What this system does in one paragraph
Every sales call is recorded. We download the recording, turn it into an English transcript, and score that transcript against the pitch our reps are supposed to follow. Before this, someone had to listen to a call to know how it went, so under 1% ever got reviewed. Now all of them do, within minutes.
The four numbers, and whose problem each one is
| Number | What it measures | Whose problem | Scale |
|---|---|---|---|
| Pitch adherence | Did the rep follow the pitch — greet properly, ask discovery questions, capture the requirement, commit to a next step? | the rep | 0–10. Under 5 red, 5–7 amber, 8+ green. |
| Customer quality | Problems with the lead we handed the rep: not a manufacturer, unserviceable area, order below minimum, material we don't stock. | the lead | 0–4 flags. More is worse. |
| Supply | Times we couldn't serve the customer — out of stock, lead time too long, spec not carried. | us | A count of calls. |
| Sentiment | How the customer sounded. Warm and engaged, or annoyed and dismissive. | the customer | −1 to +1. Around 0 is neutral. |
Two things use green / amber / red, and they mean different things
"Could we hear the call?"
About the recording. Nothing to do with the rep.
| full conversation | 6+ turns of clean audio. We can score it. |
| short or unclear | Very short, or parts were inaudible. Score it carefully. |
| no conversation | Voicemail, ringing, or a greeting and hangup.
Reported as n/a and left out of every average. |
"How well did the rep do?"
About the rep. This is the /10 score.
| 8–10 | Followed the pitch. No action. |
| 5–7 | One or two steps missing. Coaching within a week. |
| 0–4 | Several steps missing. Supervisor listens within 48 hours. |
Before you judge anyone by their score
Short calls cannot score well, no matter who is on them. The pitch has six steps — greet, hook, six discovery questions, pitch the right benefits, capture the requirement, commit to a next step. That takes minutes. In our data, no call under two minutes has ever scored green, and the best a sub-one-minute call has ever managed is 3 out of 10.
So a low score on short calls is a question — why are these calls ending so fast? A weak opener? Bad lead lists? — not a verdict on the rep. Compare reps whose calls are a similar length, not across the board. The overview page shows each rep's average call length right next to their score for exactly this reason.
What each page is for
| Page | Use it when you want to… |
|---|---|
| Overview | See how the whole floor is doing, which pitch step everyone drops, and which campaigns and reps stand out. |
| Excel transcripts | Read what was actually said on any call. |
| Upload & run | Add a new batch of calls from an Excel export, or process a single recording. Runs in the background, and shows live progress plus every past run. |
| Data quality | Check whether the recordings, the CRM tagging and the scoring can all be trusted. Every check runs automatically. |
Words you'll see
| Term | Means |
|---|---|
| Pitch type | Which of the three scripts this call should have followed — Customer-Led (they raised a requirement on the platform), New Registration (they just signed up), or Supply-Led (we called them cold). Different scripts, different scoring. |
| n/a | Deliberately not scored, with a reason recorded. Never the same as scoring zero. |
| Talk ratio | Share of the talking the rep did. Above 0.72 means they talked over the customer. |
| Objection | A concern the customer raised — price, existing supplier, credit. We record whether the rep answered it. An objection raised and not answered is a coaching moment. |
| Evidence | The exact line from the transcript that justified a score. Every step has one. No quote, no pass. |