Report 03 · Measurement
Prove It: The L&D ROI Playbook
How to move from completion rates to numbers a CFO respects: cost per learner, verified skill growth, and the retention math that makes your budget the easiest line to defend instead of the first to cut.
The short version
L&D has an evidence problem, not a value problem.Half of L&D professionals report executives worrying that employees lack the skills to execute the company's strategy, while the function meant to fix that still reports its results in hours logged and completion rates.
This playbook replaces activity metrics with three numbers, adds the turnover math finance already believes, and ends with a standing fifteen-minute readout that makes budget season boring.
Evidence: 1 LinkedIn Learning (2025)
L&D professionals reporting executive concern that employees cannot execute strategy.
1 LinkedIn Learning, 2025Learning opportunities rank first among retention strategies for the 88% of organizations worried about retention.
1 LinkedIn Learning, 2025Organizations tracking engagement as their learning impact metric, with 64% tracking retention. Few reach business outcomes.
1 LinkedIn Learning, 2025The replacement cost of one departing employee, the offset math behind the budget case.
2 Gallup and SHRM, 2025What's inside
The question is not did they attend. It is did capability move, and what did that cost.
Why L&D budgets get cut first
The CFO asks three questions. Answer all three before the meeting.Budgets do not get cut because finance dislikes learning. They get cut because, of every line on the P&L, L&D is the one that answers what did we get for it with a shrug and a satisfaction score.
Across the wider economy, 63% of employers already name the skills gap the number one barrier to transformation, so the capability question is on the board agenda whether or not L&D frames it. When a CFO reviews your line, three questions are on the table.
Evidence: 4 World Economic Forum (2025)
What did we spend, per person who used it?
Not budget divided by headcount. Spend divided by active learners. Most teams cannot produce this number inside a week, and unused seats hide inside it.
What changed because of it?
Completions measure attendance. The question is capability: which skills verifiably moved, for whom, and did anyone work measurably better.
What would break if we cut it?
If the honest answer is nothing visible this quarter, the budget is already gone and just has not been announced. The defensible answer names the gaps that would reopen and what they cost.
Engagement, tracked by 72% of organizations, and retention, tracked by 64%, are worth watching, but both are moved by a hundred things. The one number only L&D can own is verified skill change tied to cost. That is the moat.
Evidence: 1 LinkedIn Learning (2025)
The measurement ladder
Four levels, four instruments, and the level most teams never leave.The Kirkpatrick model has organized training evaluation since 1959, and it still maps cleanly onto modern tooling. The failure mode has not changed either: most organizations climb to level one and stop. Each level answers a different sentence in the CFO conversation.
Evidence: 3 Kirkpatrick Partners (2025)
| Level | It answers | Modern instrument | If you stop here |
|---|---|---|---|
| L1 Reaction | Did they like it? | Pulse rating after each learning unit | You have measured catering, not capability. |
| L2 Learning | Did they learn it? | Before and after skill checks, spaced quiz recall | Knowledge that never reaches the job. |
| L3 Behavior | Do they use it? | Manager behavior ratings 30 to 60 days out | The strongest predictor most organizations skip. |
| L4 Results | Did the business move? | Paired operating metrics: ramp time, quality, retention of trained cohorts | The summit. Pair the metric, do not claim causation alone. |
Evidence: 3 Kirkpatrick Partners (2025)
Practical counsel: do not build all four levels for everything. Run levels one and two across the whole catalog cheaply and automatically, and reserve level three and four instrumentation for your three most expensive programs. Depth where the money is, breadth everywhere else.
The unit economics of learning
Cost per active learner and the turnover offset, worked on a 200-person org.Two numbers transform the budget conversation: cost per active learner, what you actually pay for actual usage, and the turnover offset, what learning saves in exits that do not happen. The worked example below uses a 200-person organization. Swap in your own figures.
Budget divided by headcount flatters the number. Dividing annual spend by the people who actually used something last quarter is the honest denominator, and it is the one finance will trust.
| Line | Example org | Formula |
|---|---|---|
| Headcount or licensed seats | 200 | A |
| Annual L&D spend (content, platform, programs) | €90,000 | B |
| Nominal cost per head, the number in your deck today | €450 | B ÷ A |
| Active learners last quarter (used anything once) | 74 (37%) | C |
| Real cost per active learner, the honest number | €1,216 | B ÷ C |
| Replacement cost of one €55k employee at 50–200% of salary | €27.5k–€110k | D |
| Exits the program must prevent to break even | about 1 to 3 per year | B ÷ D |
Illustrative figures. The formulas are the point, not the numbers.
Evidence: 2 Gallup and SHRM (2025)
At 50 to 200 percent of salary per departure, a mid-size L&D budget pays for itself by preventing one to three exits a year. That is the sentence finance funds.
Evidence: 2 Gallup and SHRM (2025)
With learning ranked the number one retention strategy by the organizations most worried about retention, and one departure costing half to double a salary, an entire mid-size L&D budget pays for itself by preventing one to three exits a year. The uncomfortable row is utilization: at 37% active use you pay €1,216 per learner for a €450 promise. Fix usage before asking for budget.
Evidence: 1 LinkedIn Learning (2025) / 2 Gallup and SHRM (2025)
Where your spend stands
The 2025 benchmarks, with their weaknesses named.Leadership will ask for the benchmark, so bring it yourself and name its weaknesses. US organizations spent $102.8 billion on training in 2025, up 4.9%. Training Magazine puts the average at $874 per learner, the people who actually trained, while ATD reports $846 in direct spend per employee across headcount, roughly 0.9% of payroll. The denominators differ. Quote the wrong one and the CFO takes the meeting from you.
Evidence: 5 Training Magazine, ATD and TalentLMS (2026)
Source: Training Magazine 2025 Industry Report
Evidence: 5 Training Magazine, ATD and TalentLMS (2026)
Source: Training Magazine 2025, roughly half the budget is operational before strategy starts
Evidence: 5 Training Magazine, ATD and TalentLMS (2026)
First, size is destiny on unit cost: firms of 100 to 999 people pay about 2.3 times more per learner than enterprises, because fixed costs do not amortize. If you are mid-size and near the average you are not overspending, you are paying the small-org tax, which makes utilization the lever, not the rate card.
Second, the average employee got 40 hours of training in 2025, down from 47, while 70% admit to multitasking through it. The market is spending more per hour on less attention per hour. Benchmarks measure spending, not yield, and that asymmetry is your opening.
Evidence: 5 Training Magazine, ATD and TalentLMS (2026)
Objection handling
The four sentences that kill budgets, and the replies that hold.Reply: here is the line as unit economics. Spend B, active learners C, cost per learner B divided by C, set against a replacement cost of 50 to 200 percent of salary per exit. The budget breaks even at one to three prevented exits a year, and here is the utilization number I am fixing first. It works because you brought the honest denominator and a break-even, which is how finance evaluates every other line.
Evidence: 2 Gallup and SHRM (2025)
Reply: correct, so I will not claim it. What I will show is the same pairing, announced in advance, running every quarter: trained against untrained cohorts on ramp time. Three quarters pointing the same direction is a better basis for a decision than a one-off ROI percentage neither of us believes. Conceding the epistemics buys credibility for the trend, and pre-announcing the metric is what separates evidence from advocacy.
Reply: the data runs the other way. Organizations with strong learning cultures see 57% higher retention, and people who move internally stay at 70% against 45% for those who do not move. People rarely leave because they grew. They leave because growing here felt impossible. Two large-sample numbers beat one anecdote, and they reframe L&D as the supply line for internal mobility.
Evidence: 6 LinkedIn, Josh Bersin Company and IBM/BCG (2025)
Reply: a pause is not neutral. Technical skills now halve in about 2.5 years, so a 12-month freeze reopens gaps we then close at external-hire prices, three to five times the internal cost. I would rather cut the two lowest-utilization programs today and protect the loop. Pricing the pause, and offering a real cut of your own, signals you manage the portfolio rather than defend it.
Evidence: 6 LinkedIn, Josh Bersin Company and IBM/BCG (2025)
The scorecard, and ninety days to stand it up
Three numbers every month, and the plan that gets you there.Annual impact decks arrive too late to change opinions. A monthly readout with the same three numbers builds the pattern that protects budgets: this team measures itself before anyone asks. The template below uses illustrative figures.
| Metric | This month | Change vs last | What it tells leadership |
|---|---|---|---|
| Active learners, share of headcount who learned this month | 61% | up 6 points | Adoption. The denominator of every other claim. |
| Verified skill growth, average mastery delta on target skills | +12% | up 3 points | Capability. The number only L&D can own. |
| Cost per active learner, spend divided by active learners, annualized | €742 | down €98 | Efficiency. Falls as adoption rises, and finance loves it. |
Illustrative figures. Once a quarter, add one level four pairing, for example ramp time or 90-day retention of trained against untrained cohorts.
Ninety days from shrug to scorecard
Days 1 to 30: get the honest baseline
Compute cost per active learner and seat utilization, and publish both internally, uninvited. Pick the three programs that carry most of the spend for the full ladder. Retire the smile sheet as your headline metric and keep it as level one telemetry.
Days 31 to 60: instrument skill change
Add before and after skill checks to the top three programs, small and repeatable. Start manager behavior ratings 30 days after each cohort finishes. Define this year's ten target skills so verified growth has a denominator.
Days 61 to 90: book the standing readout
Ship scorecard one to the exec team: three numbers, one page, fifteen minutes. Add the turnover-offset math with your own figures. Set the quarterly level four pairing and name the metric now, because announcing it in advance is what makes it believable later.
Self-audit and takeaways
Eight questions, a scoring key, and the one-page version.Could you defend the budget tomorrow? Check every statement that is true today.
| Score | Band | What it means |
|---|---|---|
| 0 to 2 | Flying blind | Spend is going out and nobody can see where it lands. Start with utilization this week. |
| 3 to 4 | Counting activity | You measure motion, not movement. Add skill checks to the big three programs. |
| 5 to 6 | Measuring what matters | Outcomes are on the scoreboard. Tighten cadence to monthly, without heroics. |
| 7 to 8 | Receipts ready | You run learning like a portfolio. The budget conversation is already won. |
Takeaways, the one-page version
Evidence: 2 Gallup and SHRM (2025) / 5 Training Magazine, ATD and TalentLMS (2026)
This week: publish cost per active learner and seat utilization, uninvited, and draft the turnover-offset math with your own salary bands. This quarter: ship the three-number scorecard and pre-announce one cohort pairing.
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.comFuture of Jobs Report 2025
Employer survey of more than 1,000 companies. 63% name the skills gap the top barrier to transformation, the strategic backdrop for the capability case.
weforum.org2025 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.comLearning-culture, internal-mobility and skill half-life analyses
57% higher retention in strong learning cultures; three-year retention 70% for internal movers against 45% for those who do not move; external hiring three to five times internal cost; technical skill half-life near 2.5 years.
economicgraph.linkedin.com