What coaching effectiveness means
Sales coaching effectiveness is the degree to which a coaching conversation changes what a rep does on the next live call with a buyer, and whether that change is still there three weeks later. It is measured at the live call, against a scorecard the team trusts, with the coaching session as the starting point and the buyer conversation as the test. A session count sits before the thing being measured. Revenue three quarters later sits long after it. A satisfaction score measures how the session felt.
That definition rules out most of what gets reported as coaching measurement. Emails sent and certifications completed move with volume and tooling. Filler words per call move with tenure. None of them tells you whether a coaching conversation on Tuesday changed Thursday's call.
The measurement ladder
Each rung is a different claim, and each costs more to establish than the one below it. Most teams report rung 0 or rung 1 and call it effectiveness.
| Rung | Claim | How it is measured | What it proves |
|---|---|---|---|
| 0 | The session happened | Calendar, platform log | Coaching was delivered |
| 1 | The rep says it helped | Survey, 1:1 comment | The rep liked it |
| 2 | The rep can name what will change | Written at the end of the session, in the rep's words | The rep understood it |
| 3 | The change shows up on the next live call | Scorecard level on that skill, that call, against the pre-coaching level | The coaching transferred once |
| 4 | The change is still there at three weeks | Same skill, same scorecard, calls in week three | The coaching held |
| 5 | Deal evidence moved | Buyer-stated impact and an agreed next step on the deals the rep was coached on | The behavior reached the buyer |
| 6 | Outcomes moved | Stage conversion, win rate, cycle length, over one to three cycles | The business felt it |
Rung 3 is the first one that measures coaching. Rung 4 is the one that matters. Rung 5 can be read on the coached deals within weeks. Rung 6 lags by months and is confounded by territory and pricing, and everything below rung 3 is a report on the session.
Why self-report fails
Rungs 1 and 2 are the rungs most programs stop at, and the evidence against them is strong. In our field study of sales kickoffs, Game Winning Kickoffs, we read 3,582 public posts by people describing their own company's kickoff. 37% said nothing about learning. 25% said they "learned so much" and named nothing. 32% named a topic. 5% named a change or said they were assessed. Across the wider set of about 4,800 first-hand accounts, nine people, by hand count, described using something afterwards. Of 341 posts written weeks or months after the event, five describe using anything from it.
The learning science behind that pattern is old. Across the studies reviewed by Alliger and colleagues in 1997, how much trainees enjoyed a program correlated with learning at .02; how useful they judged it correlated at .26. Skills that go unused for more than a year decay with an effect size of −1.4 (Arthur et al. 1998). What happens back at work is more strongly tied to whether a skill gets used than what happened in the room (Saks & Belcourt 2006; Blume et al. 2010). Corporate Visions put it in one line about kickoffs that applies to any coaching program: "Energy is a great outcome for an SKO, but it's a lousy measurement plan."
A rep saying a session helped is rung 1. The next call is rung 3.
The number to track: drift at three weeks
Performance drift is the gap between the first live call after coaching and the rep's calls three weeks later. A training day produces transfer at best and does nothing about drift. Coaching on live deals, inside the three weeks, is what closes it, and the measurement is the gap itself. We wrote up the distinction between the know-do gap and drift in a separate piece.
To measure it you need three readings of the same skill on the same scorecard: the level before coaching (from the call that prompted the session), the level on the next live call, and the level across calls in week three. The first difference is transfer. The second is drift. A program that shows transfer and no drift is working. A program that shows transfer and full drift is a training program, whatever it is called.
The scorer has to be calibrated first
The ladder depends on the scorecard level at rungs 3 and 4 being a real reading. If the scorer moves on its own, the drift measurement is noise.
In our scoring research, two trained human assessors disagreed with each other on 18% of scores on the same calls. In the same research a redesigned prompt moved up to 40% of scores on a re-run, and each newer model generation marked reps more harshly than the last, and from GPT-5.1 on every model marked below the assessors. The final method reached 0.89 rank agreement with the experts' order, and 1.7% of its scores changed on a re-run.
Judge a scorer on ranking. A scorer can have a low average error and still put calls in the wrong order: across 34 setups, the two measures had almost no relationship (a correlation of 0.02, a separate finding from the Alliger figure above). The full write-up is Can AI judge a sales call?. Before any coaching measurement, run the scorer twice on the same marked calls and look at how many scores moved. If it is more than the managers disagree with each other, fix the scorer first.
Metrics and what each one cannot tell you
| Metric | Rung | What it tells you | What it cannot tell you |
|---|---|---|---|
| Sessions held per rep | 0 | Coaching capacity is being used | Whether any of it landed |
| Rep rating of the session | 1 | The rep's experience | Anything about the next call (enjoyment correlates .02 with learning) |
| Rep's stated change | 2 | The rep heard the point | Whether they did it |
| Scorecard level on next live call | 3 | Transfer happened | Whether it will hold |
| Scorecard level at week three | 4 | Drift, the number that matters | Why it drifted (that is the next coaching call) |
| Buyer-stated impact on coached deals | 5 | The behavior reached the buyer | Whether the deal will close |
| Stage conversion, win rate | 6 | The business result | Which coaching caused it, given everything else that moved |
| Talk ratio, filler words, monologue length | none | Conversation hygiene | Whether a skill improved; a quiet rep who asks nothing scores well |
| Emails sent, certifications completed | none | Activity and compliance | Nothing about coaching |
A metric at rung 0 to 2 is a report on effort. A metric at rung 3 to 5 is a measurement of coaching. Rung 6 is the reason the measurement exists and the worst place to look for it.
A four-week measurement to start on Monday
This needs a scorecard with whole-level descriptors, a scorer that has passed the re-run check above, and one manager.
- Week 0, pick the skill. For each rep, choose one skill from a recent call where they scored at level 2 or below. Record the level and the call.
- Week 0, hold the coaching call. The rep brings that deal and talks it through with the coach. The coach goes to the moment in the call that decided the level, names it, and agrees with the rep what they will do differently when the buyer is back on the line. The rep writes the change in one sentence (rung 2).
- Week 1, score the next live call with that buyer or the next buyer at the same stage, on the same skill. Record the level (rung 3).
- Week 3, score every call the rep had that week on the same skill. Record the median level (rung 4).
- Week 4, read the three numbers per rep. Transfer is week 1 minus week 0. Drift is week 1 minus week 3. Report both for every rep, and the share of reps whose week 3 level is at or above week 1.
A program is working when most reps show transfer and the week 3 level holds. When transfer is there and drift takes it back, the follow-up coaching is missing, and the fix is another call about the next deal, inside the three weeks. When transfer is absent across the team, the coaching itself is the problem, and that is the next section.
Measure the coach too
Coaching effectiveness has two inputs, and most measurement only looks at one. Across 1,246 pharmaceutical sales reps, managers' coaching skill predicted goal attainment, and frequent coaching from less skilled managers was linked to worse results (Dahling et al. 2016). Across 117 studies, skills held or grew when people worked on their own situations and when their managers were trained on the same thing (Taylor, Russ-Eft & Chan 2005).
So the ladder applies to the coach as well. For each manager: how many reps showed transfer at rung 3, and how many held at rung 4. A manager whose reps transfer and drift is running sessions that land and not following up. A manager whose reps do not transfer is running status updates.
This is where a team runs out of managers. One manager with eleven reps cannot hold a coaching call per rep per deal and also measure any of it.
On our platform the rep brings each live deal to one of our AI sales experts on a call, the expert holds the scorecard standard and remembers every previous conversation about that deal, and Bryce, our manager coach, gives the human manager the read across the team: which criteria each rep clears on recorded calls and roleplays, and where the team's lowest bar is. A rep who cannot talk can type. The manager spends their hour on the reps below that bar. To see your team's calls scored this way, talk to us.
FAQ
How do you measure sales coaching effectiveness?
Score the same skill on the same scorecard at three points: the call that prompted the coaching, the rep's next live call, and the rep's calls three weeks later. The first difference is transfer; the second is drift. Coaching is effective when transfer is positive and the week-three level holds. Session counts and rep ratings only measure delivery.
What sales coaching metrics should you track?
The sales coaching metrics to track are the scorecard level on the next live call after coaching, the same level three weeks later, and buyer-stated evidence on the coached deals. Examples are a quantified impact or an agreed next step with a date. These sit close enough to the coaching to attribute. Win rate and cycle length confirm later and are confounded by everything else that moved.
How long does it take to see if sales coaching is working?
It takes about three weeks to see whether sales coaching is working. Transfer shows on the next live call, usually within a week. Whether it held shows in week three. Pipeline and win-rate effects take one to three sales cycles and cannot be cleanly attributed to coaching, so use them to confirm what the week-three reading already showed.
Can rep feedback tell you if coaching worked?
Rep feedback cannot tell you whether coaching worked, because enjoying a session and using it are different things. In our study of 3,582 public posts about sales kickoffs, 25% said they "learned so much" and named nothing. Across about 4,800 first-hand accounts in the same study, only 9 people described using anything afterwards. Across the studies reviewed by Alliger et al. (1997), enjoyment correlated with learning at .02. A rep's rating is a report on the session, and the next call is the test.
Does more sales coaching mean better results?
More sales coaching means better results only when the coach is skilled. Across 1,246 pharma reps, manager coaching skill predicted goal attainment, and frequent coaching from less skilled managers was linked to worse results (Dahling et al. 2016). Measure the coach the same way as the rep: how many of their reps show transfer on the next call, and how many hold at three weeks.
What is performance drift in sales coaching?
Performance drift is the gap between the first live call after coaching and the rep's calls three weeks later. It is measured as the difference between the scorecard level on the first call after coaching and the level in week three. A training day can produce transfer but does nothing about drift; coaching on live deals, inside the three weeks, is what closes it.
Can AI measure sales coaching effectiveness?
AI can measure sales coaching effectiveness once the scorer is calibrated. In our research trained assessors disagreed on 18% of scores; our final AI method reached 0.89 rank agreement with them, and 1.7% of its scores changed on a re-run, which makes drift at three weeks a readable number. An uncalibrated scorer (a redesigned prompt moved up to 40% of scores) turns the same measurement into noise.