Train sales reps on their own deals, not on slide decks

Why doesn't quarterly sales training change behavior?

No sales manager gets through forty conversations a day. So the sample ends up arbitrary — usually whatever happened to be loudest. Feedback reaches the rep a week later, when the deal is closed and nobody remembers the details. The notes stay vague — "work on objections" — and repeat month after month, because nobody tracks that this rep already heard it three times. Coaching becomes a ritual: hours spent, behavior unchanged.

What does day-to-day coaching actually look like?

This is the cycle every rep runs through. The timings are fixed and visible in the settings panel.

1. 01

   Right after a conversation — a score

   The call is transcribed and tagged: qualified or not, first conversation or follow-up. Then it gets a score from 0 to 100, with recommendations. A review task lands in the CRM for every first conversation, and for follow-ups only when the score drops below 65.

2. 02

   6:00 pm — the day's active deals

   The day's active deals are reviewed along four axes: rep, client, deal, process. The rep score is 0.6 × conversation quality + 0.4 × process. The deal score lives in its own column and never gets mixed in.

3. 03

   Evening — a private day debrief

   One direct message to the rep: credit for what worked, plus a single growth area with a concrete angle. Not ten notes at once — one step that can actually be taken tomorrow.

4. 04

   Morning — a four-block plan

   One direct message with four blocks: deals that need attention, promises due to clients today, tasks for the day, and recommendations with buttons. The rep opens the day with a list, not with the question of where to start.

5. 05

   Monday 10:10 — the weekly review

   A personal review for each rep. Code does the math, the AI only finds facts, and every conclusion rests on a quote. Time off comes from the shift schedule, so nobody is judged against a week they were not there for.

6. 06

   Next week — the error loop

   The growth area from the review is folded into the following week's analysis. The system watches exactly what was agreed on. You can see whether the habit changed or the same thing keeps happening.

7. 07

   When it matters — a note on the deal

   If the rep axis falls under 50 or a red flag fires, a comment is added to the deal timeline. At most one per deal per day. Card comments are only appended, never overwritten.

How is the score calculated, and can you trust it?

The number does not come from a model's general impression. The rules are fixed and can be checked against any single deal.

Scale

0–100. The outcome sets the band: 90–100, 80–89, 65–79, 40–64, 1–39. Process quality decides the position inside the band.

Who does the math

Code calculates the score against fixed rules. The AI only finds facts and backs them with quotes.

Critical mistakes

They cap the score instead of subtracting points — and only when a literal quote is present. No quote, and the mistake is not counted.

Compensation

A step skipped early but closed later, or closed by the outcome itself, gets full credit. Name asked at the end but the sale happened — the step counts.

Company rules

A separate layer for banned phrasing and weak closings. Each rule is either critical, capping the score at 64, or just a signal.

Who gets the score

Whoever actually spoke. If someone other than the deal owner picked up, the report and the task go to them.

Method behind it

ISO 18295-1, COPC CX, BARS, SPIN, MEDDIC, plus Gong Labs and MIT research on response speed.

What does a rep see in the task after a call?

The task lands in the CRM and is written from the rep's point of view, with no mention of AI anywhere. The structure never changes.

• First line: what to do on this deal right now, while it is still fresh.
• Then the score and the reasoning behind it — what set the band and what moved the position.
• What went well and what went weak, tied to the actual words spoken.
• A recommendation for the next contact, plus a link to the method so the rep can re-check the score.
• A separate "Next step" task carries "why now" and "how to start", always with a deadline. A new one closes the previous one, and moving the deal closes it by itself.

Why don't reps argue with the score?

Coaching gets rejected when it feels unfair. So fairness sits in the rules, not in the wording.

• Every conversation gets a client profile: retail buyer, reseller, broker, service-only, scammer and others — each with a confidence level and quotes.
• Failing to push a client who objectively could not buy is not counted as a rep error. Politeness, though, is scored for everyone, every time.
• Distrust rule: a softening segment counts only at 75% confidence or higher and with a direct quote. Haggling, the word "expensive", asking for a discount, short replies and silence never count as proof that a client cannot pay.
• Scammers are pulled out of the follow-through stats, but the manager still sees them and they get escalated.
• The "budget order" threshold is set by the company and carries its own scoring policy.
• Terminal and won stages, reps marked "do not ping", owners outside the directory and pipelines you have not connected are never analyzed.
• A manager can mark a risk as reviewed — five reasons across two classes. The mark is reversible and fully logged in the audit trail.

What changes in the numbers after rollout?

Measured before and after rollout at active clients, one of them a premium car dealership.

Deals with a prepayment

from 4—7% to 13%

Median first response to a client

from 32 minutes to 6.5

Revenue

+33% in 4 months without adding headcount

Rollout

3—5 days, after a free audit that also takes 3—5 days

Control

33 modules with individual toggles, 85 settings that apply in under 60 seconds

Reviewing conversations is one signal source, not the product. Omnia Lab runs each lead from the moment it appears to the moment it pays: routing by shift, first-response tracking and escalation, promises on both sides, missed calls, and a check on every closure.

What does the system deliberately not do?

Half the design is about knowing where to stop. Otherwise coaching turns into a stream of complaints and people stop reading it.

• It does not create a task for every call — follow-ups are reviewed only below a score of 65.
• It does not post coaching feedback in the team chat. Credit and growth areas go to the rep privately; the manager gets the aggregate picture. Only operational signals reach the team chat — an unanswered client, or a promise more than an hour overdue.
• It does not deduct points for anything that lacks a literal quote.
• It does not bury people in notes: one growth area per daily debrief, one deal comment per day at most.
• It does not punish automatically. Pausing new leads after a conversation scored 50 or lower lasts 2 working hours, and night time does not count toward it. A rep working a shift alone keeps receiving leads. The feature is off by default.
• It does not replace the sales manager. The system brings facts, quotes and priorities; the decision stays with a person.