Linking assessment scores to training ROI for your clients
Training ROI is the net money value of a program’s benefits divided by its fully loaded cost, shown as a percentage. Assessment scores cannot produce that figure on their own. They show that learners learned something, and your client’s business data has to show that the learning changed results. This guide shows training providers how to design the assessment and the data agreement so a client’s ROI study becomes possible.
It is written for providers who deliver courses to corporate clients and get asked, “What was the return?” Most pages ranking for this topic assume an in-house L&D team that already owns the performance data. You usually do not, and that changes what you ask for before the course begins.
Every number in the worked example is invented to show the arithmetic. It is not a benchmark and not financial advice.
Key takeaways
- Training ROI needs benefits and costs in money, so assessment scores support the case at the learning level but rarely appear as a number in the formula.
- Most organizations stop short of measuring results: in CIPD’s 2021 survey, 8% evaluated the wider impact of most learning initiatives on the business or society.
- Agree the business measure, its dollar value and a comparison group with the client before the first session, because that data lives with the client.
- Report ROI as a range with a break-even point, since one conservative attribution assumption can move the result from negative to strongly positive.
What is training ROI, and where do assessment scores fit?
Training ROI is the fifth level in the Phillips methodology, which compares program benefits with program costs. Assessment scores sit at Level 2, learning. They are early evidence that the chain could work, not proof that it did, because Levels 3 to 5 use data the client holds.
The ROI Institute defines the five levels and gives the formulas as BCR = program benefits divided by costs and ROI = (benefits minus costs) divided by costs, times 100. The table adds what an assessment contributes at each level and who usually holds the data.
| Phillips level | What it measures (ROI Institute wording) | What your assessment adds | Who usually holds the data |
|---|---|---|---|
| 1. Reaction and planned action | “participant satisfaction” and planned actions | Nothing beyond a survey | You |
| 2. Learning and confidence | “changes in knowledge, skills, and attitudes” | Pre-test, post-test and retest scores by objective | You |
| 3. Application and implementation | “changes in behavior or actions” | Observation checklists or scenario items, if designed in | Client managers |
| 4. Business impact | “changes in business impact variables” | Nothing directly; scores help check that groups started alike | Client operations or finance |
| 5. Return on investment | “compares project benefits to the costs” | Nothing directly | Client finance, plus your cost records |
Kirkpatrick’s four-level model is the older frame. Kirkpatrick Partners defines Level 4, results, as “the degree to which targeted organizational outcomes occur as a result of an experience” plus performance support. Its four levels do not include a separate ROI level. Phillips added one, so “Level 4” and “ROI” are not the same thing. For what scores can show at Level 2, see our guide to assessment reports corporate clients want.
How many organizations measure results or ROI at all?
Few do. CIPD’s Learning and Skills at Work 2021 survey of 327 learning professionals found that 76% evaluated their learning initiatives in some way, but the most common method was participant satisfaction. Only 8% evaluated wider business or society impact, and 13% assessed behavior change on the job.
That survey is UK-led and a few years old, so treat it as a signal, not a global rate. Its use for you is practical: your client contact may never have run a Level 4 study. If so, you can lead, and the CIPD report is a neutral source to cite when you explain why.
The ROI Institute’s own targets are modest too. As of its 2022 slides, it recommends evaluating about 10% of programs at business impact and about 5% at ROI each year. Offer a full ROI study for the programs that carry the client’s largest budget, not for every course.
Can assessment scores predict business results?
Only if someone tests that link. A score predicts a work outcome when data show it does for that job and context. The Standards for Educational and Psychological Testing call this a testable hypothesis, so a quiz on terminology should not be sold as a forecast of performance.
The Standards (AERA, APA and NCME, 2014) say the “fundamental question is always, how accurately do test scores predict criterion performance?” and that whether a test predicts a given criterion in a given context is a testable hypothesis. For a provider, the criterion is the client’s performance measure, such as escalations or rework.
Three design choices make a later link more plausible:
- Write items from the job tasks the client wants changed, using scenarios instead of recall, as in our guide to scenario-based questions for professional training courses.
- Tag every item to a learning objective, so you can say which objective moved and which business measure it should affect.
- Keep learner IDs consistent across the pre-test, post-test and retest, so scores can be matched to client data when the agreement allows it.
Be careful with small cohorts. With 30 or 60 learners, a score-to-performance correlation is unstable, so describe it as an observation, not a validation.
What should you agree with the client before the course starts?
Agree the business measure, its dollar value, a comparison group and the data owner in writing before delivery. Kirkpatrick Partners calls the stakeholder’s definition of success Return on Expectations and says it is established in advance. Without these four items, an ROI study after the fact is guesswork.
The table is a data request you can send with the proposal. The examples are invented.
| Ask | Why you need it | Example (invented) |
|---|---|---|
| One business measure the client wants to move | The ROI formula needs a benefit in money, and that starts with one measure | “Escalated complaints per month” |
| Six or more months of history for that measure | A baseline and a trend show what would have happened anyway | Monthly counts per team, last 12 months |
| The client finance team’s value per unit | You cannot defend a dollar figure you invented | “$240 fully loaded handling cost per escalation” |
| A comparison group or a staggered cohort schedule | It separates the course’s effect from seasonality and other changes | A similar team trained 4 months later |
| Data owner, extract dates and matching rules | Results arrive late if nobody is named to pull them | Ops analyst, monthly extract, matched by employee ID under a data-sharing clause |
| The hurdle the client will accept | It turns the result into a decision instead of a debate | “Break even in year one” |
Matching employee IDs across systems is personal data processing, so involve the client’s privacy contact early. Our guide to selling assessment as a service to corporate clients covers the contract side.
How do you separate the training’s effect from everything else?
Use at least one isolation method and say which. The ROI Institute’s guiding principles require it, along with conservative choices and estimates adjusted for error. A comparison group is usually the strongest option a provider can arrange, and manager or participant estimates are the weakest.
The ROI Institute lists methods including mathematical modeling, control groups, trendline analysis, and estimates from participants, managers and senior managers. The table compares four a provider can realistically use. Strengths and weaknesses are my reading of how each works, not ROI Institute findings.
| Method | What you need from the client | Where assessment data helps | Main weakness |
|---|---|---|---|
| Control or comparison group | Outcome data for trained and untrained staff over the same months | Pre-test averages show whether the groups started alike | Groups rarely match perfectly |
| Trendline analysis | Many months of pre-course outcome data | Little beyond learning evidence | Assumes the old trend would have continued |
| Manager estimate | A short form asking what share of the change came from training, and how confident they are | Objective-level results give managers something concrete to react to | Opinion, and prone to optimism |
| Participant estimate | The same form, completed by learners | Same as above | Learners may credit a course they enjoyed |
Estimate-based methods should be adjusted for how sure the estimator is. For example, if a manager says training caused 60% of an improvement and is 80% confident, count 48%. The ROI Institute’s session slides show estimates multiplied by confidence percentages in this way.
Pro tip: Stagger Cohorts On Purpose. If the client trains 120 people, split them into two cohorts four months apart and hold the later cohort as the comparison group. Their pre-tests double as a check that both groups began at the same level, and nobody is permanently denied the course.
What does a worked training ROI calculation look like?
A worked training ROI calculation lists fully loaded costs, converts the attributed change in one business measure to money, and applies the formula. The example below uses invented numbers for a complaint-handling course taught to 60 agents, with a comparison group of 60 agents trained later.
The ROI Institute’s principles call for fully loaded costs, so participant time counts. Costs come first:
| Cost item | Calculation | Amount |
|---|---|---|
| Course fees | Agreed price | $30,000 |
| Participant time | 60 agents x 16 hours x $35 per hour | $33,600 |
| Venue, materials and admin | Client and provider records | $4,400 |
| Total fully loaded cost | Sum | $68,000 |
Now the benefit. Escalations in the trained group fell from 200 to 145 per month, a drop of 55. The comparison group fell from 200 to 175, a drop of 25, so 30 fewer escalations per month are attributed to the course. At the client’s $240 per escalation, the annual benefit is 30 x 12 x $240 = $86,400.
Net benefit is $86,400 minus $68,000, or $18,400. ROI is $18,400 divided by $68,000, times 100, which is 27.1%. The benefit-cost ratio is 1.27. The ROI Institute advises using only the first year of benefits for short-term solutions, so this example stops at 12 months.
Notice what is missing: no test score appears in the formula. The assessment’s job here was different. Pre-test averages of 54% (trained) and 53% (comparison) suggest the groups started alike, and a post-test average of 81% shows the trained group learned. Those numbers make the story credible, but the dollars come from escalation counts.
How sensitive is the result to the attribution?
Very. The attributed drop is the shakiest input, so show what happens when it is wrong. The table keeps costs and value per escalation fixed and changes only the monthly net drop in escalations.
| Case | Net drop per month | Annual benefit | ROI |
|---|---|---|---|
| Low | 20 | $57,600 | -15.3% |
| Base | 30 | $86,400 | 27.1% |
| High | 40 | $115,200 | 69.4% |
Break-even is $68,000 divided by ($240 x 12), which is about 23.6 fewer escalations per month. Show that figure to the client, because it frames the question as “is a drop of 24 a month realistic?” instead of “is the percentage good?”
Pro tip: Publish The Break-Even Before The Result. At kickoff, write the break-even drop into the agreement and ask the client whether it is plausible. When results arrive, the discussion is about a number both sides already accepted, not about whether your method was fair.
What can you honestly claim at each stage?
Claim only what the evidence in hand supports. At course end you can claim learning, a few weeks later you can claim retained learning or observed behavior, and only after client outcome data arrives can you discuss business results. A dollar figure before then is a forecast, and should be labeled one.
The table pairs each stage with a claim to make and a claim to avoid. The stages are a suggested sequence, not a standard.
| Stage | Evidence you hold | Say | Avoid |
|---|---|---|---|
| Course end | Post-test and matched pre-test scores | “Matched learners gained 27 points on the objectives tested.” | “The course will save the client money.” |
| Retest, a few weeks later | Retest scores, observation results | “Retention on Objective 3 held at 74%.” | “Learners now perform the task correctly.” |
| After client data arrives | Outcome data for trained and comparison groups | “The trained group improved 30 escalations a month more than the comparison group.” | “Training caused every improvement.” |
| After finance review | Agreed unit value and full costs | “First-year ROI was 27%, with a range of -15% to 69%.” | A single figure with no range |
The 27-point gain and 74% retention are invented. For how to compute gain, see pre and post course assessments for instructor-led training.
What should you do next?
Pick one live corporate engagement and send the six-line data request from the table above with your next proposal or contract amendment. Ask for the measure, history, unit value, comparison group, data owner and hurdle before the next cohort starts.
Then check that your assessment tool can give you what you need. SimpliAssess’s FAQ says you can create custom reports and charts, export all test scores into a spreadsheet for further analysis, and track performance over time by student or group. Its published pages do not describe an ROI calculator or a way to import client business data, so plan to do the money calculation in your own spreadsheet.
If your client’s number comes back negative, report it with the same care as a positive one and look at which objective or delivery detail to change. A transparent negative result protects the next renewal better than an inflated positive one.
Frequently asked questions
What is training ROI?
Training ROI is the financial return of a training program, calculated as net benefits divided by fully loaded costs and multiplied by 100. Net benefits are the money value of attributed improvements minus costs. The ROI Institute treats it as the fifth level of evaluation, after reaction, learning, application and business impact.
Can assessment scores prove training ROI?
No. Assessment scores evidence learning, which is Level 2 in the Kirkpatrick and Phillips models. ROI needs the client’s business results converted to money, plus a method to separate the training’s effect from other factors. Scores can show the learning happened and help check that comparison groups started alike, but they do not give dollars.
How do you calculate training ROI?
Convert the attributed improvement in one business measure to annual money, add up fully loaded costs including participant time, then compute (benefits minus costs) divided by costs, times 100. For example, $86,400 of benefits against $68,000 of costs gives $18,400 net and an ROI of 27.1%. Show a low and high case beside it.
What is a good training ROI?
No source reviewed for this guide sets one benchmark that applies to every program. Some vendor pages quote ranges without data, so treat them cautiously. A better test is the hurdle your client agrees in advance, such as breaking even in year one, set before results are known.
Is training ROI the same as Kirkpatrick Level 4?
No. Kirkpatrick’s Level 4 measures whether targeted organizational outcomes occurred as a result of the program, and the model has four levels. ROI is the fifth level added by Phillips, which compares monetary benefits to costs. You can report Level 4 results without calculating ROI.
How many programs should you measure at the ROI level?
Few. As of the ROI Institute’s 2022 slides, it recommends evaluating about 5% of programs at the ROI level each year and about 10% at business impact. Reserve full ROI studies for the largest or most strategic programs, and report learning and behavior data for the rest.
