A reliable way to measure sales training is to check whether the trained behavior shows up on reps' live calls weeks later, and whether the business number the training was meant to move changed against a baseline or a comparison group. Satisfaction surveys and quiz scores are easy to collect and say little about what changed on deals.
Decide the measure before the training
Many teams measure what is easy, and our sales kickoff field study found that enjoying a training session correlates .02 with what people learn from it, and rating it useful correlates .26 (Alliger et al. 1997).
Proof of business effect is rarer still. In the same field study we found 21 studies, surveys and case studies claiming a kickoff had an effect, and none compares people who attended with people who did not. The one formal ROI calculation we found, a 2005 franchise event, came out at minus 53% on a small sample. Only 24% of event planners have ROI metrics in policy (Amex GBT 2026 forecast).
The fix is to decide the measure before the training. Name the behavior, name the number, record a baseline, and plan to listen to real calls three or more weeks later. That window is where performance drift shows up: the gap between what a rep learned and what they do on a live call.
Four levels, and which ones count
The Kirkpatrick model is the standard frame. Kirkpatrick Partners advise starting at Level 4 and working back.
| Level | What it measures | Practical measure | Weight as evidence |
|---|---|---|---|
| 1 Reaction | Whether the session felt useful | Pulse survey | Low: enjoyment correlates .02 with learning |
| 2 Learning | Knowledge and skill at the end | Pre and post test | Medium: shows capacity under test conditions |
| 3 Behavior | Whether reps do it on real calls | Sample of live calls scored against the standard | High: the link between training and results |
| 4 Results | The business number | Win rate, deal size, cycle length vs baseline or a comparison group | Highest, when there is a comparison |
Level 3 is where most measurement breaks, because managers score calls differently. In our call-scoring research, trained experts disagreed with each other on 18% of scores. Across more than 250,000 scoring decisions, the final AI setup reached 0.89 rank agreement with the experts' order. Consistent scoring is what makes a level-3 number trustworthy.
Where the Kirkpatrick model falls short
The Kirkpatrick model tells you what to measure and leaves out how to separate the training's effect from everything else that happened in the quarter. Level 4 asks whether the business number moved. It has no built-in answer to a price cut, a new territory or a strong market landing in the same months, which is why the comparison group in this guide carries so much weight.
Two other gaps show up in sales teams. The levels read as a ladder, so teams collect levels 1 and 2 first, run out of time, and never reach 3. Starting from Level 4 fixes the order. The model also says little about the weeks after training, when managers either reinforce the behavior on live deals or let it fade. Level 3 records the outcome of those weeks, and someone has to plan them before the training starts.
Lead measures and lag measures
The four levels arrive at different speeds. Behavior on calls moves within weeks; win rate and deal size move over a full sales cycle. Split the measures so the team has something to read before the quarter ends.
| Type | Measure | When it moves | What it tells you |
|---|---|---|---|
| Lead | Share of sampled calls where the behavior appears, scored against the standard | Weeks 3 to 6 | Whether reps use the skill on the job |
| Lead | Share of deals in the CRM with the evidence the behavior should produce, such as a quantified problem | Weeks 4 to 8 | Whether the skill shows up in deal records |
| Lead | Manager notes from deal reviews on whether the skill appeared | Weekly | Whether managers see it and reinforce it |
| Lag | Win rate on deals that reached the stage the training targets | One sales cycle | Whether more of those deals close |
| Lag | Deal size or discount level | One sales cycle | Whether reps hold value |
| Lag | Cycle length from the targeted stage to close | One sales cycle | Whether deals move faster after the step |
A lead measure that moves while the lag measure stays flat says the behavior changed and did not pay off, or has not yet. A lag measure that moves while the lead measure stays flat usually means something other than the training moved it: a price change, a new segment or a strong market. Read the two together.
A measurement plan, step by step
- Name one behavior. Write it as something a listener can hear on a call: "the rep asks the buyer to put a cost on the problem". A training that covers ten behaviors gets measured on the one or two it most needs to change.
- Write the standard. Describe what weak, adequate and strong look like for that behavior, in the team's own method. This is the scorecard every scorer uses, before and after.
- Name the business number. Pick the lag measure the behavior should move, and the stage of the pipeline where it should show.
- Record a baseline. Score a sample of calls from the weeks before the training against the standard, and pull the business number for the previous quarter or two.
- Set up the comparison. Decide which reps or teams will be the comparison group (see the next section), and freeze that choice before the training starts.
- Plan the reinforcement. Decide who checks the behavior on live deals in the weeks after. Without it, the measure at week six mostly records how fast the skill faded.
- Score again at three and six weeks. Use the same standard and, where possible, the same scorers, or a scorer you have checked against them.
- Read the business number at the end of the cycle. Compare the trained group with its baseline and with the comparison group, and write down what else changed in the period.
- Report both levels. Show leaders the behavior change and the business change side by side, with the comparison. A single ROI figure with no comparison behind it invites a fair challenge.
Finding a comparison group
Most teams cannot hold half the sales force back from training. Four workable options:
- Stagger the rollout. Train one team or region this quarter and the next a quarter later. The second group is the comparison until it is trained, which is the setup in the example below.
- Match reps. Pair trained reps with untrained reps of similar tenure, territory and pipeline, and compare the pairs.
- Use the trend. If everyone is trained at once, compare the trained period with the same reps' trend over the quarters before. It is weaker evidence, because the market moves too, so say so in the report.
- Compare stages. If the training targets discovery, compare deals that entered discovery after the training with deals that entered before it, on the same reps.
Whichever you pick, write it down before the training and do not swap it afterwards for whichever group makes the result look best.
Measuring one discovery training (illustrative)
This example is illustrative. A team trains 20 reps on one behavior: getting the buyer to put a cost on the problem during discovery. Before the training, enablement scores a sample of recent discovery calls against that behavior to set a baseline. A second team, trained a quarter later, is the comparison group.
After the training, each rep brings a live deal to a coached deal review on a call, with their manager or an AI sales expert, and the behavior is checked on that deal. Three and six weeks later, enablement scores a new sample of real discovery calls with the same scorecard. At the end of the quarter, it compares the share of deals with a quantified problem and the win rate between the two teams.
The one-page plan for this training reads:
| Item | Entry |
|---|---|
| Behavior | Rep asks the buyer to put a cost on the problem during discovery |
| Standard | Weak: no cost asked. Adequate: rep offers an estimate. Strong: buyer states their own figure |
| Business number | Win rate on deals that pass discovery |
| Baseline | Sample of discovery calls from the six weeks before, scored on the standard |
| Comparison | The team trained next quarter |
| Reinforcement | Coached deal review on a live deal for each rep, weekly |
| Measures | Calls scored at three and six weeks; win rate at quarter end |
| Other changes to note | Pricing, territory or headcount changes in the period |
Reporting the result to leadership
Report a sales training result on one page, with the behavior change and the business change side by side and the comparison next to each. Leaders tend to ask the same three questions: did reps change, did the number move, and how sure are we the training did it. A layout that answers them in that order looks like this:
| Line | Trained group | Comparison group |
|---|---|---|
| Behavior on scored calls, baseline | ||
| Behavior on scored calls, week 6 | ||
| Business number, previous quarter | ||
| Business number, this quarter | ||
| Calls scored per rep | ||
| Other changes in the period |
Add two sentences under the table: what you conclude, and what would change your mind. If the comparison group improved as much as the trained one, say so plainly. That result is still useful, because it stops the team paying for the same training again on the strength of a number the market produced.
When the result is mixed, lead with the lead measure. A behavior that moved on calls while win rate stayed flat is a reason to wait one more cycle, and the report should say which cycle and what you will check then.
Common mistakes
Measuring the day after. A post-test on the last afternoon shows what reps can do in a classroom. The question is what they do on a live call weeks later, when nobody is watching and the deal is messy.
Changing the scorecard halfway. If the standard changes between the baseline and the follow-up, the two numbers measure different things. Fix the wording before the baseline and leave it alone.
Letting each manager score their own team. Managers score differently, as the 18% expert disagreement above shows. Use one scorer or a checked scoring setup across every team, including the comparison group.
Counting completion as success. A course finished by every rep is an activity number. It belongs in the report as context, below the behavior and business measures.
Sampling too few calls. A handful of calls per rep can swing on one good or bad meeting. Score enough calls per rep that one outlier does not decide the result, and report the sample size.
Related reading
The sales kickoff field study has the evidence on measurement and ROI. The difference between what training and coaching each move is in sales training vs sales coaching, and how to reinforce sales training covers the weeks after. If level 3 keeps stalling because nobody has time to review each rep's deal after the training, see how our AI sales experts work with enablement teams.
FAQ
What metrics show sales training is working? Track how often the trained behavior appears on real calls three or more weeks later, scored against a written standard. Pair it with the business number the training targeted, such as win rate or deal size, compared with a baseline or an untrained group. Completion rates and satisfaction scores only tell you about activity and mood.
How long after sales training should you measure? Measure behavior at three weeks and again at six or more, and measure results at the end of the quarter or the sales cycle. Skills that are not used fade, so a late measure shows whether the training reached the job.
Is the Kirkpatrick model good for sales training? Yes, as a frame, if teams use all four levels and start from the result they want. Kirkpatrick Partners recommend starting with Level 4. Many sales teams stop at reaction and learning, which are quick to collect and say little about change on deals.
Can you calculate ROI on sales training? You can, but it needs a comparison group or a credible baseline, and few teams set one up. In our field study, none of 21 claims about kickoff effects compared attendees with non-attendees. Decide the business measure and the comparison before the training, or the ROI number will rest on assumptions.
What is the difference between lead and lag measures of sales training? Lead measures move within weeks and show whether reps changed what they do, such as how often a trained behavior appears on scored calls. Lag measures move over a sales cycle and show whether the business changed, such as win rate or deal size. Track both, because a lag number alone cannot say whether the training caused it.
Is a post-training satisfaction survey worth running? A short survey is worth running for feedback on the session itself: pace, relevance, what to cut. Do not read it as evidence of learning. In the research cited above, enjoying a session correlated .02 with what people learned, and rating it useful correlated .26 (Alliger et al. 1997).
How do you measure sales training without a control group? Stagger the rollout so one team waits a quarter, match trained reps with similar untrained reps, or compare the same reps against their own trend from earlier quarters. The last option is the weakest, because the market changes too, so note what else happened in the period.
Who should score the calls after sales training? Use the same scorer or scoring setup for the baseline, the follow-up and the comparison group, working from one written standard. Managers scoring only their own teams will drift apart, which makes the change hard to read.