Training results: how to measure in 4 levels and 90 days

You measure the results of sales training at four levels, reaction, learning, behavior and results, and the measurement that convinces sales leadership starts before the class, with a baseline pulled from the CRM, behaviors agreed with the managers and, whenever possible, a group that has not yet gone through the program. The evidence that corporate training works exists: a meta-analysis of 397 effects measured in organizations found a medium to large effect at every level, but only 26 of those effects came from business results criteria (Arthur, Bennett, Edens and Bell, 2003). Most evaluations stop before the point the head of sales cares about.

At the close of the program, L&D presents a high satisfaction average, the board asks what changed in the pipeline, and nobody answers because nobody recorded what the pipeline looked like before. In most of the companies I work with, that conversation decides whether the program gets a second wave or becomes a cut in next year’s budget. The design of the program, from diagnosis to leadership reinforcement, is in sales training that changes behavior, and here I deal only with evaluation.

What is the Kirkpatrick model and what does each level measure in a sales operation?

Donald Kirkpatrick proposed the four steps in a series of articles for training directors that began in 1959, one step per article. For the behavior level, he already called for a before-and-after appraisal, made by more than one source (the participant, the superior, the subordinates and the peers), three months or more after the training and with a control group that did not receive it (Kirkpatrick, 1959, 1979 reprint). The current version, maintained by Kirkpatrick Partners, adds confidence and commitment to level 2, speaks of critical behaviors and performance support at level 3 and asks for leading and lagging indicators at level 4 (Kirkpatrick Partners). The same page recommends planning from level 4 down to level 1: you start with the pipeline indicator that needs to move and only then decide what goes into the room.

The ROI Methodology of the ROI Institute, chaired by Jack Phillips, organizes evaluation into five levels, with return on investment as the fifth, and proposes a method to separate the effect of training from other factors and convert it into money (ROI Institute). The table translates the five levels into a sales operation and works as the evaluation plan HR approves with the board.

Level	Question for the operation	What to measure	Instrument and source	When
1. Reaction	Does the content serve this quarter’s deals?	Perceived relevance and intention to apply	Short survey with a question about a real deal	End of each session
2. Learning	Can the person do it in a simulation?	Execution, confidence and commitment	Recorded role-play scored with a rubric	Before the class and at the end of the program
3. Behavior	Does the person do it with the real customer?	Critical behaviors agreed with the managers	CRM fields, a sample of recordings, manager observation	30, 60 and 90 days
4. Results	Did the operation’s indicator change?	Stage conversion, cycle, discount or renewal	CRM reports against the baseline and the comparison group	90 days and after one sales cycle
5. ROI	Did the gain pay for the cost?	Isolated margin benefit over total cost	Phillips calculation with assumptions approved by finance	After the level 4 reading

Reaction and learning: what to ask and how to test

Reaction is the easiest level to measure and the one that says least about the other three. The authors of the 2003 meta-analysis write that there is very little reason to believe that liking a training program tells much about how much people learned, whether they changed what they do on the job or what the company gained (Arthur, Bennett, Edens and Bell, 2003). Keep the survey short and include a relevance question: “in which open deal will you use this next week?”. Blank answers warn the provider at the first session, in time to adjust.

In sales training, you measure learning with the seller performing. Record a role-play with a real customer’s script before the first session, score it with the rubric the manager will use afterwards and repeat it at the last session. The rubric describes what the evaluator observes, such as probing the financial impact before presenting a solution and closing the conversation with a next step and a date. A multiple-choice test on pipeline stages measures memory and predicts little of what happens in the meeting with the customer.

Behavior and results: where most evaluations stop

A meta-analysis of 89 studies on transfer found a positive relationship between applying what was learned and a work environment that supports that application, stronger for open skills, such as leadership development, than for closed skills, such as using a piece of software (Blume, Ford, Baldwin and Huang, 2010). Consultative selling is an open skill: each customer calls for a different combination of questions, and the seller repeats the new behavior when the manager asks about it in the pipeline review and in the 1:1. That is why level 3 enters the plan with an owner in sales leadership.

How do you build the baseline before the training?

The baseline is the snapshot of the team’s pipeline and behavior before the first session, recorded with date, filter and definition. In the programs I follow, it is the step that gets postponed most, because it takes work before the class exists. Without it, any number 90 days later becomes opinion, and each party picks the reading that suits it.

Start with level 4. Choose with the board one or two indicators the program needs to move, such as conversion from discovery to proposal or the discount on the first proposal. With the managers, define two or three critical behaviors that explain that indicator and that someone can observe in the CRM or in a recording. Pull the last two quarters with the same cut of team, segment and deal type, save the reports with name and date and freeze the definitions, because a stage criterion changed in the middle of the program ruins the comparison.

For behavior, draw five to ten discovery meetings per seller from the last quarter and score them with the role-play rubric, using evaluators other than the seller. The transfer meta-analysis showed that when the same source reports predictor and outcome in the same measurement context, the relationships come out inflated (Blume, Ford, Baldwin and Huang, 2010).

Which behavior indicators should you pull from the CRM and from recordings?

A behavior indicator needs to be observable, depend on the seller and connect to the chosen result. “Records pain, impact and decision maker before moving the deal to proposal” works, because the CRM shows the field and the recording shows whether the conversation happened. The exit criteria for each stage of the sales playbook are usually the best source of these behaviors.

From the CRM

The CRM history shows four behaviors with little argument: the share of deals that reach proposal with pain, impact and decision maker recorded; the share of meetings that end with a next activity scheduled the same day; the time between the meeting and the record; and the loss reason filled in with a category and a sentence from the customer.

From meeting recordings

In the sample of discovery meetings, measure the minute at which the seller first mentions the product, how many questions they ask about the impact of the problem and whether the meeting ends with a next step and date agreed with the person who decides. AI transcription helps locate the passages, and the score stays with the trained evaluator.

In CRM and AI training

The structure holds and the behaviors change. In a CRM adoption program, measure meetings recorded within 24 hours, open deals with a next activity and managers’ use of saved lists. In an applied AI program, measure meeting summaries approved without edits, use of the prompt library and incidents, as I detail in AI training for companies.

Comparison group: how do you isolate the effect of training?

Market conditions, seasonality and changes in pricing or commission move the pipeline in the same quarter as the training. The Phillips methodology lists the control group as the first technique to isolate the effect, alongside trend line analysis, forecasting and estimates (ROI Institute). The 2003 meta-analysis excluded studies with a single group measured before and after (Arthur, Bennett, Edens and Bell, 2003), a sign of how much that evidence rests on comparison.

Inside the company, the comparison group comes from the order of the classes. If the program covers two regions or two squads, the second wave starts ninety days after the first and serves as the comparison during that interval. Compare each group’s change against its own baseline, because the teams start from different portfolios and markets. If you cannot split the teams, use the trend of the last four quarters and the managers’ estimate of the share of the gain that came from training, with each manager’s confidence level, and record the assumptions.

Why measure at 90 days?

The window follows Kirkpatrick’s recommendation to appraise behavior three months or more after the training. Filled-in fields and scheduled next activities move in the first weeks, and behavior in the recordings settles between 60 and 90 days if the manager reinforces it. Revenue follows the sales cycle: in a 60-day sale, the 90-day reading already shows conversion and discount; in a six-month sale, it shows only the early stages, and the revenue reading comes after a full cycle.

Training ROI: how to apply the Phillips methodology

The ROI Institute describes six steps: identify the improvement in impact measures, isolate the part that came from the program, convert that part into money, tabulate the fully loaded cost, identify the intangible benefits and compare benefit and cost. The benefit-cost ratio divides benefits by costs, and ROI divides net benefits by costs, as a percentage (ROI Institute). The fully loaded cost includes needs assessment, design, delivery, participants’ hours away from the operation and the evaluation itself.

In a hypothetical example, a team of ten sellers closed R$ 2,000,000 in the 90 days before the training and R$ 2,300,000 in the 90 days after, a 15% increase. The comparison team, which joins the second wave, grew 5% over the same period. The attributable effect is 10 points over the baseline, or R$ 200,000 in revenue, and at a 45% gross margin the benefit reaches R$ 90,000. The fully loaded cost adds up to R$ 60,000: R$ 40,000 in fees, R$ 15,000 in team hours in class and R$ 5,000 in managers’ time and analysis. The benefit-cost ratio is 1.5, and the ROI is 50% (R$ 30,000 in net benefit over R$ 60,000).

With the gross increase of R$ 300,000, without comparison and without margin, the same math reaches 400%, a number finance takes apart in minutes. Team confidence, morale and record quality for the forecast go into the report as intangibles, outside the calculation.

Post-training dashboard: what to track at 30, 60 and 90 days?

The dashboard brings levels 2 to 5 onto one screen, with the baseline alongside, and it is the instrument HR, the board and managers open at the monthly review. You read levels 3 and 4 for the trained group and for the second wave, and the structure below fits consultative selling.

Indicator	Level	Source	Baseline	30 days	60 days	90 days	Owner
Role-play rubric score	2	Recorded role-play	Before the class	Last session			L&D
Proposals with pain, impact and decision maker	3	CRM fields	Two quarters	Reading	Reading	Reading	Sales manager
Meetings with a next activity the same day	3	CRM history	Two quarters	Reading	Reading	Reading	Sales manager
Minute of first product mention	3	Recording sample	Previous quarter		Sample	Sample	Evaluator
Discovery to proposal conversion	4	CRM report	Two quarters	Indicative	Reading	Reading	RevOps
Discount on the first proposal	4	CRM report	Two quarters	Indicative	Reading	Reading	RevOps and finance
ROI and benefit-cost ratio	5	Phillips calculation	Total cost			Estimate	Finance and HR

Read conversion together with the cycle and the volume of proposals. Good discovery training often makes sellers disqualify earlier, and the number of proposals may drop while proposal conversion rises.

How does the CRM support training measurement?

Create a field in the user or team record with the training wave and the class date, so you can filter trained and comparison groups in any report. Stage criteria become required fields at stage change, and the loss reason gets a closed list plus a text field for the customer’s sentence. Duplicate each baseline report for the 30, 60 and 90-day readings without touching the original.

The routine turns the dashboard into management. In the pipeline review, the manager opens the list of deals in proposal with no decision maker and asks about the behavior the training taught, the kind of environmental support the transfer meta-analysis links to application. Once a month, HR, sales leadership and whoever owns RevOps read the dashboard for half an hour and decide what to adjust in the reinforcement.

What are the most common mistakes when measuring training results?

Treating the satisfaction score as proof

The reaction survey shows whether the class found the content relevant and whether the instructor delivered well. Presented as a result, it teaches the board to distrust L&D.

Starting to measure after the class

Without a baseline, HR rebuilds the “before” from memory or pulls a report with different filters, and the difference mixes training with a change in criteria.

Comparing levels instead of change

A region with a more mature portfolio converts more than the other before and after any training. What matters is how much each one changed against its own baseline over the same period.

Asking for revenue at 30 days in a six-month sale

The number does not move, the board ends the program and the new behavior fades before it reaches revenue. Agree in the plan on the timing of each level according to the cycle.

Measuring behavior through self-assessment

A seller asked whether they now probe better tends to say yes. Use the CRM, recordings and manager observation, and keep self-assessment for measuring confidence at level 2.

Counting gross revenue in the ROI

Revenue without margin, without isolation and without the team’s hours produces an ROI nobody can defend in front of finance. Record margin, isolated effect, fully loaded cost and assumptions.

What should you ask the provider before hiring?

A provider who knows how to measure arrives with an evaluation plan alongside the agenda and leaves the instruments with the company. The questions below complement the hiring checklist in sales training that changes behavior and deal only with evaluation.

For HR and L&D: how to run the evaluation

HR recognizes the need when the previous program was evaluated only by satisfaction, when the board questions the training budget with no data to answer or when the same request comes back every year with a different provider. The plan involves sales leadership, who choose the level 4 indicator, the managers, who define and reinforce the behaviors, whoever administers the CRM, who builds the reports, and finance, who approves the ROI assumptions. L&D runs the plan, keeps the baseline and schedules the readings in the order of the dashboard.

30-day roadmap for leadership

1. Choose the results indicator. Sales leadership and HR define one or two level 4 indicators and the reading window according to the sales cycle.
2. Define the critical behaviors. The managers write two or three observable behaviors and the rubric to score them.
3. Pull the baseline. Whoever administers the CRM saves the reports for the last two quarters with frozen filters, and the evaluator scores the sample of recordings.
4. Set up the comparison group. Split the classes into waves by region or squad and record each person’s wave and date in the CRM.
5. Run the initial role-play. Each participant records the role-play with a real customer’s script and gets the rubric score.
6. Take the behavior into the rituals. From the class onward, the manager asks about the critical behaviors in the pipeline review and in the 1:1.
7. Do the first reading and schedule the next ones. On day 30, HR and the board read the leading indicators against the baseline and book the 60 and 90-day readings.

This evaluation plan is part of the in-company consultative selling training: I define with the board and HR the level 4 indicator and the critical behaviors, pull the baseline from the CRM and the recordings, run the practice with rubric-scored role-plays and come back for the 30, 60 and 90-day readings. The team finishes with the consultative selling playbook and the needs discovery framework written on its own deals, and HR keeps the rubric and the dashboard for the next classes.

Frequently asked questions

What is the Kirkpatrick model?

It is a training evaluation model with four levels: reaction, learning, behavior and results. Donald Kirkpatrick published it in a series of articles that began in 1959, and the current version recommends planning from level 4 down to level 1. In a sales operation, you measure level 3 in the CRM and in recordings, and level 4 in pipeline indicators.

How do you calculate the ROI of sales training?

With the Phillips methodology, you isolate the part of the improvement that came from training, convert that part into margin and divide the net benefit by the program’s fully loaded cost. The cost includes fees, team hours in class, managers’ time and the evaluation. Gross revenue without a comparison group inflates the result.

How long should you wait to measure training results?

Filled-in fields and scheduled next activities move in the first weeks. You read behavior between 60 and 90 days, and Kirkpatrick recommended appraising three months or more after the training. Revenue follows the sales cycle, and in a long sale the results reading comes after a full cycle.

Do you need a comparison group?

It is the most reliable way to separate the effect of training from market conditions and seasonality. The practical route is to split the classes into waves and use the second wave as the comparison for 90 days. If that is not possible, use the trend of recent quarters and the managers’ estimate, with the assumptions recorded.

How do you measure the results of CRM or AI training?

With the same level structure and different behaviors. In CRM, measure records within 24 hours, deals with a next activity and managers’ use of lists. In AI, measure summaries approved without edits, use of the prompt library and incidents, always against a baseline.

How do you prepare the team for the training evaluation?

Before the first session, explain what will be measured, why some meetings will be recorded and sampled and that the rubric serves development, with no link to variable pay. Show the team the rubric and the CRM fields that go into the dashboard. With the criteria clear, the initial role-play becomes a starting point and the 90-day reading becomes a 1:1 conversation.