Scoring every sales call

Scoring every sales call

A residential developer's sales team works through a lead base — hundreds of calls a day. The CRM shows how many calls were made, to whom and with what status. What the sales rep actually said on the call is invisible — and that's the core problem.

What breaks with call recordings

Recordings sit in the phone system, but they're reviewed selectively and with no shared standard: one supervisor focuses on the greeting, another only on the outcome. Gut-feel scoring can't be reproduced — the same call gets a five today and a three next week. New hires learn by trial and error, while the best techniques of top reps stay in their heads.

Why the CRM and manual listening didn't help

Telephony and CRM report statuses: "successful", "no answer", "lead". But behind a "successful" status there may be a call where the rep never asked about budget, never handled an objection and never offered a meeting. Listening to hundreds of daily calls by hand is simply not an option: the supervisor has neither the time nor the desire to sit through several hundred recordings every day, and a single call review takes minutes — even a spot check burns hours. In the end only a tiny fraction of calls gets reviewed, and conclusions are drawn from a sample.

What we did: score every call against one script

We connected an AI agent: every 15 minutes it pulls new calls from the phone system, transcribes them and scores each conversation against the sales script — greeting and self-introduction, clean speech without stop phrases, qualification (for yourself or family, floor area, budget, mortgage, timeline, developments already viewed, reservation), objection handling, value proposition, offering a concrete meeting, and next steps.

• Each call is scored across 20 criteria — up to 36 points for a new lead and 16 for a follow-up.
• Every score comes with a transcript and a per-criterion comment: not just "low score", but "didn't ask about budget" with a quote from the call.
• Stop phrases ("is it convenient to talk now", "don't want to buy", "sorry to bother you") are flagged automatically.
• The dashboard shows team and rep summaries, green/yellow/red traffic lights, weekly trends and growth areas for every rep.
• Scoring criteria live in a separate file: the supervisor can change the rules without a developer, and all subsequent calls are scored under the new rules.

A call ends — 15 minutes later its score and transcript are in the dashboard. The supervisor opens not a "monthly sample" but the full picture of every conversation.

No need to listen to everything

Instead of listening to recordings at random, the supervisor opens the rep ranking, sees which criterion is lagging for whom, and jumps straight to the call with a ready transcript. Performance reviews are built on facts — "in this call you didn't ask about the reservation", with a quote — not on "I feel like you don't ask enough". A rep moving from the red zone into the green is visible in the trends, not in impressions.

What stays with the human

The agent tags 100% of calls, tracks stop phrases and collects growth areas. The supervisor makes the decisions: who to praise, who to coach and on what exactly, and updates the criteria when the sales script changes. The machine counts and sorts, the human runs the team — the routine goes to the agent.