Announcement return on investment needs a counterfactual before it needs a formula

An ROI calculation is meaningful only after the buyer defines the money, period, eligible outcome and comparison. This article gives no figures and no benchmark because no campaign or finance records were supplied.

What to take away

Label four different claims

Observed publication is a dated record that an identified page or outlet carried defined content. Supplier attribution is an event assigned under a supplier's stated rule. Reconciled financial outcome is a qualifying inflow or avoided outflow matched to finance records. Causal effect is the difference the activity made compared with a credible counterfactual.

A supplier receipt, published page, index entry, monitoring result or media mention is not revenue.

Report these in separate fields. Do not upgrade one into another. Independent editorial publication remains an editor's decision even when it occurs after transmission. Paid placement must be labelled and costed on its own route; CAP's recognition rules address identification of marketing communications and advertorials.

Define the accounting view

Choose cash or an approved accounting measure before calculating. Record currency, VAT basis, start and end dates, payment timing, refunds, credits, variable fulfilment cost and internal labour. HMRC's VAT rates page shows why the tax treatment of the actual supply needs review rather than assumption.

Use this blank formula only if the terms match the decision:

cash return ratio = (eligible reconciled net cash inflow - programme cash outflow) / programme cash outflow

The numerator is buyer-defined cash, net of specified refunds and incremental fulfilment costs. The denominator includes approved supplier cash and other programme cash outflows on the same VAT basis and over the same period. Full economic return needs a separately documented labour and internal-resource measure. Never mix that denominator with a cash-only numerator without labelling the mismatch.

Worked example (hypothetical, no supplied records). Money: net cash inflow from new subscriptions. Period: 1 March 2025 to 31 March 2025. Eligible outcome: reconciled net cash inflow. Comparison: matched pre-campaign quarter. Currency: GBP.

Programme cash outflow is £10,000. Eligible reconciled net cash inflow is £40,000. The ratio is (£40,000 - £10,000) / £10,000 = 3.0.

A second case with £8,000 inflow gives (£8,000 - £10,000) / £10,000 = -0.2, a negative return. A third case with £10,000 inflow gives zero return. These figures are illustrative only and are not a benchmark.

If the denominator is zero, the ratio is undefined. If outcomes arrive outside the window or in another currency, state the treatment rather than silently moving or converting them. A finance and tax reviewer should approve the populated definitions.

Where to source outcome data

Populate buyer-owned inputs from three records. Web or product analytics gives sessions, downloads and form starts. A CRM gives lead, opportunity and closed-won stages with dates. The finance ledger gives invoiced cash, refunds and fulfilment costs. Use one consistent ID to join them. Common tools include Google Analytics 4, HubSpot or Salesforce, and Xero or QuickBooks.

Name the source beside every amount.

Build the counterfactual before interpreting change

The comparison question is: what would probably have happened to the same eligible outcome without the activity? A before-and-after difference alone may also contain seasonality, other campaigns, product changes, media events or sales activity.

HM Treasury's updated Magenta Book covers evaluation scoping, methods and interpretation for government. Its principles can discipline a private buyer's design, but they do not validate a PR claim. Name the comparison population or time series, assignment process, confounders, missing data and minimum evidence needed. Stop causal language when those assumptions fail.

Show uncertainty and sensitivity

Create low, base and high cases for buyer-owned inputs: eligible cash inflow, attribution share if used, fulfilment cost, programme outflow, lag and missingness. They are decision sensitivities, not industry forecasts. Change each uncertain field independently, then inspect plausible combinations.

If attribution share is used, choose a rule before results. Methods include last touch, first touch, linear and position-based. Published benchmarks vary by sector, so test illustrative sensitivities such as 10%, 25% and 50%.

Keep an evidence ledger linking each amount to an invoice, accounting record, approved time record or disclosed model. A monitoring dashboard may support an observed-output row, but it cannot fill the cash numerator. Do not use advertising-value equivalence or a transmission count as money.

Set a decision rule

Before observing results, write the review date, minimum data completeness, acceptable uncertainty, threshold action and stop authority. Pause interpretation if transaction reconciliation fails, the counterfactual is contaminated or a supplier changes its attribution rule.

Beside every result, the final record should say "observed", "attributed", "modelled" or "causal estimate". This draft promises neither a return nor a valid causal estimate.

Before you act

Common questions

Why is a before-and-after difference not enough to prove effect?

It cannot separate the activity from seasonality, other campaigns, product changes, media events or sales activity.

What should happen if the denominator is zero or outcomes fall outside the window?

The ratio is undefined. If outcomes arrive late or in another currency, state the treatment; do not move or convert them silently.