Every quarter, somewhere, a slide goes up: 94% course completion. 4.6/5 satisfaction. 2,300 learning hours delivered.
Everyone nods. Someone says, "Great engagement." The deck moves on.
And not one of those numbers proves that a single person got better at their job.
That's the open secret of learning ROI: most of us aren't lying on purpose. We just report the metrics that are easy to collect and hope they stand in for the ones that are hard to prove. This guide is about telling the difference, and about measuring training ROI the way an auditor would, so the number survives the one question that kills most L&D reporting: "How do you know?"
Why most learning ROI metrics are quietly fake
Think like a forensic accountant for a second. A fraudulent balance sheet rarely contains a single made-up figure. It contains real figures, arranged to imply something that isn't true.
Learning and development measurement fails the same way. Every number on the average L&D dashboard is real; completion really is 94%. It's the implication that's fake: that completion equals capability, that satisfaction equals impact, that hours equal outcomes.
Nobody signs off on a lie. We just let the easy numbers do the talking, because the honest ones are expensive to produce.
The 3 vanity metrics that break training ROI
Completions. A completion indicates that someone has reached the end of a video. It says nothing about whether they can now do the thing. Netflix has a 99% completion rate too, and we don't call that learning.
Satisfaction scores. The "happy sheet" measures enjoyment, not capability. Some of the most valuable training scores badly in the moment because it's hard and uncomfortable. If your best number is a smile score, you're measuring the catering.
Hours delivered. An input dressed as an outcome. "We delivered 2,300 hours" is the L&D version of a chef bragging about how much gas the stove burned.
None of these are useless. They're just the lobby of measurement, and too many reports never leave the lobby.
How to measure training ROI honestly: the Evidence Ladder
Here's a simple way to grade any learning metric you own. Every piece of "evidence" sits on one of five rungs, weakest to strongest:
- Rung 1: Activity. Seats filled, hours logged, modules opened. This is an alibi, not evidence. It proves the training happened, not that it worked.
- Rung 2: Reaction. Satisfaction, NPS, "would recommend." A review, not a result.
- Rung 3: Recall. They passed the quiz. Real, but it's memory: knowing, not doing.
- Rung 4: Behaviour. They're doing the thing differently on the job, and someone can observe it. This is the first rung that counts.
- Rung 5: Result. A business metric moved, and you can show the training moved it. The summit.
Most organisations live on Rungs 1 and 2 and call it ROI. The whole game is climbing to 4 and 5.
If this feels like Kirkpatrick's four levels wearing a new jacket, good eye: it is, just sharpened. Phillips later added a fifth level on top for monetary ROI. The Kirkpatrick model and Phillips ROI methodology have been right for 50 years; we've simply been ignoring the top of the ladder because the bottom is easier to reach.
Rule of thumb: if your headline metric lives below Rung 4, you don't have an ROI story. You have an attendance register.
The isolation test: where learning ROI numbers get inflated
This is where most ROI fraud actually happens: not in the data, but in the attribution.
Your sales rose 12% after the training. Great. But you also changed pricing and hired two reps, and it was the festive season. Claiming the training caused all 12% isn't a measurement. It's theft.
Honest measurement asks one brutal question: "What else could explain this number?" Then it subtracts. You don't need a lab. Three cheap ways to isolate the effect:
- A comparison group. Trained cohort versus an untrained one. The gold standard, and often free, since you rarely train everyone at once.
- A trend line. If the metric was already climbing 2% a quarter before training, don't claim that 2%.
- Estimation, honestly labelled. Ask managers to estimate the training's share, then, using the Phillips trick, discount it by their confidence level. A 60%-confident 40% becomes 24%. Ugly. Honest. Defensible.
Report the range, not the trophy. "Between 4% and 7%, most likely 5%" beats a suspiciously clean "12%" every time. Showing your working is what separates an ROI number from an ROI claim.
The one shift that fixes most of the fakery
Stop measuring whether people attended. Start measuring whether they can do the thing.
That's the leap from Rung 3 to Rung 4, and it lives or dies on one word: assessment. A completion certificate says someone was present. A graded demonstration of competency says someone is capable. The moment your learning is tied to real, verified proof of skill, your ROI numbers stop being vibes and become defensible.
This is exactly why assessment-backed skilling beats a shelf full of content, and it's the principle we built LiveSkills around: skilling you can prove, where competency is verified rather than assumed. When the certificate at the end means the learner cleared a real bar, "we trained 200 people" quietly becomes "we have 200 people who demonstrably can," and that's a sentence you can take to a CFO without flinching.
You don't even need to name a platform to steal the principle: if your programme can't produce evidence that a learner can perform, you're not measuring ROI. You're taking attendance.
A 5-question audit for honest L&D metrics
Before your next learning report goes out, put every headline number through this:
- Would it survive a sceptical CFO asking "so what?" "Engagement was high" fails.
- Do I have a baseline? No before, no improvement claim.
- Can I name a behaviour that changed? Observable, not self-reported confidence.
- Have I isolated the training from everything else? Even roughly.
- Is there proof of capability, not just proof of presence? An assessment, a work sample, a live demo.
Fewer than three passes? Keep that metric in the appendix, not the headline.
Measure less. Prove more with LiveSkills. And make sure the certificate at the end actually means something - that's the standard we hold ourselves to at LiveSkills, and the one we'd encourage every L&D team to demand.
Upskill your teams with LiveSkills today!