silent, or the product is explained wrong. That’s a signal for you, not a reason for individual reviews.

The best conversations from your samples are a ready-made textbook: real chats from your own company, not someone else’s scripts.

During probation, the scorecard is at its fairest. The new hire sees it on day one and knows what they will be judged on. The sample is larger, reviews more frequent, and the final decision rests on how scores moved over the whole period, not on one bad conversation.

A review, not surveillance

Surveillance isn’t about how much you read. It’s about the rules.

• Company work accounts only. Personal chats are never reviewed.
• The team knows the criteria in advance. The scorecard and examples are open.
• You judge the conversation, not the person. “In this conversation no next step was set,” not “you can’t close.”
• Two people see the result. The rep and their team lead, not the whole team.
• People decide about people. Whoever gives the first read — a colleague or AI — it’s a reason to reread the conversation yourself, not a verdict.

As for the law: put the fact that work chats are reviewed into company rules or the employment contract. Personal data rules differ by country; if in doubt, ask a lawyer.

You settle all of this before you think about tools: who reviews, who sees the result, what goes into the rules. Answers first, choices second. How it looks on our side is a conversation for those who were brought in.

Short checklist

• A scorecard of five to seven criteria is written down and known to the team
• The scale is simple: yes, partly, no, with the spot in the chat marked
• A few conversations per rep are reviewed every week, random and important
• Reviewers compare scores on the same conversations once a month
• A review starts with a good example and ends with one step
• Recurring misses go into training, not punishment
• A new hire on probation knows the scorecard from day one
• Only company work accounts are reviewed

A good scorecard looks different six months in. If it has never changed, chances are nobody is using it.

---

The conversation gets the score. The person gets better.