The ROI of Compliance Training: Why the Number Is Wrong and What to Measure Instead

Return on investment is the wrong frame for most compliance training, and pretending otherwise is why so many training proposals lose credibility with finance. The honest position is that the benefit is largely unmeasurable, that the few available benchmarks are unreliable, and that there is a better discipline available — one a regulator has demonstrated at considerable expense.

A worked example of how badly these forecasts miss

In 2021 the FCA introduced rules ending price walking in home and motor insurance, supported by a cost-benefit analysis projecting a £4.2bn saving to consumers across both markets over ten years.

In July 2025 the FCA published an evaluation of what actually happened. In motor insurance it found a statistically significant average reduction of £6.63 per policy, giving a central ten-year estimate of £1.59bn within a range of £163.2m to £3.02bn. In home insurance it found no statistically significant effect and estimated no monetary saving at all.

So a well-resourced regulator, forecasting a change it designed and controlled, with access to the entire market’s pricing data, overestimated by roughly a factor of three in one market and could not measure an effect in the other. It then published the finding.

That is the benchmark against which an internal training ROI model should be judged. If the FCA cannot forecast the benefit of its own intervention within an order of magnitude, a spreadsheet projecting a percentage uplift from a compliance course is not a forecast. It is a decoration.

Why the standard ROI models do not transfer

Three structural problems defeat training ROI in a regulated setting.

The counterfactual is unobservable. ROI requires knowing what would have happened without the training. In compliance the desired outcome is usually an absence — no breach, no misselling, no safeguarding failure — and absences cannot be attributed. You cannot evidence the incident that did not occur.

The training is mandatory. Where a rule requires it, there is no do-nothing option to compare against, so there is no return to calculate. Asking for the ROI of a legally required control is like asking for the ROI of filing accounts.

Confounding is total. Compliance outcomes move with supervisory attention, management changes, product mix, market conditions and sheer luck. Isolating a training effect from that would require a control group a regulated firm cannot ethically or lawfully run.

What can actually be measured

Four things, all of which are real numbers rather than modelled ones.

1. Evidence coverage. The proportion of your regulatory training obligations you can evidence on demand, measured before and after. This is a count, not an estimate, and it is what an inspection tests. The specific questions are: can you produce the requirement per role and when it was last reviewed; the gap between required and completed; evidence that someone acted on a gap; and current supervision status.

2. Time to competence. Where staff must be supervised until assessed as competent, the time taken to reach that assessment has a payroll cost attached. Reducing it is a measurable productivity gain with a defensible calculation, and nobody usually tracks it.

3. Remediation hours avoided. If you have ever reconstructed training records retrospectively, you have an actual figure for what that cost in senior compliance time. Use your own historic number, not an industry one.

4. Deadline compliance. Where a population faces a fixed qualification date — Irish credit union staff must hold a recognised qualification by 1 October 2028, with no grandfathering — the measure is simply whether you will make it, and what the shortfall is today. Binary, dated and unarguable.

Three numbers to refuse to produce

Avoided fines. The FCA published 19 fines in 2026 across a sector of thousands of firms, and 17 were against individuals rather than firms. No per-firm probability can be derived from that, and a model that does will be dismantled by the first person who checks.

Industry productivity multipliers. A published figure for the return on training spend in general is not evidence about your organisation, your workforce or your obligations.

Engagement as proxy for outcome. Completion rates and satisfaction scores measure activity. CQC’s guidance expects completed and required training to be tracked with action taken on gaps — which means the regulator itself treats completion alone as insufficient.

Reframe the question

The productive question is not “what is the return?” but “what is the exposure, and what does closing it cost?” That has answers.

The exposure is specific and often personal: of the FCA’s 2026 fines, most were against individuals for integrity, due skill, care and diligence, and fitness and propriety, typically accompanied by prohibition. CQC’s fixed penalties attach to the registered manager as well as the provider. Naming that exposure, listing the obligations you cannot currently evidence, and costing the work to close them produces a proposal built on facts.

Our guide to building the business case sets out that structure, and our guide to what regulators actually imposed in 2026 provides sourced figures to use in place of invented ones. Our CPD course library covers the training itself.

Frequently asked questions

Can you calculate ROI on compliance training?

Not credibly in most cases. The counterfactual is unobservable because the desired outcome is an absence, the training is often mandatory so there is no alternative to compare against, and compliance outcomes are confounded by supervision, management and market changes.

What should be measured instead?

Evidence coverage against your regulatory obligations, time to competence where supervision applies, remediation hours drawn from your own historic experience, and progress against any fixed qualification deadline.

Why not use avoided fines?

Because the data does not support it. The FCA published 19 fines in 2026 across a sector of thousands of firms, 17 of them against individuals. That supports no per-firm probability estimate.

Is there evidence that compliance cost-benefit forecasts are unreliable?

Yes, from a regulator. The FCA’s cost-benefit analysis for its general insurance pricing rules projected £4.2bn in consumer savings over ten years; its own July 2025 evaluation found a £1.59bn central estimate in motor and no statistically significant effect in home insurance.

Does this mean training spend cannot be justified?

No — it means it should be justified as risk and obligation management rather than as an investment with a return. Where a rule requires competence and you cannot evidence it, the justification is the gap, not a projected yield.