How to Measure L&D ROI: A Practical Guide for HR and L&D Leaders

"If your ROI calculation runs on completion rates and satisfaction scores, you are measuring the training. Not the development."

What I actually see, conference talks aside, is this. Most organisations aren't measuring ROI. They're measuring activity, dressed up nicely.

Hours delivered. Heads through the room. A satisfaction average with a decimal point on it, because decimal points look rigorous. And then the CHRO asks whether the leadership spend worked, and out comes a carefully assembled bundle of proxies that all point in a promising direction while proving absolutely nothing.

Nobody is lying. Everybody knows.

So here's how to measure it in a way you could defend under actual questioning - the standard framework, where it runs out of road, and what to do instead when behaviour is the thing you're buying.

The Kirkpatrick Model: What It Measures and What It Misses

You know this one. Reaction, Learning, Behaviour, Results. It's dominant for two reasons: it's logical, and the first two levels are easy.

Level 1 (Reaction). Did they enjoy it. The most collected number in corporate L&D and the least predictive thing in the building. High satisfaction tracks facilitation quality, room comfort, and whether the session finished on time. It barely tracks change at all.

Level 2 (Learning). Did they take something in. Pre- and post-tests, assessments, reflection sheets. Also popular, also weak. Knowing the correct answer about active listening and actually listening to someone who is annoying you are two different skills, and only one of them is on the test.

Level 3 (Behaviour). This is where it gets hard and where nearly everyone stops. You need follow-up in the real job, by people who can watch the leader lead. Manager observation, 360s, interviews with direct reports. It takes time, it takes coordination, and it has an unfortunate habit of showing the programme did less than Levels 1 and 2 promised. That's the real reason it gets skipped, and everybody knows that too.

Level 4 (Results). Behaviour joined up to business outcomes. Retention, promotion, team performance, how the customers of a participant's team feel compared to a similar team that never went. This is the level the CFO actually cares about, and it is almost never done properly, because attribution is genuinely difficult and the wait is long.

Why Most L&D Measurement Stops at Level 2

It's not laziness. It's incentives.

L&D teams get judged on delivery - hours, heads, scores. Every one of those is a Level 1 or 2 metric. The person who signed off on the programme is almost never the person standing next to the leader six months later, and the systems that would catch it were never built.

So the spend goes out, the programme runs, everyone is pleased. Strong evidence the training happened. Decent evidence people enjoyed it. Almost no evidence anything changed.

A More Honest Measurement Approach

It starts before the programme, not after. Which sounds obvious and almost nobody does it.

Get L&D and the business sponsor in one room and agree three things in advance. What behaviour should change. How anyone would see it. And what amount of change means this was worth the money.

That conversation is harder than it looks, because sponsors often can't say what they want. They want "better leaders" and "stronger collaboration" and "more strategic thinking". Your job at that table is to keep pushing until the words get specific enough to be wrong.

"Better leaders" is not measurable. "Managers who run a structured one-on-one with each report at least fortnightly" is measurable. "Stronger collaboration" is not measurable. "Cross-functional teams that settle resource fights without escalating to a director" is measurable.

It feels pedantic right up until the day you try to prove something happened. Then the pedantry is the only thing keeping you afloat.

Practical Measurement Approaches for Different Interventions

For skills-based training - technical skills, process adoption - pre- and post-assessment plus one check at 60 days will do. If it isn't being used at 60 days, either the learning didn't happen or it didn't travel. Both of those tell you something worth knowing.

For behaviour-focused development - leadership, influence, communication, decisions - you have to go to Level 3, and you need a baseline before you start. 180s or 360s with the participant's team before and again three to six months later. Structured conversations with reports and peers. Manager assessment against named behaviours. Facilitator observation across a series, if there is one.

For team dynamics and culture, accept that the window is longer and attribution is messier. Performance metrics, engagement, retention - all relevant, none clean. You'll want either a comparison group or enough time to tell the signal from the ordinary noise of a company being a company.

The ROI Calculation

If somebody needs a number, the Phillips model bolts a Level 5 onto Kirkpatrick and turns results into money. Formula's the easy bit: ROI (%) = ((Benefits - Costs) / Costs) x 100.

The hard bit is deciding what the behaviour was worth. Putting a rupee figure on a conflict that got resolved instead of escalated, on a resourcing decision made well, on a new joiner who got useful in six weeks instead of twelve. Those numbers do exist. They just don't fall out of any system you already have.

What works: pick one or two behaviours where the money is most visible and calculate only those. A manager who settles cross-functional fights is preventing escalations that eat director and VP hours - and those hours have a rate. Estimate low. Write down exactly how you got there. Present a range, not a point.

Ranges are more honest, and here's the funny part: finance people trust them more than a precise number. A number with three digits after the decimal has never fooled a CFO in the history of business.

What Good L&D ROI Measurement Requires of the Intervention Itself

Last thing, and it's the one that gets missed.

You can only measure an intervention that was built to be measured. If nobody defined a behavioural objective, there is nothing to check against. If the programme was designed around content and engagement, the measurement will report that back to you very accurately: nice scores, unclear outcomes.

Which is why design and measurement are the same conversation. Build a thing around a named, watchable behaviour and it is both easier to measure and more likely to do what you paid for. Serious games start from a specific behaviour pattern they're built to bring out. So the facilitator can see whether it showed up, the debrief can pin it to real work, and the follow-up six months later has one concrete thing to go looking for.

Design the measurement while you design the programme. Do that and ROI stops being the uncomfortable question at the end of the year and starts being just another line in the delivery plan.

If you want to build one with the measurement baked in from the start, come and talk to me. Or have a look at the games and the specific patterns each one is built to surface.

With love from Bengaluru, this is Arvindh saying over and out.