Journal · AI agents

Sales call analysis: what to check and how to coach

Almost every CRM stores call recordings, yet few teams get much out of them: nobody has time to listen, and after listening it is not clear what to tell the rep. Call analysis works when you have criteria, numbers against them and a clear way to review.

Score a sales call by the stages of the conversation (opening, discovery, pitch, objections, next step) plus the rep's delivery. MakeBiz call speech analytics scores every call against your own script from AED 1,000 a month for 5,000 minutes, and the sales manager coaches one skill at a time using real moments from those calls.

MakeBiz Technologies · 11 November 2026 · 7 min read

Call analysis starts with criteria, not with listening

If a manager listens to a call without a list of criteria, the score depends on their mood and on whether they liked the rep's voice. A month later those scores cannot be compared across people or over time.

So first you break the sales script into stages and write down what counts as done for each one. Not "good discovery", but "asked at least two open questions about the client's problem and played it back in their own words". The more specific the wording, the fewer arguments in the review. How a dropped conversation gets tied to a specific stage is covered in our news piece on where sales teams lose clients inside the call.

Six blocks of criteria cover a typical sales call

The set below works for most phone sales, B2B and B2C. Trim it or extend it for your process, for example with a separate criterion for follow-up calls.

BlockWhat we checkRed flag
OpeningIntroduced themselves, named the company and the reason for the call, checked it is a good timeThe client asks who is calling
DiscoveryOpen questions, the client's problem, timing, who makes the decisionJumps to the pitch within a minute
PitchTalks about value for the client's problem, not the product in generalA monologue with no questions
ObjectionsFinds out what is behind the objection and answers the real concernAn argument, or an instant discount
Next stepA specific agreement: date, action, channel"I'll call you back" with no date
DeliveryPace, interruptions, the rep's share of talk timeThe rep does nearly all the talking

MakeBiz Technologies baseline criteria, November 2026. The final list is agreed against your script.

The blocks do not carry equal weight. A missed greeting rarely costs a deal, while a call with no next step almost always ends in silence. So criteria get weights: discovery and the next step score more, delivery scores less. Agree the weights with the head of sales before the first report, or the numbers will be disputed.

Read call scores by rep and by stage

An overall team average tells you very little. Two views are useful:

  • by stage across the team. If almost everyone fails the next step, the problem is the script or the training, not individual people
  • by rep on a single stage. If one rep keeps losing points on objections, that is the topic of their one-to-one

A third useful view is client questions and objections across all calls. Repeated questions show what to add to the script and the website, and repeated objections point to where the offer itself needs work. Much of this never makes it into the CRM, as we wrote in our news piece on what you hear on a call but never see in the deal.

We score every call against your script

Criteria built around your process, scores by rep and stage, top objections. From AED 1,000 a month for 5,000 minutes

See it on our calls

Coach one skill at a time

A common coaching mistake is listing everything that went wrong. The rep remembers one or two points and hears the rest as criticism. This order works:

  1. Pick one stage where the rep consistently scores low
  2. Find 2 or 3 calls where the mistake is obvious, and one where they got it right
  3. Listen to the clips together, starting with the good one
  4. Agree on one concrete action: which question to ask, which phrase to drop
  5. A week later, check the same criterion on new calls

A short session every week beats an hour-long meeting once a month: the rep has time to try one change and see it in their own numbers.

Training is built from real call clips

Once calls are scored, you end up with a library: strong answers to common objections, good discovery questions, examples of a clean next step. That is material for new hires and for team practice sessions.

  • collect 3 to 5 strong clips for each stage
  • add typical mistakes with a short breakdown, with rep names removed
  • refresh the set every month from new calls
  • have new hires listen to the set before their first solo calls

If your team sells in more than one language, build a separate library for each: a line that lands well in Russian does not always work in English. We train teams on the CRM and on call reports in our team training sessions.

Software scores every call, the manager reviews the flagged ones

No manager can listen to every conversation, and a random sample misses patterns. So the split is simple: call speech analytics transcribes and scores every call against your criteria, shows average scores by rep and stage, and suggests who needs coaching. The manager spends time only on the calls the system has flagged. The agent works on the same CRM data as our other AI agents.

Pricing: 5,000 minutes for AED 1,000 a month, 10,000 minutes for AED 1,700, and from AED 4,000 for 30,000 minutes or more. When a sample is enough and when you need full coverage is covered in our guide on call quality control. If your calls mix languages, we test transcription quality on your own recordings before launch.

For any of this to work, recordings have to land in the CRM together with the client record. What to check with your telephony provider before connecting is in our news piece on telephony for a CRM in the UAE.

A worked example: a team of 10 reps

A hypothetical Dubai company sells services over the phone and is sure it loses clients on price. Once every call is scored across the six blocks, the picture changes: most reps score low on the next step, and calls end without a date. For two months the sales manager coaches everyone on that stage only, and the script gets a mandatory question about the date of the next contact. After a month the manager checks whether that score has moved, and only then picks the next skill.

Follow-up after the call is picked up by the AI sales manager: it reminds a rep who has not messaged the client and replies when the rep goes quiet. All prices are on the pricing page.

Questions and answers

What criteria should I use to score sales calls?

A solid baseline is opening, discovery, pitch, objection handling, next step, and delivery. For each criterion, write down what counts as done so that different reviewers give the same score.

How many calls should I review with a rep?

For a one-to-one, 2 or 3 calls showing the same mistake plus one good call is enough. Regularity matters more: a short weekly session on one skill.

How much does call speech analytics cost?

With MakeBiz Technologies, 5,000 minutes cost AED 1,000 a month, 10,000 minutes AED 1,700, and 30,000 minutes or more from AED 4,000. Prices exclude VAT.

Can you analyse calls in English and Russian?

We work in English and Russian, and we write the scoring criteria in your team's language. We test transcription quality on your own recordings before launch, especially where calls mix languages.

Does the software replace the sales manager?

No. It scores conversations and shows where the problem is, but rewriting the script and coaching the rep is still a person's job.

Let us look at your team's calls

Tell us how many reps are on the phone and which telephony you use. We reply within a business day

By clicking the button you agree to the privacy policy

CallWhatsAppRequest