How to Measure Training ROI and eLearning Impact
We design eLearning that collects the data your ROI study needs.
Get a Free Quotation
Scoped proposal delivered within 48 hours
Training ROI in short
Training ROI compares the money a program returns with what it costs: ROI (%) = (monetary benefits minus program costs) ÷ program costs × 100. To measure it, pick one business measure before the course is built, record a baseline, track results level by level, isolate the training's share of the change, convert it to money and subtract fully loaded costs.
What is training ROI, and how do you calculate it?
Training ROI (return on investment) is the net money value a training program creates, shown as a percentage of what it cost. An ROI of 0% means the program broke even. Anything above 0% means the benefits were larger than the costs. An ROI of 100% means the program returned its full cost plus the same amount again.
The ROI Institute, whose ROI Methodology is often called the Phillips ROI Methodology, reports two main measures side by side, ROI and the benefit-cost ratio, and its materials also list a third, the payback period, which answers the finance question "how soon do we get the money back?"
| Measure | Formula | What it tells you |
|---|---|---|
| ROI (%) | (Program benefits minus program costs) ÷ program costs × 100 | Net return per unit spent, as a percentage. 0% = break-even. |
| Benefit-cost ratio (BCR) | Program benefits ÷ program costs | Total return per unit spent. A BCR of 1.0 = break-even. |
| Payback period | Program costs ÷ monthly benefit | How many months until the benefits cover the costs. |
A common mistake is reading ROI and BCR as the same number. A BCR of 2.0 is an ROI of 100%, not 200%. Another is treating an ROI under 100% as a loss: a 20% ROI still means the program paid for itself and returned 20% more.
Should you calculate ROI for every course?
No. A full ROI study takes time and money, so save it for programs where the decision is big. The ROI Institute suggests measuring nearly every program at the reaction level, but only 5 to 10% of programs at the ROI level. These are "best practice minimum" targets from its published model.
| Evaluation level | Share of programs to evaluate |
|---|---|
| 1. Reaction and planned action | 90 to 100% |
| 2. Learning | 60 to 90% |
| 3. Application and implementation | 30 to 40% |
| 4. Impact | 10 to 20% |
| 5. ROI | 5 to 10% |
Pick a program for a full ROI study when it is:
- expensive, or reaches a large audience;
- linked to a clear business problem or strategic goal;
- long-lived (it will run for years, so the result guides future spend);
- visible to senior leaders, who want to see its value;
- under review, so the result decides whether it continues, changes or stops.
For everything else, Level 1 to 3 data and a short impact story are usually enough.
How to measure ROI for training: 9 steps
Measuring training ROI works best when it is planned before the course is designed, not added after launch. The steps below follow the logic of the Phillips ROI Methodology, which builds on the four Kirkpatrick levels and adds a fifth level for ROI.
- 01
Start with the business measure, not the course.
Ask what problem the training must solve and which number shows it: error rate, time to competency, incidents, sales conversion, support tickets, audit findings. A proper training needs analysis checks that training is the right fix before you spend on it.
- 02
Set objectives at every level.
Write what success looks like for reaction, learning, on-the-job behavior and business impact. If the course cannot be linked to an impact objective, it is not a candidate for ROI.
- 03
Record a baseline.
Capture the business measure for a stable period before launch, for the group you will train and, if possible, for a similar group you will not train yet.
- 04
Build measurement into the course.
Decide what the eLearning must record: scores per objective, attempts, scenario choices, time on key tasks. With custom eLearning development, this is part of the design, not an afterthought. SCORM gives you completion, pass or fail, score and time; xAPI can record much more (see learning analytics).
- 05
Collect data at each level, at the right time.
Reaction and learning data come during or right after the course. Behavior and impact data come later, once people have had time to apply the skills (see the table below).
- 06
Isolate the effect of the training.
Business numbers move for many reasons: a new system, a new manager, the season, the market. Use at least one isolation method (table below). Skip it, and finance teams will rightly question the result.
- 07
Convert the impact to money.
Use values your finance team already accepts: profit per extra sale, cost of an error or a rework, cost of an incident, employee time saved (at salary plus benefits), cost of a lost hire. Anything you cannot convert credibly stays as an intangible benefit and is reported, not forced into the number.
- 08
Count fully loaded costs.
Include every cost, not just the vendor invoice (checklist below).
- 09
Calculate, then tell the story.
Report ROI, BCR and payback together, with the method, the assumptions, the intangible benefits and what you will change. A number without its method will not survive a budget meeting.
Kirkpatrick and Phillips levels: what to measure and when
The Kirkpatrick Model has four levels and was created in the 1950s. The Phillips ROI Methodology uses the same four ideas (with slightly different names) and adds Level 5, ROI, which compares the monetary benefits with the costs.
| Level (Kirkpatrick / Phillips) | What it measures | Typical data source in eLearning | When to collect |
|---|---|---|---|
| 1. Reaction / Reaction and planned action | Whether learners find the course engaging and relevant to their job | End-of-course survey, planned-action question | At the end of the course |
| 2. Learning | Knowledge, skills, attitude, confidence and commitment gained | Pre and post assessment, scenario scores, simulations | During and right after the course |
| 3. Behavior / Application and implementation | Whether people perform the critical behaviors on the job | Manager checklists, observation, system data, follow-up survey | After people have had time to apply it, often weeks to months |
| 4. Results / Business impact | Whether the targeted organizational outcomes occur | Business systems: quality, sales, safety, support, HR data | When the measure has had time to move; compare with the baseline |
| 5. ROI (Phillips only) | Monetary benefits compared with fully loaded costs | Level 4 data converted to money, plus the cost tally | After Level 4, usually on first-year benefits |
Kirkpatrick Partners point out that results in complex organizations rarely come from one program alone. Their alternative to a full ROI figure is Return on Expectations (ROE): key stakeholders agree up front what success looks like, and the evaluation reports against that.
How to isolate the effect of training
| Method | How it works | When to use it |
|---|---|---|
| Control group | Compare a trained group with a similar group trained later | Phased rollouts across sites, regions or teams. The most credible option. |
| Trend line analysis | Project the pre-training trend forward and compare it with actual results | The measure was stable before training and nothing else major changed |
| Forecasting | Model the measure from its known drivers, then credit training with the gap | You already have a solid model of what drives the measure |
| Estimates from credible sources | Participants, managers or experts estimate the training's share and their confidence | No other method is possible. Adjust for error, as below. |
When you use estimates, the ROI Institute asks the most credible people three questions: what caused the improvement, what percentage of it is due to each factor, and how confident they are in that estimate (as a percentage). The training's share is then reduced by the confidence level. This follows two of its guiding principles: use at least one method to isolate the effects, and adjust estimates of improvement for potential errors of estimation.
Fully loaded cost checklist
- Needs analysis and design time (prorated if the course is used for years)
- Course development, media and any translation
- Facilitator or coordinator costs (for blended programs)
- LMS administration and testing time
- Participant time: salaries and benefits for the hours spent in training
- Travel, venue and meals (for any live sessions)
- Administrative and overhead costs
- The cost of the evaluation itself
Be conservative. The ROI Institute's guiding principles include choosing the most conservative alternative in calculations, using only the first year of annual benefits for short-term solutions, and assuming little or no improvement where no data is available.
Planning a new course and want the measurement built in?
Worked example: training ROI from start to finish
Illustrative example (invented numbers, not client data)
A company rolls out a software simulation course to 200 claims processors to cut processing errors that need rework. Money values are in the company's own currency.
- Business measure and baseline: 1,000 rework errors a month before training.
- After training (3-month average): 700 errors a month, so 300 fewer a month, or 3,600 fewer over the first year.
- Isolation: a new form was released in the same quarter, so the team asked supervisors to estimate the training's share. Their average estimate: 50% of the improvement, with 80% confidence.
- Errors credited to training: 3,600 × 50% × 80% = 1,440 a year.
- Money value: finance values one rework error at 40, so the first-year benefit is 1,440 × 40 = 57,600.
- Fully loaded costs: 48,000 (development, LMS testing, participant time, administration and the evaluation).
| Result | Calculation | Answer |
|---|---|---|
| ROI | (57,600 minus 48,000) ÷ 48,000 × 100 | 20% |
| Benefit-cost ratio | 57,600 ÷ 48,000 | 1.2 |
| Payback period | 48,000 ÷ (57,600 ÷ 12) | 10 months |
What the report would also say: the method (supervisor estimates adjusted for confidence), the intangible benefits that were not converted (for example, fewer customer complaints about delays), and what would raise the return next year, such as reusing the course for new hires at almost no extra development cost.
Notice what the conservative steps did. Without isolation and the confidence adjustment, the same data would claim 3,600 errors, a benefit of 144,000 and an ROI of 200%. That bigger number is exactly the kind a finance team will challenge.
Training ROI calculator
Your numbers (money in your currency)
Results
- Benefit credited to training
- 57,600
- ROI
- 20%
- Benefit-cost ratio
- 1.2
- Payback
- 10 months
- Benefit = improvement × value × share × confidence
- ROI = (benefit minus costs) ÷ costs × 100
- BCR = benefit ÷ costs
- Payback months = costs ÷ (benefit ÷ 12)
Use it to forecast before you build (with estimated inputs) and again after launch (with measured inputs). If the forecast ROI is only positive with optimistic inputs, rethink the scope before development starts.
What is learning analytics, and how does it support training ROI?
Learning analytics is the collection, analysis, interpretation and communication of data about learners and their learning, used to improve learning and teaching. That is the definition of the Society for Learning Analytics Research (SoLAR). The field started in education, at the first Learning Analytics and Knowledge (LAK) conference in 2011. In corporate learning, the same data answers Levels 1 to 3 and links them to business data for Levels 4 and 5.
Types of learning analytics
SoLAR lists four types of analytics. Here is what each looks like in workplace training.
| Type | Question it answers | Workplace example |
|---|---|---|
| Descriptive | What happened? | Completion and pass rates by team, average score per objective |
| Diagnostic | Why did it happen? | One scenario question most people fail, pointing to a confusing step |
| Predictive | What is likely to happen? | Flagging learners at risk of not completing before a compliance deadline |
| Prescriptive | What should we do? | Sending extra practice on the objective a learner struggled with |
What your LMS can and cannot tell you
Most LMS learning analytics come from SCORM, the standard that lets a course run in most LMSs. SCORM reports completion, pass or fail, score and time. That covers Levels 1 and 2 well, but not what people do on the job. xAPI (also called Tin Can) records activity as simple "noun, verb, object" statements ("I did this") in a Learning Record Store (LRS). It can track mobile learning, simulations and real-world activity, and it does not need a constant connection. xAPI is now an IEEE standard (IEEE 9274.1.1-2023).
| Aspect | SCORM in an LMS | xAPI with an LRS |
|---|---|---|
| What it records | Completion, pass or fail, score, time, answers to questions | Any activity written as a statement: course steps, scenario choices, simulations, offline and on-the-job tasks |
| Where data lives | Inside the LMS | In an LRS, which can run on its own or inside an LMS |
| Best for | Completion and assessment reporting (Levels 1 and 2) | Linking learning to behavior and business data (Levels 3 to 5) |
| What you need | An LMS that supports the SCORM version you publish | An LRS (standalone or in your LMS) and a plan for which statements to record |
If your current LMS only reports completions, the question is not "which dashboard?" but "what data do we need for our impact measure?". That is often where LMS implementation and course design need to be planned together.
Choosing a learning analytics solution
Learning analytics providers range from the reporting built into your LMS, to standalone LRSs, to BI dashboards fed by HR and business systems. Before you compare learning analytics companies, answer these questions:
- Which business measure must the data explain, and where does that measure live today?
- Do you need SCORM data only, or xAPI statements from courses, simulations and on-the-job tasks?
- Can the tool join learning data with HR or business data, and who in your team will do it?
- Who will read the learning analytics dashboards, and what decision will each view support?
- How is learner data protected, and who can see individual records?
AI in learning analytics
AI can speed up predictive and prescriptive analytics, for example by spotting learners who need help or suggesting practice. SoLAR also flags the risks: data privacy, opaque "black box" algorithms, biased training data and unequal access. Use AI suggestions as input for people to review, and tell learners what data is collected and why.
Building a learning analytics business case
Treat analytics like any other program: name the decision it will improve, estimate the time or cost it saves (for example, manual reporting hours or wasted seats in courses nobody needs), list the costs, and apply the same ROI formula above. Start small with one high-value program.
Customer and channel partner training ROI
Customer and partner training ROI uses the same formula, but the business measures sit outside your workforce. The money value usually comes from support costs, product adoption, renewals and partner sales, so you need data from your support, product and CRM teams.
| Audience | Business measures to track | How to isolate |
|---|---|---|
| Customers (product and software training) | Support tickets per account, time to first value, feature adoption, renewal rate, expansion revenue | Compare trained and untrained accounts of similar size and age |
| Channel partners and resellers | Time to first deal, deal size, win rate, certified staff per partner, support escalations | Compare partners before and after certification, or certified with not-yet-certified partners |
To measure the ROI of customer training programs, agree with your support and customer success teams on one measure and its money value before launch. Customers who choose to take training may already be more engaged, so a comparison group of similar accounts matters even more here.
Common training ROI mistakes
- Measuring after the fact. Without a baseline, there is nothing to compare with.
- Crediting training with all of the change. Always isolate, and say how you did it.
- Counting only the vendor invoice. Participant time (salaries and benefits) is a real cost and can be a large one.
- Using completion as proof of value. Completion is a delivery measure, not an outcome.
- Forcing every benefit into money. Report credible intangibles separately instead.
- Reporting one big number. Show ROI, BCR and payback with the method and assumptions.
- Running a full ROI study on every course. Save it for the few programs where the decision is big.
How IKHYA helps you measure eLearning ROI
IKHYA is a custom eLearning development company that has delivered eLearning since 2010, with 120+ specialists, 500+ projects and clients in 40+ countries. Learning analytics and ROI measurement (xAPI, Kirkpatrick) is one of our services, so measurement can be planned into a course from the first design meeting.
Measurement designed into the course
Through custom eLearning development, we build courses that record the data your impact measure needs, published as SCORM 1.2, SCORM 2004, xAPI or HTML5.
Data that works in your LMS
From LMS implementation to course testing, we make sure tracking works in your platform. LMS platforms we have delivered courses to include Moodle, TalentLMS, Docebo and Cornerstone.
A measurement plan before you build
Our eLearning consulting helps you pick the business measure, the evaluation levels and the data sources before development starts.
An example of a learning measure: in a compliance overhaul we delivered, the completion and pass rate rose from 61% to 82% across 1,200 employees. That is the kind of Level 2 result a measurement plan should capture and then connect to the business measures that matter to your leaders.
Frequently asked questions
There is no universal benchmark. Any ROI above 0% means the program returned more than it cost. The ROI Institute suggests setting the target before you start: break-even (0%), the same return you expect from other investments, slightly above it, or what the program's sponsor expects.
Measure reaction and learning straight away, and behavior and impact only after people have had time to apply the skills, often weeks to months. Most ROI studies of short programs use the first year of benefits. See the level-by-level table.
It is the ROI Institute's evaluation process. It uses five levels (reaction, learning, application, impact, ROI), requires isolating the training's effect and converts impact to money before comparing it with fully loaded costs. See the 9 steps.
ROI is a financial ratio: net benefits divided by costs. Return on Expectations (ROE), from Kirkpatrick Partners, measures results against what key stakeholders defined as success before the program started. ROE is useful when results are hard to convert to money.
Learning analytics is the collection, analysis, interpretation and communication of data about learners and their learning, used to improve it (SoLAR). In training, it supplies the data for ROI. See learning analytics.
Not usually. An LMS reports completions, scores and time, which covers learning but not business impact. ROI also needs business data, isolation and costs. See what your LMS can and cannot tell you.
Yes. Learning analytics and ROI measurement (xAPI, Kirkpatrick) is one of our services, and we publish courses as SCORM 1.2, SCORM 2004, xAPI or HTML5. See how IKHYA helps.
SCORM is enough if you only need completions and scores. Choose xAPI and an LRS when you need to track scenario choices, simulations, mobile or offline activity, or link learning data to business systems. See the SCORM and xAPI comparison.
We quote each project. The cost depends on the course length, interactivity, which data points you track, whether you need xAPI and an LRS, LMS testing and how many reports you need. Send us the details and we will reply with a quote.
Need this built for your team?
Tell us about the program and the business result it must move, and we will suggest how to measure it.
