How to measure training effectiveness: 7 steps, metrics, and examples

What training effectiveness means

Training effectiveness is the degree to which a program produces the change it was built for: new knowledge or skill, different behavior at work, and a better business result. Training efficiency is a separate question about resources, such as cost per learner and time to roll out. A course can be cheap and fast to deliver and still change nothing.

Attendance and activity data sit below both. Enrollments and completions tell you the rollout reached people. Will Thalheimer's Learning-Transfer Evaluation Model (LTEM) puts attendance and activity in its two lowest tiers because neither one shows that anyone learned. Track them to manage the rollout, and do not report them as proof.

The four levels you measure

Most training measurement uses the four levels Donald Kirkpatrick introduced in the 1950s. Kirkpatrick Partners, which maintains the model, defines them roughly as follows (checked October 2026):

1. Reaction: learners find the training favorable, engaging, and relevant to their work.
2. Learning: learners gain the intended knowledge, skills, attitude, confidence, and commitment.
3. Behavior: learners perform the critical behaviors on the job, with support and accountability.
4. Results: the targeted organizational outcomes occur.

Each level needs a different kind of record. A quiz score is Level 2 evidence. It cannot show Level 3, however high it is. In the demo below, file each piece of evidence from a software rollout at the level it supports: a scenario score of 8 out of 10, a workflow used in 7 observed tasks, or errors that fell from 18 to 11 per 1,000 tasks.

Kirkpatrick Partners updated the model in 2026. Vanessa Milara Alzate added the performance environment: the conditions around learners that support or block progress at every level. The update also describes Return on Expectations (success as stakeholders define it at Level 4) and contributive ROI, which credits training with a share of a result instead of trying to prove it was the only cause.

You will also see three other models:

• The Phillips ROI Methodology from the ROI Institute adds a fifth level, return on investment, and a step that isolates the program's effect from other factors.
• Robert Brinkerhoff's Success Case Method surveys participants to find the few who got the most and least out of the training, then interviews both groups to learn what helped or blocked them.
• LTEM uses eight tiers, from attendance up to the effects of transfer on the organization, and treats realistic decision-making and task performance as stronger evidence than recall.

For a side-by-side comparison of these models, see our guide to training evaluation methods.

How to measure training effectiveness in 7 steps

This measurement strategy works for a single course or a full program. The order matters: define the result first, then design every measure backward from it. Kirkpatrick Partners gives the same advice, starting at Level 4 and planning down.

1. Agree on the business result

Meet the person who owns the result before you build anything. Ask which metric should move if the training works, by how much, by when, and what else will affect that metric in the same period. Then write the objective so anyone can check it later:

• Vague: "Improve sales team communication." Measurable: "Cut the average sales cycle for Product X from 45 days to 38 days within two quarters of the objection-handling course."
• Vague: "Make new managers better leaders." Measurable: "Lower first-year voluntary turnover on new managers' teams from 30% to 20% within 12 months."

If no business metric fits, the problem may not be a skills gap. A training needs assessment checks that before you spend on content.

2. Name the critical behaviors

List the few things people must do differently at work for the result to move. Write each one so a manager could see it happen: "asks about the buyer's timeline before the demo," "logs a root cause on every escalated ticket," "locks out the machine before clearing a jam." This list becomes your Level 3 checklist, and it tells you what the Level 2 assessment has to test.

3. Record a baseline

Measure before training starts. For learning, give a short pre-test that uses the same items, or parallel items, as the post-test. For behavior, have managers score a sample of real work with the checklist. For results, pull the metric for several months before launch so you can see its normal trend and any seasonal pattern.

4. Choose a small set of metrics at each level

Pick measures tied to the result and behaviors you wrote down. The metrics table after these steps lists common options and where the data lives. A clear chain from one learning measure to one behavior to one result makes a stronger case than a long list of metrics.

5. Plan a comparison

A before-and-after change can come from a new product or a busy season. Build a comparison into the rollout:

• Train some teams, stores, or regions first, and compare them with similar groups that train later. A staggered rollout gives you a comparison group without withholding training from anyone.
• Compare the trained group with its own trend line from before training.
• When neither is possible, ask participants and their managers how much of the change came from training and how confident they are in that estimate, then discount the estimate by that confidence. Say in the report that you used estimates.

6. Collect data on a schedule

• During training: in-lesson questions and quick polls (Levels 1 and 2).
• At the end: a short reaction survey and the post-test.
• About 30 days later: a short knowledge check to see what learners kept.
• After people have had time to use the skill, often 30 to 90 days depending on how often the task comes up: the behavior checklist, system data, and a follow-up survey.
• Quarterly: the business result against the baseline and the comparison group.

7. Report the chain and fix the program

Report each level and say plainly what the data supports. Then use the gaps to improve the training, because the pattern across levels usually points to the cause.

Common measurement patterns and what to change

What the data shows	Likely cause	What to change
High satisfaction, little use on the job	Content did not match real work, or nobody reinforced it	Add job aids, manager follow-up, and practice on real cases
Most learners pass the quiz, but errors stay high	Learners can recall facts but cannot apply them under pressure	Replace recall questions with scenario practice
Many learners miss the same question	That part of the content is unclear	Rewrite or re-record that segment
One team improves far more than the others	Its manager reinforces the skill	Find out what that manager does and ask other managers to do it
Managers see change, but learners report low confidence	Learners know what to do but need more practice	Add coached practice or peer review

Training effectiveness metrics at each level

Use this table to choose metrics to measure training effectiveness. Pick from each row, and start with the rows closest to the business result.

Training effectiveness metrics by level, with data sources and timing

Level	Question it answers	Example metrics	Data source	When
Activity (not a level)	Did people take part?	Enrollment, completion rate, time in course, share of each video watched	LMS, video platform	Ongoing
1. Reaction	Did learners find it relevant and usable?	Relevance rating, confidence rating, intent to apply	Short survey, in-lesson poll	During and right after
2. Learning	Did they gain the knowledge or skill?	Pre-to-post score gain, first-attempt correctness by question, scenario decision accuracy, scored skill demonstration	Scored quizzes, scenario questions, recorded demonstrations	Before, at the end, and about 30 days later
3. Behavior	Do they do it at work?	Checklist observation rate, process adoption, quality assurance score, 360-degree feedback	Manager observation, QA reviews, CRM and help desk logs	30 to 90 days after
4. Results	Did the business metric move?	Sales cycle, win rate, error rate, incident rate, CSAT, turnover, time to proficiency	CRM, help desk, quality, HR, and ERP systems	Quarterly, against baseline and comparison group
5. ROI (Phillips)	Was the result worth the cost?	ROI percentage, benefit-cost ratio	Level 4 results converted to money, finance cost data	After Level 4 data is in

Why reaction and completion data are not enough

Satisfaction scores are the easiest data to collect, so many programs stop there. A 2008 meta-analysis by Sitzmann and colleagues in the Journal of Applied Psychology found that trainee reactions were only modestly related to knowledge right after training and unrelated to knowledge measured later. Reactions did predict gains in motivation and self-efficacy. Keep Level 1 to improve the course and track confidence, and do not present it as evidence of learning.

Survey data shows how often teams stop early. In ATD's 2019 Effective Evaluation study of 779 talent development professionals, 54% of organizations measured behavior (Level 3), 38% measured results (Level 4), and 16% calculated ROI. In ATD's April 2025 report with Zensai, 4% of the 277 talent development professionals surveyed rated their organization excellent at using learning program data to make business decisions.

How to measure learning inside training videos

A completion flag on a video module says little. A learner can let the module play in another tab and still count as complete. Scored questions placed inside the video, at the point where each concept is taught, turn the module into a Level 2 measure.

In Mindstamp, a question pauses the video and can require an answer before playback continues. The records you get for measurement:

• Every answer, with a time stamp, stored against the learner's email or employee ID.
• An answer breakdown for each question, which shows the concept most learners miss.
• A score for each view: correct answers divided by all answers to scored questions. With Repeat Until Correct turned on, every attempt counts, so a miss followed by a correct retry still lowers the score and the first miss stays visible.
• Drop-off by section, which shows where learners stop watching.
• A completion report with one row per learner and a score column for each video, exported as CSV.
• A SCORM package that reports completion and score to your LMS, on supported plans.

To join these records with performance data later, identify each learner the same way everywhere. Our guide to tracking a training course shows how to pass an email or employee ID through lead capture, a personalized link, or a URL parameter, and the training records export explains which columns each report includes.

Recall questions are weak evidence for skills. Two formats give you stronger Level 2 data:

• A branching scenario shows a realistic situation, lets the learner choose, and plays the consequence. The report records the path each learner took.
• A video response question asks the learner to record a pitch or walk through a procedure on camera. The recording is attached to the learner, so a manager can score it with the same checklist used on the job.

For more ways to design these checks, see our learning assessment strategies.

How to measure behavior change on the job

Fewer organizations measure behavior than reaction or learning, yet behavior is the link that explains whether results moved. Use more than one source:

• Manager observation with the checklist from step 2, scored on real work such as call recordings, ticket reviews, or floor walks.
• System data that shows the behavior: CRM fields completed, a new workflow used, tickets with a root cause logged.
• Quality assurance scores before and after, for support, manufacturing, or clinical roles.
• 360-degree feedback for leadership and communication skills, collected before training and again a few months later.
• A short follow-up survey that asks about specific use.

Good follow-up questions ask for a specific event:

• "Describe one situation in the last two weeks where you used a technique from the training."
• "What has made it hardest to apply what you learned?"
• "On a scale of 1 to 5, how confident are you using the new process for your core tasks?"

A survey tool such as Google Forms or SurveyMonkey is enough for these surveys. Send them on a schedule so you see whether the behavior lasts. A small focus group a month after training adds the reasons behind the numbers.

When behavior does not change, look at the performance environment before you rewrite the course. People who learned the skill may lack the time, the tools, or a manager who expects it.

How to connect training to business results

Choose results that the critical behaviors would plausibly move, and agree on them before launch.

Example business metrics by training program

Training program	Business metrics to track
Consultative sales training	Sales cycle length, win rate, average deal size
Customer service de-escalation	CSAT, escalation rate, first-contact resolution
Safety training	Recordable incident rate, near-miss reports, audit findings
Process or quality training	Defect rate, cycle time, rework cost
Leadership development	Team engagement scores, voluntary turnover, internal promotion rate
New software rollout	Task time, error rate, support tickets about the tool

These numbers sit outside the learning stack: in the CRM, the help desk, the HR system, the quality system, or an ERP (enterprise resource planning) system that holds operations and finance data. HR teams already define workforce metrics such as turnover rate, retention rate, and revenue per employee. Use their definitions so your numbers match the ones leadership already sees. Then join the data by employee ID, which is why consistent learner identity matters.

Worked example: separating the training effect

Say one sales region completes objection-handling training in March and a similar region trains in July. From April to June, the trained reps' average deal size rises by $5,000 and the untrained reps' average rises by $500. The market lifted both groups, so the most you can credit to training is the $4,500 difference per deal. Report that figure with what else differed between the regions, such as product mix or headcount changes, and measure only opportunities created after training, because a deal that closes in April may have started in January. These figures are illustrative.

When to calculate training ROI

ROI is the fifth level in the Phillips ROI Methodology. The ROI Institute's formula is: ROI (%) = (program benefits minus program costs) divided by program costs, multiplied by 100. The benefit-cost ratio is program benefits divided by program costs. An ROI above 0% means the monetary benefits exceeded the costs.

The hard part is the inputs. Benefits come from your Level 4 results, after you isolate the training effect and convert it to money. Costs should be fully loaded:

• Design and development time
• Platform, content, and license fees
• Facilitator and subject matter expert time
• Participants' time away from their work
• The cost of the evaluation itself

Save full ROI studies for programs that are expensive or highly visible. In the demo below, pick one of three illustrative cases: $60,000 in benefits on $20,000 in costs, $30,000 on $30,000, or $18,000 on $24,000. Their ROIs are 200%, 0%, and -25%.

For cost categories, monetizing benefits, and worked cases, read our guide to measuring the ROI on training, or run your own numbers in the training ROI calculator.

How to measure customer training effectiveness

Customer training uses the same levels, but the results sit in product and account data. Compare trained and untrained customers, or the same accounts before and after training, on measures like these:

• Adoption of the features the training teaches
• Time to first value, from signup to the first meaningful outcome
• Support tickets per account on topics the training covers
• Renewal and expansion rates
• Pass rates on admin or partner certifications

Customers who choose to take training are often more engaged to begin with, so a raw gap between trained and untrained accounts overstates the effect. Compare accounts of similar size and starting usage, or compare each account with its own trend before training. For program design, see our customer education page and our guide to customer onboarding best practices.

How to measure retail training effectiveness

Retail training is easiest to measure at store level, because stores make natural comparison groups. Train a set of pilot stores first and compare them with matched stores of similar sales volume and format that train later. Store metrics to watch:

• Conversion rate (transactions divided by store visitors)
• Average transaction value and units per transaction
• Mystery shop or service audit scores
• Shrink and compliance audit results
• Time to proficiency and 90-day turnover for new hires

For short, frequent store tasks such as a new returns process or a safety check, pair a short in-video knowledge check with a manager's sign-off on the floor. That gives you Level 2 and Level 3 evidence for each associate.

How does Walmart measure training effectiveness?

Walmart has not published a full evaluation framework, but its public statements about virtual reality (VR) training show what it tracked. In September 2018, Walmart's corporate news site reported that VR training at its Academies improved test scores by 10 to 15 percent and boosted confidence and retention, and that associates who watched others use the training saw the same retention gains. In October 2018, a blog post about Walmart's Oculus Go rollout, now on Meta's site, reported 30% higher training satisfaction with the VR modules and said 70% of employees who trained in VR outperformed groups trained with other materials. Those are reaction and learning measures with a comparison group. Neither post reports store-level results such as sales or shrink, which would be Level 4.

How to measure sales training effectiveness

Sales results are fast and countable, which makes it tempting to credit training for every good quarter. Measure the whole chain instead:

• Learning: scenario scores on discovery and objection handling, and a recorded pitch scored with a rubric.
• Behavior: call review scorecards from recordings, discovery questions asked per call, and next steps logged in the CRM.
• Results: win rate, stage-to-stage conversion, sales cycle length, average deal size, quota attainment, and ramp time for new reps.

Compare cohorts trained in different months, and compare each rep with their own results from the quarters before training. For program ideas, see our sales training best practices and the sales enablement page.

Common questions

How do you evaluate training effectiveness once you have the data?

Measuring collects the numbers. Evaluating compares them with the objective you set in step 1 and leads to a decision: keep the program, change it, or stop it. Write the decision and the evidence behind it in the same report, so the next program starts from what you learned.

How do you measure soft skills training?

Measure behavior instead of recall. Use scenario questions in which the learner chooses how to respond, recorded role-plays scored with a rubric, and 360-degree feedback before and after training. Then link the behaviors to a business metric they influence, such as team turnover or customer complaints.

What is a good ROI for training?

No single benchmark fits every program. Any ROI above 0% means the benefits exceeded the fully loaded costs. Agree on a target with finance before the program starts, and report ROI as a range along with the method you used to isolate the training effect.

Start with one program

Pick one program that launches next quarter and has a clear business owner. Fill in this plan with that owner before the content is final, and use the employee training tracker guide to keep the completion data you will need.

One-page training measurement plan, with an example for a support triage course

Item	Example
Business result	Escalation rate on tier 1 tickets falls from 22% to 15% by the end of Q2
Critical behaviors	Agents check the triage checklist before escalating and log a root cause on every escalation
Baseline	Escalation rate for the six months before launch, pre-test score, checklist audit of 20 tickets per agent
Learning measure	First-attempt correctness on triage scenarios inside the training video
Behavior measure	Share of escalated tickets with a logged root cause, from the help desk, at 30 and 60 days
Comparison	Day shift trains in April; night shift trains in June
Review dates	End of course, day 30, day 60, end of Q2

Sources and further reading

• InStride publishes its own guide on how to measure training effectiveness.
• Devlin Peck keeps a sourced list of employee training statistics on training investment, engagement, and retention, last updated in July 2026.
• eLearning Industry tracks spending and trend data in its employee training statistics page, updated in May 2026.