Return on corporate communications: write the counterfactual before the result

Corporate communications return on investment is calculable only after the buyer defines both eligible financial value and full cost. A published statement, employee message, mention, click or enquiry is not itself cash. This method uses blank fields for Cedar Vale Engineering Ltd's fictional England site-consolidation communication and makes no return claim.

The review date is 6 September 2026. It applies evaluation discipline to one private buyer decision; it does not turn government guidance into a company accounting rule.

What to take away

A return figure needs both eligible financial value and full cost defined before any calculation.
Keep observed values, attributed values, scenario outputs and causal increments in separate columns.
Write the counterfactual before the result so alternative explanations can be tested.
A favourable ratio does not prove causal impact and cannot override privacy, legal or security failures.
Without reconciled buyer inputs and a defensible comparison, withhold any return or causation conclusion.

Record what happened before assigning credit

Start with observed data under named ownership. Finance records eligible cash inflows, cash outflows, credits and write-offs for the stated period. Communications records controlled outputs and corrections. Analytics records defined events with missingness and deduplication. None of those records explains cause on its own.

Four columns of evidence

Observed

Source: Reconciled
Ownership: Finance
Move rule: No

Attributed

Source: Rule
Ownership: Comms
Move rule: No

Scenario

Source: Assumptions
Ownership: Analytics
Move rule: No

Causal

Source: Counterfactual
Ownership: Reviewer
Move rule: Reviewer + evidence

Keep four columns:

Record what happened

observed value from a reconciled source;
value assigned by a stated attribution rule;
scenario output created from explicit assumptions;
causal increment estimated against a credible counterfactual.

Do not move a value between columns without a reviewer, method and evidence ID.

Define cost consistently

The denominator must name currency, VAT status, period and included activities. Include supplier cash, internal labour, implementation, content production, accessible formats, rights, legal and privacy review, security, monitoring, correction, incident response, maintenance and exit where relevant. State exclusions beside the formula.

Cost denominator must name

Currency and VAT status
Period covered
Included activities
Supplier cash and internal labour
Implementation and content production
Accessible formats, rights, legal, privacy
Security, monitoring, correction, incident, exit

Cost sanity check

Most programmes spread cost across supplier fees, internal labour, licences, production, accessibility, review and exit. The split varies with scope, so no single range fits every buyer. Test the denominator by asking whether internal labour, correction work and exit are inside it. A denominator that omits them will usually understate full cost.

HM Treasury's Green Book 2026 distinguishes appraisal from evaluation and calls for options, risk and uncertainty to be considered in government decisions. Applied to Cedar Vale, that means costing more than one option and recording which assumptions would change the choice.

Write the counterfactual before the result

The question is not merely what followed communication. It is what would plausibly have happened under the chosen comparison, such as the unchanged manual route or a different authorised intervention. Record the comparison population, period, exposure rule, other changes, missing data and reasons the groups may differ.

Can you claim causation?

Does the counterfactual address material alternative explanations?

Yes

Report causal increment with evidence

Stop causal claim; report descriptive or attributed value

The Magenta Book, updated 15 May 2026, is central-government evaluation guidance. It covers scoping, design, conduct and interpretation, and says evaluation should be considered through design and delivery. Its frame treats evaluation as designed in from the start, so comparison groups are fixed before launch.

Stop a causal claim when the counterfactual cannot address material alternative explanations. The buyer may still report descriptive observations or a transparent attribution rule, labelled correctly.

Use blank formulas

observed financial balance = eligible realised cash inflows - eligible realised cash outflows

Four blank formulas

Observed financial balanceeligible realised cash inflows - eligible realised cash outflows
Attributed contributionobserved eligible value x documented attribution share
Estimated causal incrementobserved outcome - estimated counterfactual outcome
Financial return ratio(eligible value - full eligible cost) / full eligible cost

attributed contribution = observed eligible value x documented attribution share

estimated causal increment = observed outcome - estimated counterfactual outcome

financial return ratio = (eligible value - full eligible cost) / full eligible cost

Every term needs a unit, period, population, source, owner and uncertainty. The ratio is undefined when the denominator is absent, inconsistent or zero. Tax and accounting treatment requires qualified review.

Worked example with illustrative placeholders

All amounts below are illustrative placeholders in pounds sterling, not Cedar Vale data. Cedar Vale's real inputs must come from its own reconciled records.

Eligible realised cash inflows: £240,000. Eligible realised cash outflows: £40,000. Observed financial balance: £200,000.

Documented attribution share: 0.30. Attributed contribution: £200,000 x 0.30 = £60,000.

Observed outcome: £200,000. Estimated counterfactual outcome: £150,000. Estimated causal increment: £50,000.

Full eligible cost: £80,000. Financial return ratio: (£60,000 minus £80,000) / £80,000 = -0.25.

At an attribution share of 0.55, attributed contribution is £110,000, giving a ratio of 0.375.

How to set the attribution share

The share must come from a written rule agreed with finance, not from preference. What determines it: which events the rule credits, the claim window, and whether a sales owner confirms the deal.

Sources include platform conversion reports, CRM owner confirmation and matched control groups. Version the rule, record the reviewer, and do not reuse another programme's share without testing it against this evidence.

Test sensitivity and the decision

Vary one uncertain assumption at a time: eligible value, attribution share, counterfactual outcome, internal labour, correction work or exit cost. Do not insert illustrative numbers. Record the switching value at which the decision changes and the evidence that would narrow uncertainty.

A scenario result is not a forecast. Accessibility and factual-approval checks sit outside any ratio.

The final record should say proceed, hold or stop for the specified decision only. Record the reviewer, method and date beside it.

Before you act

Name currency, VAT status, period and included activities for the denominator.
State exclusions beside the formula.
Record the comparison population, period, exposure rule and other changes.
Give every term a unit, period, population, source, owner and uncertainty.
Vary one uncertain assumption at a time and record the switching value.
Label descriptive observations and attribution rules correctly.

Common questions

What must be defined before a corporate communications return can be calculated?

The buyer must define both eligible financial value and full cost. A published statement, employee message, mention, click or enquiry is not itself cash. The method uses blank fields for Cedar Vale Engineering Ltd's fictional England site-consolidation communication and makes no return claim.

Why should the counterfactual be written before the result?

The question is what would plausibly have happened under the chosen comparison, such as the unchanged manual route or a different authorised intervention. Record the comparison population, period, exposure rule, other changes, missing data and reasons the groups may differ. Stop a causal claim when material alternative explanations cannot be addressed.

What should the final record say about the decision?

The final record should say proceed, hold or stop for the specified decision only. A scenario result is not a forecast. It does not settle causation on its own.