Measuring Training ROI Without Making the Numbers Up

Training ROI calculations are often works of creative fiction. Someone multiplies an estimated productivity gain by a headcount, subtracts the course cost, and produces a percentage nobody believes — including them.

There is an honest version. It's less impressive and considerably more useful.

The formula, and its problem

ROI = (benefit − cost) ÷ cost, expressed as a percentage. The cost side is genuinely knowable. The benefit side is where the fiction enters, because attributing an organisational outcome to training specifically is hard and usually impossible to do cleanly.

Get the cost right first

Most calculations understate cost by ignoring the largest component. Include:

• Development or purchase cost.
• Delivery — facilitator time, venue, platform fees.
• Learner time. 500 people × 45 minutes at loaded salary cost. This usually dwarfs everything else and is the number most often left out.
• Management and administration time.

Being rigorous about cost buys you credibility for whatever you claim about benefit.

What can be measured honestly

Directly attributable

• Avoided cost — a penalty not incurred, a certification maintained, an audit passed. Clean when it's genuinely contingent on the training.
• Displaced spend — bringing training in-house from an external provider. Simple arithmetic.
• Time saved — where you can measure task duration before and after.

Reasonably attributable, with care

• Error and incident rates — measurable, but confounded by everything else happening.
• Time-to-competence for new starters — good when you have a defined standard and a baseline.
• Reduction in help-desk or supervisor queries on the trained topic.

Not honestly attributable to training alone

• Revenue growth.
• Staff engagement or retention.
• Culture change.
• Anything where a dozen other initiatives ran concurrently.

You can report movement in these. You cannot claim your course caused it, and claiming it is how L&D loses credibility with finance.

The one thing that makes this possible

A baseline, captured before the training. Almost every honest ROI story depends on knowing what the number was beforehand — and almost every dishonest one exists because nobody measured first.

If a sponsor wants ROI, the conversation happens at commissioning, not at the end: what number are we trying to move, what is it now, and when will we look again? Retrofitting a baseline is impossible.

Isolating the effect, roughly

Where a clean comparison is impossible, partial approaches help: a staggered rollout creating a natural comparison group; trend analysis showing whether the change breaks an existing trajectory; or asking managers to estimate the training's contribution and stating plainly that it's an estimate.

None is perfect. All are better than asserting causation, and all survive scrutiny better because you've named the limitation yourself.

How to present it

Give the stakeholder three things: what you measured and what it shows; what you can reasonably attribute and what you can't; and what you'd need in order to say more. Presenting a confident single figure that collapses under one question does more damage than a modest, defensible one.

Confidence you can't support is worth less than a smaller number you can defend.