Assessing marketplace software return from an observed baseline and net contribution

Marketplace seller software return on investment cannot be calculated from gross marketplace revenue or a supplier dashboard. For one England seller, define the eligible product and order population, observe the current route, estimate causal incremental contribution and subtract reviewed whole-life economic cost.

Worked figures below are illustrative placeholders. Research closed on 6 September 2026. This is a method, not financial or tax advice and not a return promise.

What to take away

Freeze the observed baseline

Choose a period, Europe/London timezone, authorised account and stable product population. Define each event and its source. Keep listing publication, stock update, order placement, seller acceptance, payment authorisation, capture, fulfilment, retention, cancellation, return, refund, dispute and settlement apart.

Record eligible events, missing or late data, duplicates, staff activity, reversals and corrections. Document staff work using actual time records and approved cost sources. A baseline with changed products, prices or fulfilment conditions may not be comparable. Reconcile settled cash to VAT records and marketplace settlement reports.

Separate four kinds of evidence

Observed evidence is what the seller's reconciled systems recorded. Attribution assigns an event to a channel or tool under a stated rule. A buyer scenario combines explicit assumptions. Causal evidence estimates what changed because the intervention occurred.

HM Treasury's QPIE guidance is a methodological check, not a merchant software rule. It asks for a counterfactual and a proportionate design. A practical test: run causal evidence when the whole-life cost is large enough that being wrong would outweigh the cost of the test. If not, report cost and operational outcomes without an ROI claim.

Define benefit as contribution, not receipts

An additional order is not automatically a benefit. Start from cash settled for eligible retained orders, then deduct refunds, disputes, taxes on the reviewed basis, product cost, fulfilment, customer support and other variable costs. Keep this net contribution separate from accounting profit.

Possible avoided labour needs an observed task reduction and evidence that released capacity had another useful application. Prevented error or risk remains a scenario unless the event frequency, consequence, counterfactual and uncertainty are supported. Absence of a complaint or incident does not by itself show a saving.

Use blank equations:

Incremental contribution = [net contribution with intervention] minus [counterfactual net contribution]

Net return = [causal incremental contribution] + [supported avoided economic cost] minus [whole-life economic cost]

ROI ratio = [net return] divided by [whole-life economic cost]

Currency, VAT basis, population, period and uncertainty must accompany every input. Do not calculate when the denominator is zero, negative or unsupported.

Worked example with placeholder numbers. Baseline net contribution: £40,000 over 12 months. With intervention: £46,500. Incremental contribution: £6,500. Supported avoided cost: £500. Whole-life cost: £5,000. Net return: £2,000. ROI ratio: £2,000 divided by £5,000 = 0.40, or 40p per £1 of whole-life cost. All figures are illustrative placeholders.

Include the full cost and failed gates

Whole-life cost includes supplier charges, internal labour, implementation, product-data work, consumer and safety assurance, privacy and PECR review, security, accessibility, support, correction, incident recovery, migration and exit. The current process is an option, not a free benchmark.

HM Treasury's Green Book 2026 is a methodological check on appraisal, lifetime costs, risk and sunk spending. It is not a private investment rule.

Consumer, product-safety, tax, privacy, security, accessibility and exit failures are stop conditions. A positive model cannot buy its way past them.

Decide under uncertainty

Show buyer-owned low, central and high scenarios only after documenting each assumption. Vary one uncertain input at a time and state the value that would change the decision. Recheck when the service edition, meter, volume, product mix, marketplace interface, tax view or test validity changes.

The final record may support proceed, rework, hold or stop.

Before you act

Common questions

How does the worked example calculate the ROI ratio?

Baseline net contribution is £40,000. With the intervention it is £46,500, so incremental contribution is £6,500. Add £500 supported avoided cost and subtract £5,000 whole-life cost to give £2,000 net return. £2,000 divided by £5,000 gives an ROI ratio of 0.40. All figures are placeholders.

What must be documented about the observed baseline?

Record eligible events, missing or late data, duplicates, staff activity, reversals and corrections. Document staff work using actual time records and approved cost sources. A baseline with changed products, prices or fulfilment conditions may not be comparable, so the period, timezone, authorised account and stable product population must be fixed.

When should a seller report cost and operational outcomes without an ROI claim?

When the practical test shows the cost of causal evidence is likely to exceed the value of the decision. The final record may support proceed, rework, hold or stop.