A training ROI figure that holds up in the room

Most published training ROI figures do not survive one question from a finance director. The question is always some version of how do you know it was the training?, and it is fair, and the usual answer — a confident percentage with no visible working — makes it worse rather than better.

There is a way to build a figure that holds. It is not more sophisticated than the usual method. It is less sophisticated, and it works because it makes the assumptions loud instead of hiding them inside a calculation.

The shape of the claim

Every return figure is the same four steps.

1. A behaviour changed.
2. That behaviour is connected to a business number.
3. Some part of the change in that number is attributable to the intervention.
4. That part, converted to money, is compared to what the intervention cost.

Nearly all the credibility lives in step three. Steps one, two and four are arithmetic. Step three is a judgement, and the difference between a figure that holds and one that does not is whether the judgement is stated or smuggled.

Isolate before you convert

The cheapest form of isolation is a comparison group, and it is more often available than people assume. Most programmes roll out in waves. Cohort one goes in March and cohort two in September, which gives you six months where one population has been through the session and a comparable one has not.

If the waves were not planned as an experiment they will not be clean. Say that too. A rough comparison group, described honestly, is worth more than a precise figure with no comparison at all.

Where there is genuinely no comparison group, the fallback is estimation with named sources. Ask the participants' managers, separately, two questions: how much of the change in this number do you attribute to the programme, and how confident are you in that estimate. Multiply. A manager who says "sixty per cent, and I am about half confident" has given you thirty per cent, and they have given it to you in a form you can attribute to them by name.

That is a weak method. It is a weak method that is visibly weak, which is the property that matters.

Convert conservatively, and show the rate

Converting behaviour to money needs a rate: what an hour of rework costs, what an escalation costs to handle, what a week of cycle time is worth. The organisation usually has these numbers somewhere in finance, and the ones finance already uses are the ones to use, even where you think they are wrong.

Where you have a range, take the bottom of it. A figure built on the most conservative available rate is much harder to argue with than one built on the most plausible, and the difference is rarely large enough to change the decision.

Count the whole cost

The denominator is where credibility quietly leaks. The invoice is not the cost. The cost is the invoice plus the participants' time at a loaded rate, plus travel, plus the venue, plus the internal hours spent organising it.

Participant time usually dominates. Two hundred people out for a day is a large number before anyone has been paid for anything, and leaving it out is the single most common reason a published ROI figure reads as promotional.

Putting it in has a useful side effect: it makes the group-size decision visible as a cost decision rather than a logistics one.

State the assumptions where they can be checked

A defensible return figure is a paragraph, not a number. The paragraph says: we compared these two cohorts; the number moved by this much; we attribute this proportion, on this basis; we valued it at this rate, which is finance's own; the total cost including participant time was this; here is the ratio; and here is the one assumption that, if wrong, changes the answer most.

That last clause is what converts the document from marketing into analysis. Every model has a load-bearing assumption. Naming yours before the finance director finds it is the difference between a discussion about the programme and a discussion about your credibility.

When not to do this at all

If the behaviour is genuinely hard to connect to a number — most leadership development, most culture work — an ROI figure is a fiction with a decimal point. Do not build one.

Report level 3 instead: the behaviour changed, here is who observed it, here is what they observed, here is the trace it left. That is a real finding. A sponsor who wanted a return figure and receives a well-evidenced behavioural one is usually satisfied, because what they actually wanted was to be able to defend the spend, and a specific observed change defends it better than a percentage nobody believes.