Measurement

How to prove training ROI to a CFO

Finance is not hostile to L&D. It is hostile to numbers that cannot survive a follow-up question — and most training ROI presentations are built entirely from those.

8 min read
In short

CFOs accept cost avoidance, cohort comparison and waste reduction. They discount satisfaction scores, industry benchmarks applied to your organisation, and any ROI multiple presented without a stated counterfactual. Lead with the cost of training that failed, because that number is defensible and already in your budget.

Why the usual pitch fails

Only 29% of L&D leaders feel confident proving ROI, and 67% report struggling to demonstrate impact to executives. That is not a measurement problem so much as a credibility problem, and it is usually self-inflicted.

Three things get presented that finance discounts immediately:

Satisfaction scores. "Learners rated the programme 4.2/5." A CFO hears: people enjoyed a day away from their desk. Reaction data measures the experience, not the effect, and finance knows the difference even when L&D presents them as equivalent.

Borrowed benchmarks. "Industry research shows training delivers 200–600% ROI." The obvious question is whether it applies here. It does not, and citing it signals that you have not measured your own.

Unfalsifiable multiples. "This programme delivered 340% ROI." Where is the counterfactual? What would have happened without it? If you cannot answer, the number is decoration.

The pattern is the same in each case: a figure that cannot survive one follow-up question. Present three of those and finance stops listening to the fourth.

What finance actually accepts

Argument Why it lands
Cost avoidance Money not spent is the most defensible category in finance
Waste reduction Uses spend already in your budget; no attribution needed
Cohort comparison A real counterfactual, with stated limits
Time-to-competency Directly convertible to salary cost
Incident and error rates Already measured by someone else, so not your number to defend
Retention differential Replacement cost is a known figure in every HR function

The unifying feature is that each starts from a number finance already has and trusts. You are not asking them to accept a new measurement regime — you are reinterpreting data already on their books.

The waste calculation

This is the strongest opening argument available to most L&D functions, and almost nobody uses it.

You already know two numbers: what you spend on training, and how many hours of employee time it consumes. What you have not calculated is how much of that reached people who were not paying attention.

The calculation

Wasted spend = (licence cost + delivery cost + loaded hourly cost × hours) × disengagement rate

Where disengagement rate comes from content-level engagement data — the proportion of learners who were not attending during a meaningful part of the module.

This works for three reasons.

It uses money already committed. You are not asking for budget. You are describing what happens to budget you already have, which is a fundamentally easier conversation.

It requires no attribution argument. You do not need to prove training caused a revenue change. You need to show that a proportion of a known spend did not land.

It is verifiable. Finance can check the licence cost and the headcount. The only number they have to take from you is the engagement rate, and that comes from instrumentation rather than a survey.

Cohort comparison done honestly

When you do need to demonstrate outcome impact, cohort comparison is the only method most organisations can run credibly.

The structure:

  1. Pick one metric that matters and is already collected. Error rate, incident count, time-to-productivity, sales conversion. Do not invent a metric for this exercise.
  2. Compare trained against not-yet-trained. Phased rollouts give you this for free — the people scheduled for Q3 are your control group in Q2.
  3. Control for what you can. Tenure, team, role level, region.
  4. State what you could not control for. Explicitly, in the deck.

That fourth step is the one people skip, and it is the one that builds credibility. A CFO who sees you name your own confound trusts the rest of the analysis considerably more. A CFO who spots an unnamed confound discounts everything.

A worked example

Here is the shape of a presentation that works. Numbers are illustrative — the structure is the point.

Situation. Mandatory data-handling training. 3,200 employees. Licence cost £41,000. Average completion time 45 minutes at a loaded cost of £38/hour.

Total investment: £41,000 + (3,200 × 0.75 × £38) = £132,200

What completion said: 94% completion. Green on the dashboard.

What instrumentation showed: 38% of learners had the window unfocused for more than half the session. 61% disengaged during the incident escalation section.

Waste: 38% of £132,200 = £50,236 delivered to people who were not attending.

Intervention: Rewrote the escalation section — split into three shorter segments, added a scenario-based check. Cost: 6 days of instructional design, £3,400.

Result: Engagement through that section rose from 39% to 74%. Assessment scores on related questions rose 22 points. Data-handling incidents in the following quarter fell from 14 to 9.

What we are not claiming: that the training caused the entire incident reduction. Q3 also saw a policy change and a new starter cohort. The engagement and assessment improvements are directly attributable; the incident figure is contextual.

That last paragraph is what makes the rest believable.

Phrases to avoid

"Learners loved it." Reaction data. Finance discounts it entirely, and using it early signals you have nothing stronger.

"Studies show..." followed by an industry ROI figure applied to your programme. If you had your own number you would have used it.

"Training is an investment in our people." True, and unfalsifiable, which makes it useless in a budget conversation. It reads as a request to stop asking for evidence.

Any ROI percentage without a stated counterfactual. The first question will be "compared to what?" Have the answer before you show the number.

"We can't really measure this." Sometimes said as a pre-emptive defence. It concedes the argument. You can always measure engagement and waste, even when you cannot measure outcome.


The reframe that changes these conversations: stop trying to prove training created value, and start showing what it costs when it fails. The first requires an attribution argument you will probably lose. The second uses money already in your budget, and it is unarguable.

Frequently asked questions

What ROI figure should I claim for training?
Ideally none as a headline. Published figures of 200–600% exist for rigorously measured programmes, but quoting an industry range for your own programme invites the obvious question of whether it applies to you — and it does not, until you have measured it. Lead with cost avoidance and waste reduction, which are calculable from your own data.
How do I attribute a business outcome to training?
Cohort comparison is the only method most organisations can run credibly. Compare a trained group against an untrained or later-trained group on a metric that matters, control for what you can, and state plainly what you could not control for. Stating your limitations increases credibility with finance rather than reducing it.
What if I don't have any outcome data yet?
Start with waste. You already know your licence spend and your delivery hours. If you can show what proportion of that spend reached people who were not engaged, you have a defensible number on day one that requires no attribution argument at all.
Should I use the Phillips ROI model?
It is a sound framework for converting Kirkpatrick Level 4 results into a financial figure, and it is worth knowing. But the model is not the hard part — the hard part is having credible Level 3 and 4 data to feed into it. Most ROI presentations fail on inputs, not arithmetic.

See where your content loses people

Book a walkthrough and we will show you the engagement data on your own content.

Request a demo
Get Started

See what completion rates can't tell you

Find out exactly where your content works, where it fails, and what disengagement looks like before people leave.

Request a Demo