Costs and pricing
Part of A measurement budget method that keeps three separate ledgers and tests its benefits
Measurement return on investment: the causal share at which spending stops
Work out measurement return on investment from incremental contribution, then find the causal share where the result flips and set a stop rule.
What to take away
- Return on investment is incremental contribution plus evidenced cashable savings, less whole-life economic cost, divided by that cost.
- The causal share, the part of observed revenue that the change really caused, moves the result more than any other input.
- The switching value is the causal share at which ROI equals the decision threshold.
- Report the switching value beside ROI, and cap spending when one unverified assumption carries the result.
Calculate return on investment from net benefit
Return on investment is net financial benefit divided by cost, with both in pounds over the identical period. Revenue shown in an analytics report is not the numerator. Only incremental contribution counts, because it is the margin on revenue the change caused, after variable cost.
ROI (%) = ((incremental contribution + evidenced cashable savings - whole-life economic cost) / whole-life economic cost) x 100
Incremental contribution (£) = eligible incremental revenue (£) x contribution margin (%) - incremental variable cost (£)
Use eligible incremental revenue only after the evaluation method supports it, and apply the organisation's approved contribution margin. Exclude attributed revenue without a counterfactual, transfers from another channel and unvalidated modelled uplift. The whole-life economic cost in the denominator comes from the budget, not from this calculation.
Work one example: a checkout event fix
The figures below are illustrative only, not a benchmark or a claim about a real programme. They show how the pieces combine: contribution first, then savings, then cost.
- Eligible incremental revenue is £120,000, the approved contribution margin is 40% and incremental variable cost is £8,000. Incremental contribution is £120,000 x 40% - £8,000 = £40,000.
- Evidenced cashable savings from fewer manual reconciliations are £6,000.
- Whole-life economic cost is £30,000 over 12 months.
- ROI = ((£40,000 + £6,000 - £30,000) / £30,000) x 100 = 53.3%.
That 53.3% assumes the whole of the £120,000 was caused by the fix. In practice only a share of it is. The rest of the method asks how small that share can be before the answer changes.
Find the causal share where the decision flips
Treat the causal share, s, as the unknown. If the fix caused only share s of the £120,000, incremental contribution is 48,000s - 8,000, and ROI = (48,000s - 32,000) / 30,000. Set ROI at a placeholder decision threshold of 30% and s = 85.4%. That is the switching value.
Below a causal share of 85.4%, the programme fails that threshold. The share is the least certain input, because it depends on a comparison that someone must design and defend, so a modest error in it can change the verdict quickly.
Stress the other assumptions as well
Run low, central and high cases for contribution margin, causal share, adoption, delivery delay, staff effort, recurring charge and exit cost. These are bounded conditions, not forecasts. A case failing only in the low column is a risk to monitor; one failing centrally needs a new plan.
Stress the assumptions
| Assumption | Low | Central | High |
|---|---|---|---|
| Contribution margin | 30% | 40% | 45% |
| Causal share | 50% | 85% | 100% |
| Delivery delay | 0 months | 2 months | 6 months |
| Staff effort | 40 hours | 80 hours | 140 hours |
| Exit cost | £0 | £5,000 | £15,000 |
The central causal share of 85% sits just under the switching value of 85.4%, so the central case narrowly misses the 30% threshold. A result that close to the line calls for a test of the share, not a confident verdict.
The Green Book is UK government guidance on appraisal. Government appraisers use it to confront uncertainty and optimism bias, and a private team can copy that caution without importing discount rates. The stress table applies the same caution to a single measurement change.
Fix the counterfactual before the money is spent
Fix the counterfactual before the change starts, because a causal share can only be defended if the comparison was planned. The counterfactual is the result expected over the same period without the measurement change. A before-and-after difference can reflect seasonality, media, price, product, competitor activity or data-definition changes.
The Magenta Book is HM Treasury guidance on designing an evaluation. Monitoring an outcome is not the same as determining whether an intervention caused it, and attribution output is not causal proof on its own. Check what each attribution method really means before choosing a causal share.
The evidence plan fixes the eligible population, outcome definition, observation window, comparison group and treatment of missing records. Preserve the baseline code and report version so the same calculation can be rerun when someone challenges the share. Crediting a channel is a different claim from showing it was incremental, a slip the reporting mistakes list describes.
Write the stop rule into the decision line
Report the contribution and cost components beside ROI, not only the percentage, and include the low case and the switching value. If the result depends on one unverified attribution assumption, say so in the decision line and cap further spending until the planned test resolves it.
- Name the operational change and its financial effect.
- Use eligible incremental revenue and the approved contribution margin.
- Record the switching value next to ROI.
- State the low case and which assumption it stresses.
- Stop spending when one unverified assumption carries the result.
Keep the calculation in the planning template record, so the decision, the threshold and the evidence plan travel together. Exit cost and the other items a licence quote leaves out belong in the denominator, and the hidden cost categories list what to model.
Common questions
What is the correct numerator for measurement ROI?
Incremental contribution plus evidenced cashable savings, less whole-life economic cost. It is net financial benefit in pounds, not revenue displayed in an analytics report, and the denominator covers the same period.
What is a switching value?
It is the level of one stressed assumption at which ROI equals the decision threshold. In the worked example, a causal share of 85.4% is the switching value for a placeholder threshold of 30%.
Why is a before-and-after difference not enough to prove causation?
It can reflect seasonality, media, price, product, competitor activity or data-definition changes. A counterfactual fixed before the change starts is what lets the team defend a causal share.
What should happen when ROI lands just under the threshold?
Treat it as a prompt to test the causal share, not as a settled failure. In the example, the central causal share of 85% sits just under the 85.4% switching value, so the team records the switching value and caps spending until the planned test resolves the question.
