Report 17 · Measurement
The L&D ROI Report 2026
The complete method for proving what learning returns: cost per active learner, verified skill movement, the retention offset, sensitivity ranges, a board slide and a 90-day pilot plan.
The short version
L&D does not have a value problem. It has an evidence problem.Executives already believe skills matter. Half of L&D professionals say their leaders worry the workforce cannot execute strategy, and learning opportunities rank first among retention strategies. Yet most teams still report activity: hours, enrolments, completions and smile sheets.
This report replaces activity with a small set of numbers finance already trusts: cost per active learner, verified skill movement, and the retention offset. It shows the formulas, works them through illustrative examples you can replace with your own inputs, and ends with a board slide and a 90-day pilot.
Evidence: R1 LinkedIn Learning (2025)
L&D professionals reporting executive concern that employees cannot execute strategy.
R1 LinkedIn Learning, 2025Learning opportunities rank first among retention strategies for organizations worried about retention.
R1 LinkedIn Learning, 2025Organizations tracking engagement as their learning impact metric.
R1 LinkedIn Learning, 2025Replacement cost of one departing employee, rising with seniority.
R2 Gallup and SHRM, 2025Lead with three numbers, not thirty: cost per active learner, verified skill movement, retention offset.
What is in this report
Fifteen chapters, built to be used in the next budget conversation.| Part | Chapters | You leave with |
|---|---|---|
| The case | 2 Why budgets get cut, 3 The measurement ladder | The language finance uses, level by level |
| The math | 4 Unit economics, 5 Retention offset, 6 Time value, 7 Sensitivity | Formulas you can fill in tonight |
| The context | 8 Benchmarks, 9 Worked scenarios | Where your numbers sit, and three illustrative orgs |
| The decision | 10 Procurement, 11 Board slide, 12 Pilot playbook | A buying checklist, one slide, a 90-day plan |
| The defence | 13 Objections, 14 Self-audit, 15 Appendix | Answers, a scorecard and the full method |
Three kinds of figure, always labelled
Third-party benchmarks are reproduced as published and cited in the Sources page. Worked examples are illustrative: every input is stated so you can replace it. Frameworks are Omie working frameworks, offered as a structure, not as research findings. This report contains no customer case studies and no Omie outcome claims.
Read chapters 4 to 7 with your own numbers in hand; everything else supports that page.
The line nobody can defend
When results are reported as activity, the budget looks like a cost with no return attached.Budget reviews rank lines by how clearly they connect to revenue, cost or risk. Learning usually arrives with hours delivered and satisfaction scores, which answer "did people attend" and not "what changed". In a tight year, that line is the easiest to trim.
The pressure is visible in the field. In Fosway's 2025 research, most L&D teams saw budgets fall or hold flat, while leaders say productivity has to rise and 80% of workers say they lack the time or energy.
Evidence: R20 Fosway Group (2025) / R8 Microsoft WorkLab (2025)
L&D teams whose budgets decreased or stayed the same in 2025, matching 2024.
R20 Fosway Group, 2025Leaders who say productivity must increase.
R8 Microsoft WorkLab, 2025Workers who say they lack the time or energy, a global knowledge-worker signal.
R8 Microsoft WorkLab, 2025You are not competing with other trainings. You are competing with every other line that can show a return.
Measured, but measuring the wrong thing
Most teams measure. Few measure anything a finance partner would recognise as an outcome.LinkedIn's 2025 survey shows the most common learning impact measures are engagement and retention signals, not business results. McKinsey's 2026 HR Monitor adds a perception gap: employees report far fewer training days than HR records, so even the activity numbers are contested.
Evidence: R1 LinkedIn Learning (2025) / R7 McKinsey & Company (2026)
Organizations using engagement as their learning impact measure.
R1 LinkedIn Learning, 2025Days reported by employees versus by HR in the same sample.
R7 McKinsey & Company, 2026Employees reporting no employer training in the prior year.
R7 McKinsey & Company, 2026If the people you trained and the team that trained them disagree on how much training happened, no downstream number built on hours will survive a finance review. Start from usage you can log, not hours you schedule.
Replace scheduled hours with logged, repeated use: it is the only activity number both sides accept.
Three questions finance asks
Answer these before the meeting and the meeting changes.What does it cost per person who actually uses it?
Not per licence, per seat or per enrolment. Per active learner, over a defined period. Chapter 4.
What changed that we can see?
A skill signal that moved, measured the same way before and after, on a defined group. Chapters 3 and 6.
What would it cost us not to do it?
The retention offset: what a small change in avoidable exits is worth against the program cost. Chapter 5.
Bring to the meeting
If you can only bring one number, bring cost per active learner.
Kirkpatrick, read as a ladder
Four levels, each one harder to reach and more convincing to finance.| Level | Question | Typical evidence |
|---|---|---|
| 1 Reaction | Did people find it relevant? | Short rating after use, comments |
| 2 Learning | Did knowledge or skill change? | Check questions, scenario scores, before and after |
| 3 Behavior | Do people work differently? | Manager observation, applied tasks, self-report checked by others |
| 4 Results | Did a business number move? | Retention, time to productivity, quality, cost |
Evidence: R3 Kirkpatrick Partners (2025)
Most programs stop at level 1 or 2 because the data is easy to collect there. The ladder is useful because it makes the gap explicit: every rung you skip is an assumption a finance partner is entitled to question.
Name the highest level you can honestly reach, and say which rungs are assumptions.
Level five: return on investment
Phillips adds a fifth level that converts results to money and compares them with cost.The benefit must be isolated to the program and converted to money with a stated method. If you cannot isolate it, report the range and the assumptions instead of a single percentage.
Evidence: R14 ROI Institute (2024)
Isolation methods, from strongest to weakest
Evidence: R13 HM Treasury and Evaluation Task Force (2026)
A modest ROI with a clear isolation method beats a large one with none.
Where each level breaks
Every level has a typical failure. Plan around it before you collect a single data point.| Level | Typical failure | Fix |
|---|---|---|
| 1 Reaction | Ratings measure mood, not use | Ask about relevance to this week's work |
| 2 Learning | One test, right after the lesson | Retest after days or weeks; recall decays fast |
| 3 Behavior | Self-report only | Add a manager or peer signal on the same behaviour |
| 4 Results | Credit taken for every change | Use a comparison group or a trend line |
Share of new information forgotten within seven days without reinforcement.
R15 Ebbinghaus forgetting-curve tradition (Indegene synthesis), 2025Long-term retention gain from spaced versus massed practice, across 800+ experiments.
R16 Spaced-practice research (Cepeda et al. and successors), 2025A level 2 score taken the same afternoon tells you about short-term memory, not skill.
What the research can and cannot carry
Use evidence for what it shows. Do not stretch it into a promise.A 2025 systematic review of 40 microlearning studies found positive signals across cognitive, behavioural and affective outcomes, but did not establish a universal workplace ROI. LinkedIn reports that companies with strong learning cultures show higher retention and internal mobility, an association rather than proof that learning caused it.
That is the honest position for any vendor or internal team: the direction of the evidence is favourable, and your own measurement has to supply the size of the effect in your organization.
Evidence: R11 Heliyon (2025) / R12 LinkedIn Learning (2024)
Positive signals, no universal workplace ROI established.
R11 Heliyon, 2025In companies with strong learning cultures; an association, not an intervention effect.
R12 LinkedIn Learning, 2024In companies with strong learning cultures; an association.
R12 LinkedIn Learning, 2024Quote research for direction. Quote your own data for size.
Cost per active learner
The single most useful number in this report.Use the full cost: licences, admin and manager time, content you bought or built. Use people who actually used it repeatedly, not people who were given access.
| Input | Illustrative value | Your value |
|---|---|---|
| Annual licence cost | $36,000 | |
| Admin time (0.2 FTE at $70,000) | $14,000 | |
| Content bought or built | $10,000 | |
| Total program cost | $60,000 | |
| Active learners (3 of 4 weeks) | 150 | |
| Cost per active learner | $400 per year |
Illustrative example. Every figure is an assumption to replace with your own.
If your cost per active learner is rising while seat cost is flat, adoption is falling.
Seats bought versus seats used
The gap between licences and use is usually the largest saving on the table.Unused licences are common across business software. Zylo's 2026 index reports a large share of licences going unused against its benchmark, and a median annual software spend per employee in the thousands of dollars. Learning tools are not exempt.
Evidence: R21 Zylo (2026)
Against Zylo's benchmark; a vendor dataset, not a universal rate.
R21 Zylo, 2026Median annual SaaS spend per employee across Zylo-managed organizations.
R21 Zylo, 2026Below roughly half, the conversation is about adoption before it is about content. Track it monthly.
Utilisation is the first number to fix: it lowers cost per active learner without a new purchase.
Cost per verified skill gain
Once usage is real, price the change, not the activity.Define the threshold before you measure, for example one level on a five-level scale, checked by a scenario or a manager. Count each learner once per skill.
| Step | Illustrative value |
|---|---|
| Program cost | $60,000 |
| Learners measured twice on a target skill | 120 |
| Learners who moved one level or more | 54 |
| Cost per verified skill gain | $1,111 |
Illustrative example. Compare it with the cost of a course or coaching session that delivers the same verified move.
A verified gain is a number a hiring manager understands: it is the cost of a skill you did not have to hire for.
Hidden costs that change the answer
The licence is rarely the largest line.| Cost | Where it hides | How to estimate |
|---|---|---|
| Learner time | Payroll, not the L&D budget | Minutes per week × loaded hourly rate |
| Manager time | Coaching, approvals, follow-up | Hours per month × loaded rate |
| Admin | Enrolment, reporting, content updates | Share of an FTE |
| Integration | IT time for SSO and HR data | One-off hours, amortised over 3 years |
| Switching | Migration and retraining on a new tool | One-off, only if you change |
Long courses carry a large time cost even when the licence is cheap. Short, frequent practice keeps the time cost low enough that people actually do it, which is what makes the cost per active learner fall.
Put learner time in the model. Finance will, if you do not.
The math finance already believes
Replacement cost is a well-documented number. Use it.Replacement cost of one employee, rising with seniority.
R5 Gallup, 2025Gallup's estimate of what voluntary turnover costs US businesses.
R5 Gallup, 2025Voluntary exits the Work Institute classifies as preventable.
R6 Work Institute, 2024Leavers who say their departure was preventable; a retrospective view, not causal proof.
R10 Gallup, 2024Use the low end of the replacement range (0.5) unless you have your own number. Exits avoided must come from a comparison, not from hope.
Evidence: R2 Gallup and SHRM (2025)
Use the conservative factor. A defended small number is worth more than a contested large one.
A worked offset, end to end
Illustrative: a 200-person organization, every input stated.| Input | Illustrative value |
|---|---|
| Headcount | 200 |
| Voluntary turnover | 15% (30 exits a year) |
| Average salary | $70,000 |
| Replacement-cost factor | 0.5 (low end of the range) |
| Cost of one exit | $35,000 |
| Exits avoided if turnover falls by 1 point | 2 |
| Retention offset | $70,000 a year |
Illustrative example. The factor 0.5 is the low end of the published 50 to 200% range (R2).
Against the $60,000 program in chapter 4, two avoided exits cover the cost. That is the claim worth testing: not that learning cuts turnover in general, but whether your program moves your turnover by at least the break-even amount.
Evidence: R2 Gallup and SHRM (2025)
State the break-even: how many avoided exits pay for the program. Then measure against it.
Why learning is a retention lever
The link runs through growth and the manager conversation.Higher among high-potential employees without growth opportunities; association, not causation.
R9 DDI, 2025Higher intent to leave with ineffective coaching; association.
R9 DDI, 2025Voluntary leavers who had no proactive discussion in the three months before leaving.
R10 Gallup, 2024Track turnover separately for people who used the program regularly and people who did not, in the same roles. The difference is not proof, because motivated people both learn and stay, but it is the comparison finance will ask for. Chapter 7 shows how to present it with a range.
Learning and manager conversations reinforce each other: measure them together.
Valuing time saved, carefully
Time saved is real value only if it is measured and only to the extent it is redeployed.Use loaded salary (pay plus employer costs) if finance gives you one; otherwise use base salary and say so. 1,880 hours is a common planning figure for a full-time year.
The redeployment share is the part of saved time that goes to productive work. If you do not know it, use 50% and show the result at 25% and 75% (chapter 7).
This model is a planning structure, not a research finding. Its inputs are yours. The only measured input it needs from a learning program is whether a specific task takes less time after the skill moved, checked on the same people.
Never multiply saved minutes by headcount without a redeployment share.
A worked time value
Illustrative: 120 people, one measured task.| Input | Illustrative value |
|---|---|
| People in scope | 120 |
| Average salary | $85,000 |
| Hourly value ($85,000 ÷ 1,880) | $45.21 |
| Hours saved per person per month (measured) | 2 |
| Redeployment share | 50% |
| Annual time value (2 × 12 × $45.21 × 120 × 0.5) | $65,100 |
Illustrative example. Replace each input; the 2 hours must come from a before and after measure of a named task.
Two hours a month is about six minutes a working day. That is a modest claim, which is the point: small, measured time savings on a real task survive review. Large estimated savings across "productivity" do not.
Pick one task per role, time it twice, and value only that.
Choosing the task to measure
The task decides whether the time value is believable.A good task is
| Role | Skill taught | Task to time |
|---|---|---|
| Team lead | Running a focused 1:1 | Prep plus meeting time per 1:1 |
| Support agent | Structured troubleshooting | Handle time on a ticket type |
| Sales rep | Discovery questions | Calls to a qualified next step |
| Analyst | Writing a clear brief | Review rounds per brief |
Examples of pairings, not measured results.
One well-chosen task per role is enough to make the model honest.
Putting the three values together
Time value, retention offset and avoided spend, added only where they do not overlap.| Value line | Illustrative annual value | Evidence needed |
|---|---|---|
| Time value (chapter 6) | $65,100 | Before and after timing on one task |
| Retention offset (chapter 5) | $70,000 | Turnover comparison, users versus non-users |
| Avoided spend | $12,000 | A course or tool you stopped buying |
| Total benefit | $147,100 | |
| Program cost (chapter 4) | $60,000 | |
| ROI | 145% | (147,100 − 60,000) ÷ 60,000 |
Illustrative example combining the earlier worked figures. Do not count the same person's time twice.
A single ROI figure invites a single objection. Chapter 7 turns this into a low, base and high case so the conversation is about assumptions, not about whether the number is real.
Show each value line with its evidence. Let finance remove the line it doubts and still see a return.
Low, base and high
Change the assumptions that matter and show what survives.| Assumption | Low | Base | High |
|---|---|---|---|
| Hours saved per person per month | 1 | 2 | 3 |
| Redeployment share | 25% | 50% | 75% |
| Exits avoided | 1 | 2 | 3 |
| Replacement-cost factor | 0.5 | 0.5 | 1.0 |
| Time value | $16,300 | $65,100 | $146,500 |
| Retention offset | $35,000 | $70,000 | $210,000 |
| ROI (with $12,000 avoided spend, $60,000 cost) | 5% | 145% | 514% |
Illustrative example built from the chapter 4 to 6 figures; replacement factors within the published 50 to 200% range (R2).
Evidence: R2 Gallup and SHRM (2025)
If the low case still breaks even, lead with the low case.
Which assumption moves the answer
Test one input at a time to find the number worth measuring first.Source: Illustrative calculation from the chapter 6 base case.
In this example the retention assumption carries the most weight, so it deserves the best evidence: a comparison of users and non-users in the same roles, tracked for at least two quarters.
Spend your measurement effort on the assumption that moves the answer most.
Break-even, stated plainly
The most persuasive sentence in a budget meeting is the break-even.With the illustrative $60,000 cost, $70,000 salary and 0.5 factor, the program pays for itself if it helps keep about 1.7 people a year.
With the chapter 6 inputs, about 1.8 hours a month per person (roughly five minutes a working day) covers the cost on time value alone.
"The program pays for itself if it keeps two people a year, or saves each person about two hours a month on one task. Here is how we will measure both."
Break-even turns an argument about benefits into a test you can pass or fail.
Where your spend stands
Benchmarks set expectations. They do not replace your own numbers.Industry surveys from Training Magazine and ATD publish spend per learner and hours of training each year. Use them to check that your budget is in a normal range, not to predict outcomes. Compare like with like: company size, industry and whether figures include staff time.
Evidence: R4 Training Magazine, ATD and TalentLMS (2026)
| Benchmark to look up | Where | Use it for |
|---|---|---|
| Training spend per learner | Training Industry Report (R4) | Is the budget in range for our size? |
| Hours of training per employee | ATD State of the Industry (R4) | Is our time ask realistic? |
| Share of budget by category | Training Industry Report (R4) | Are we over-invested in one area? |
| L&D budget direction | Fosway (R20) | Is a flat budget normal this year? |
Evidence: R4 Training Magazine, ATD and TalentLMS (2026) / R20 Fosway Group (2025)
Quote the benchmark with its year and population, or do not quote it.
Adoption benchmarks for learning tools
The most common failure is not content. It is use.LMS owners who would not recommend their platform.
R18 Brandon Hall Group, 2024Actively seeking a replacement at survey time.
R18 Brandon Hall Group, 2024Said their learning system fit the modern workforce, in Fosway's mostly European sample.
R20 Fosway Group, 2025Buyers meeting Capterra's strict definition: no disruption and no purchase regret.
R19 Capterra, 2026Low satisfaction with existing systems is why utilisation (chapter 4) belongs in every report. A tool people do not open has a cost per active learner close to infinity.
Benchmark adoption before you benchmark outcomes.
Three organizations, one method
Illustrative scenarios, not customers. Each shows how the same method behaves at a different size.| Input | Team of 40 | Company of 200 | Group of 1,000 |
|---|---|---|---|
| Program cost a year | $9,000 | $60,000 | $240,000 |
| Active learners (3 of 4 weeks) | 30 | 150 | 650 |
| Cost per active learner | $300 | $400 | $369 |
| Average salary | $60,000 | $70,000 | $75,000 |
| Voluntary turnover | 12% | 15% | 13% |
| Break-even exits a year | 0.3 | 1.7 | 6.4 |
Illustrative scenarios. No real organization is described; replace every input. Break-even uses the 0.5 replacement factor (R2).
Evidence: R2 Gallup and SHRM (2025)
Cost per active learner usually falls with size, but adoption is harder to keep high across many teams. The break-even exit count grows with cost, so larger programs need stronger evidence on retention.
Find the column nearest to you and replace its numbers first.
Scenario A: a team of 40
Illustrative. A manager-led team buying its first learning program.Cost
$9,000 a year, no admin hire, the manager runs a 10-minute weekly review of progress.
Measure
Utilisation monthly; one task timed before and after for the two most common roles.
Break-even
One avoided exit every three years covers the cost; any measured time saving is upside.
Risk
Small numbers: one leaver changes the turnover rate a lot, so report time value first.
Fill in for your team
Illustrative scenario; the worksheet is yours to complete.
In small teams, lead with time value; retention is too noisy to carry the case alone.
Scenario B: a company of 200
Illustrative. The worked case used throughout chapters 4 to 7.| Line | Base case |
|---|---|
| Program cost | $60,000 |
| Cost per active learner | $400 |
| Time value (2 hours a month, 50% redeployed) | $65,100 |
| Retention offset (2 exits avoided) | $70,000 |
| Avoided spend | $12,000 |
| ROI, base case | 145% |
| ROI, low case | 5% |
Illustrative scenario assembled from chapters 4 to 7.
At this size the program needs its own owner for about a fifth of a role, and the retention comparison becomes meaningful after two quarters. The low case still breaks even, which is the line to lead with.
At 200 people, the low case is the headline and the base case is the plan.
Scenario C: a group of 1,000
Illustrative. Several business units, one central L&D team.Cost
$240,000 a year including a full-time program owner and integration work in year one.
Adoption
Utilisation varies by unit; report it per unit and per manager, not only in total.
Evidence
Use a staggered start across units as a natural comparison group for retention.
Break-even
About six avoided exits a year at a 0.5 factor, a small share of annual leavers at this size.
Rolling out unit by unit is usually done for practical reasons anyway. Planned on purpose, it gives you the strongest isolation method on the ladder: a comparison group that started later.
At scale, design the rollout as the experiment.
Buying with the ROI in mind
Most regret starts at purchase. Decide how you will measure before you choose.Successful versus disappointed buyers who documented budget and must-haves up front.
R19 Capterra, 2026Successful versus disappointed buyers who created a formal implementation plan.
R19 Capterra, 2026Regretful buyers who also had implementation disruption.
R19 Capterra, 2026Before you see a demo
A written measurement plan is the cheapest risk control in the purchase.
Questions to ask every vendor
Ask for evidence you can check, not claims you have to trust.| Ask | Good answer | Warning sign |
|---|---|---|
| How do you report active use? | Per person, per week, exportable | Logins or seats only |
| How is skill change measured? | Before and after on the same people | Completions and badges |
| Can we see our raw data? | Yes, in a standard export | Dashboards only |
| What does setup need from us? | Named hours from IT and HR | "Nothing, it just works" |
| What happens to unused seats? | Clear true-up or flex terms | Annual lock-in only |
Integration problems are a leading source of disappointment in software purchases, and the share of organizations using the learning module inside their HR suite is small. Ask early what connects to what.
Evidence: R19 Capterra (2026) / R20 Fosway Group (2025)
If a vendor cannot export your data, you cannot prove your ROI.
A one-page buying checklist
Print it, fill it in, attach it to the purchase request.Purchase request: measurement plan
Omie working framework. Adapt the fields to your procurement template.
Shortlist rule
Evidence: R19 Capterra (2026)
Buy the tool that makes the plan measurable, not the one with the best demo.
One slide, three numbers
The whole report, compressed to what a board will read.| Slide area | What goes there | Example wording (illustrative) |
|---|---|---|
| Headline | Break-even in one sentence | Pays for itself if it keeps two people a year |
| Number 1 | Cost per active learner, trend | $400, down from $520 last quarter |
| Number 2 | Verified skill movement | 45% of measured learners moved one level |
| Number 3 | Retention comparison, with range | Users left at 9% versus 14%; low case still breaks even |
| Ask | One decision | Extend to two more units with a staggered start |
Illustrative wording. Every figure on the real slide must come from your own measurement.
Hours delivered, number of courses, satisfaction averages and vendor case studies. Keep them in an appendix for questions; they weaken the slide.
Headline, three numbers, one ask. Nothing else.
Presenting a range without losing the room
How to show uncertainty and still get a decision.Start with the low case
If it breaks even, say so first. It removes the strongest objection before anyone raises it.
Name the assumption that matters
From chapter 7: usually retention. Say how you are measuring it and when the next reading comes.
Offer the decision as a test
Ask for a time-boxed extension with a stated pass mark, not a permanent budget line.
Linking a metric to pay can distort it: in one study of 140 organizations, survey scores rose after they were tied to manager pay while turnover also rose. Keep learning measures informational, not compensated, so they stay honest.
Evidence: R17 MIT Sloan Management Review (2025)
A range with a test date is more fundable than a single number with a promise.
Days 1 to 30: set the baseline
Nothing in the first month is about results. It is about measuring the starting point.By day 30
The 30, 60 and 90 day structure is a planning frame from Omie's rollout playbook, not a research finding. Shorten or stretch it to your calendar; keep the order.
If the baseline is not written down by day 30, the pilot cannot prove anything.
Days 31 to 60: build the habit
Use is the leading indicator. Protect it.Spaced, repeated practice is what turns a lesson into a durable skill; a single session decays quickly. The second month is about frequency: short, regular use that survives a busy week.
Evidence: R15 Ebbinghaus forgetting-curve tradition (Indegene synthesis) (2025) / R16 Spaced-practice research (Cepeda et al. and successors) (2025)
Weekly utilisation by manager
Share it with managers, not only with L&D. Teams follow their manager's attention.
One nudge, not five
A single weekly reminder tied to the role's real work beats daily generic prompts.
Fix the bottom quartile
Talk to the least active teams first; the reason is usually time or relevance, both fixable.
By day 60 you should know your utilisation and why the lowest teams are low.
Days 61 to 90: measure and decide
Take the second readings and compare them with the baseline you set.| Measure | Baseline (day 30) | Reading (day 90) | Decision rule |
|---|---|---|---|
| Utilisation | Above your floor: continue | ||
| Cost per active learner | Falling: adoption is working | ||
| Task time per role | Below break-even minutes: report time value | ||
| Skill measure | Share moving one level or more | ||
| Retention comparison | Too early at 90 days: set the 6-month date |
Worksheet. Fill it with your own readings.
At day 90, decide on utilisation and time; schedule retention for month six.
Reporting the pilot honestly
A pilot report that admits its limits is more credible, not less.Every pilot report states
Government evaluation guidance such as the UK Magenta Book sets out how to describe a counterfactual, uncertainty and limitations. Borrowing its structure makes an internal report read as rigorous to finance.
Evidence: R13 HM Treasury and Evaluation Task Force (2026)
Say what you do not know yet. Then say when you will.
Four objections, four answers
The sentences to have ready.| Objection | Answer |
|---|---|
| "People stay for other reasons." | Agreed. We compare users with non-users in the same roles and report a range, not a single cause. |
| "Time saved is not money saved." | We count only the redeployed share, on one timed task, and show 25% as the low case. |
| "Completions were high last year and nothing changed." | That is why we report active use and skill movement, not completions. |
| "We can cut it and see." | The break-even is two people a year. One avoidable exit costs more than half the program. |
Illustrative figures from the chapter 5 to 7 worked example.
Voluntary exits the Work Institute classifies as preventable.
R6 Work Institute, 2024Replacement cost of one employee.
R5 Gallup, 2025Agree with the true part of each objection, then show the measure that answers it.
When the answer is "not yet"
Sometimes the honest reply is that the data is not in. Say it well.Name the date
"Retention needs two quarters. The first reading is on [date]."
Show the leading indicator
"Utilisation is above the floor we agreed, and cost per active learner is falling."
Offer the cheaper path
"We can hold scope flat until the reading, instead of extending now."
Quoting vendor case studies or industry averages as if they were your result. They answer a different question, and a finance partner will notice.
A dated "not yet" keeps the budget; an undated "trust us" loses it.
Eight questions, scored
Score one point for each yes. Six or more: ready for the budget meeting.| Score | Reading |
|---|---|
| 0 to 3 | Activity reporting. Start with chapter 4. |
| 4 to 5 | Partial evidence. Add the comparison and the range. |
| 6 to 8 | Fundable. Build the board slide (chapter 11). |
Omie working framework.
Fix the lowest-numbered "no" first; the questions are in order of dependency.
The monthly scorecard
Three numbers every month, three more every quarter.| Cadence | Number | Owner |
|---|---|---|
| Monthly | Utilisation (active ÷ licensed) | Program owner |
| Monthly | Cost per active learner | Program owner with finance |
| Monthly | Share of active learners by manager | Managers |
| Quarterly | Verified skill movement | L&D |
| Quarterly | Task time per role | Role owners |
| Half-yearly | Retention comparison, with range | People analytics |
This month
A scorecard that fits on one page gets read every month.
Methodology
How the numbers in this report were built.Evidence: R2 Gallup and SHRM (2025)
Measured, illustrative and framework figures are always labelled as such.
Formulas in one place
Copy these into your own sheet.| Measure | Formula |
|---|---|
| Cost per active learner | Program cost ÷ people active 3 of the last 4 weeks |
| Utilisation | Active learners ÷ licensed seats |
| Cost per verified skill gain | Program cost ÷ learners who moved by the threshold |
| Hourly value | Loaded annual salary ÷ 1,880 |
| Time value | Hours saved a month × 12 × hourly value × people × redeployment share |
| Retention offset | Exits avoided × salary × replacement factor |
| ROI % | (Total benefit − program cost) ÷ program cost × 100 |
| Break-even exits | Program cost ÷ (salary × replacement factor) |
Evidence: R14 ROI Institute (2024)
Every formula uses inputs you can look up or measure in your own organization.
Glossary
The terms used in this report, defined once.| Term | Meaning here |
|---|---|
| Active learner | Someone who used the program in at least 3 of the last 4 weeks |
| Verified skill gain | A pre-agreed move on a skill measure, checked twice on the same person |
| Redeployment share | The part of saved time that goes to productive work |
| Replacement-cost factor | Cost of replacing a leaver as a share of salary |
| Comparison group | Similar people who did not use the program, or started later |
| Break-even | The smallest effect at which benefit equals cost |
| Illustrative | An example with stated inputs, not a measured result |
Run your own figures through the ROI calculator on tryomie.com, or talk it through with the Omie team. Every formula here works with any learning program, not only Omie.
Use the same definitions in every report so the numbers stay comparable.
Sources and method
Every external numeric claim in this report points to one of these 2024 to 2026 sources. Forecasts and self-reported surveys are labelled so they are not mistaken for causal proof.
Workplace Learning Report 2025
Survey of 937 L&D and HR professionals and 679 learners. Source for executive skill concerns, retention rankings, and impact-measurement practices.
learning.linkedin.comTurnover and replacement-cost research
Replacement cost of 50 to 200 percent of annual salary. Gallup estimates US voluntary turnover near $1 trillion a year, and SHRM benchmarks about $4,700 in direct cost per hire.
gallup.comThe Kirkpatrick four-level evaluation model
Four-level training evaluation model, first published in 1959 and maintained by Kirkpatrick Partners. The ladder here modernizes the instruments, not the levels.
kirkpatrickpartners.com2025 Training Industry Report; ATD State of the Industry 2026; TalentLMS 2026
$102.8B US spend; $874 per learner; size-band spend $1,091, $782, $468; 40 average hours down from 47; budget mix compliance 13%, management 13%, IT 11%, onboarding 11%; ATD $846 direct per employee near 0.9% of payroll; 70% multitasking during training.
trainingmag.comThe Trillion-Dollar Cost of Voluntary Turnover
Replacement cost of one-half to two times annual salary; roughly $1 trillion a year in US voluntary turnover; manager-exit and one-to-one engagement findings.
gallup.comRetention Report
About 75 percent of voluntary turnover classified as preventable, coded from employee exit interviews.
workinstitute.comHR Monitor 2026: A Turning Point for the People Function
HR and employee surveys across ten countries, with a strong European emphasis.
mckinsey.comWork Trend Index Annual Report: Executive Summary
Survey plus Microsoft 365 signals. Knowledge-work context may not generalize to every role.
cdn-dynmedia-1.microsoft.comGlobal Leadership Forecast 2025
Global HR and leader survey. Multipliers reported here are associations.
media.ddiworld.com42% of Employee Turnover Is Preventable but Often Ignored
Voluntary-leaver sample and manager-conversation synthesis. Associations are not guarantees.
gallup.comMicrolearning Beyond Boundaries: A Systematic Review and a Novel Framework for Improving Learning Outcomes
Systematic review of 40 studies across cognitive, behavioral, and affective outcomes. It does not establish universal workplace ROI.
doi.orgWorkplace Learning Report: L&D Powers the AI Future
Company-level learning-culture associations. The comparisons are not causal intervention effects.
linkedin.comMagenta Book: Central Government Guidance on Evaluation
Guidance on theory-based, process, impact, and value-for-money evaluation.
gov.ukMaking ROI Part of Your Story
ROI methodology reference for isolating effects, converting benefits, counting cost, and calculating return.
roiinstitute.netThe forgetting curve in workplace learning
About 50% of new information lost within an hour, about 70% within 24 hours, and up to 90% within a week without reinforcement. Averages; individual curves vary with meaning and stakes.
indegene.comSpacing-effect syntheses across 800+ experiments
Spaced practice improves long-term retention by roughly 200% versus massed practice, with a two to three times recall advantage commonly reported. Retrieval practice adds a further, separate gain over re-reading.
pubmed.ncbi.nlm.nih.gov140-organization study on metrics and compensation
When survey scores were tied to manager compensation, scores rose 12% while voluntary turnover rose 8% (reported 2023).
sloanreview.mit.eduLMS Trends benchmark survey
31% of owners would not recommend their LMS and 38% were actively seeking a replacement, the most recent large public satisfaction benchmark.
brandonhall.comSoftware Buying Trends Report 2026 and SaaS Buying Guide
Survey fielded in 2025 across 11 countries. Disruption and regret are distinct outcomes.
capterra.comDigital Learning Realities Research 2025
Learning-market research with a primarily European context.
fosway.com2026 SaaS Management Index
Vendor benchmark covering 40 million licenses and $75 billion in managed spend.
zylo.com