Costs and pricing
Part of How to build a paid-social budget from dated evidence, not benchmarks
Why paid social ROI stays blank until the counterfactual is defined
Calculate paid-social ROI only after defining full economic cost, incremental contribution, a credible counterfactual, uncertainty and decision thresholds.
An ROI figure is causal only when its incremental contribution has a credible estimate of what would have happened without the campaign. Dividing platform-attributed value by media spend answers a different question. It also omits agency work, creative, measurement, control and exit costs.
What to take away
- An ROI figure is causal only when it includes a credible estimate of the no-campaign outcome.
- Full economic cost covers agency, media, creative, measurement and internal time.
- Leave the ROI formula blank until the counterfactual and full economic cost pass evidence checks.
- Set decision thresholds before examining the result and name who may continue, restrict, pause or stop.
Define the outcome outside the platform
Choose an outcome recorded in the business system and specify its population, England or wider geography, period, currency, returns, cancellations, variable costs and data owner. Decide whether the calculation uses revenue, gross contribution or another value. Do not switch between them after seeing the result.
Define outcome before analysis
- Population and geography
- Period and currency
- Returns and cancellations
- Variable costs
- Data owner
- Revenue or gross contribution
Keep platform reporting as a separate evidence stream. Google's current attribution-model guidance describes attribution as assigning conversion credit among advertising interactions. That is a first-party account method. It does not independently establish the no-campaign outcome.
Reconcile event counts, timestamps, currency, attribution window, model, exclusions and duplicates before analysis. Preserve the source export and configuration date.
Specify the counterfactual
Write the question plainly: what contribution would the eligible population have produced in the same period without this paid-social activity? Pre-period sales alone may be affected by seasonality, other marketing, price, stock or wider demand.
Is a credible counterfactual available?
Can the comparison answer the same outcome question?
Proceed with incremental analysis
Report observed and platform-attributed results only
The 2026 Magenta Book discusses experimental, quasi-experimental and theory-based evaluation. The accompanying quality guide explains that outcome monitoring does not by itself show that an intervention caused the change. These are UK government evaluation sources, not proof that a commercial campaign used a valid design.
Concrete designs include a geo holdout, a randomised controlled trial at customer or account level, and a matched control period with a pre-registered comparison.
A qualified analyst should select a proportionate approach before launch. Record eligibility, allocation, comparison group or time period, contamination, other interventions, missing data and assumptions. If no credible counterfactual is available, report observed and platform-attributed results without the word incremental.
Test whether the proposed comparison could answer the same outcome question. Check that both groups or periods use the same definitions, observation window and contribution treatment. Record exposure that crosses between them, changes in stock or pricing, and other campaigns. If those differences cannot be measured or bounded, lower the strength of the conclusion before anyone sees the result.
Include the full economic cost
Use the same period and scope on both sides:
Full economic cost = agency + media + creative + measurement + internal time + specialist controls + correction + renewal and exit allocation
Every input needs currency, VAT status, unit, quantity, period, source date, inclusions, exclusions, confidence and owner. Finance should decide tax treatment and whether amounts are cash or economic cost. Do not replace missing lines with a percentage allowance.
UK cost lines are usually quoted in pounds sterling. A paid-social retainer is commonly a fixed monthly fee separate from media, while measurement tooling is often a separate licence or subscription.
The amount depends on scope, channel count, creative volume, reporting cadence and tool tier. Ask for the retainer, media fee model and tool licence in writing. No single public benchmark covers every scope, so compare written quotes for the same brief.
Leave the ROI formula blank until evidence passes
Incremental contribution = observed contribution - counterfactual contribution
ROI = (incremental contribution - full economic cost) / full economic cost
Illustrative example, not a benchmark: observed contribution is £180,000; the counterfactual estimate is £120,000; incremental contribution is £60,000. Full economic cost is £45,000. ROI is (£60,000 - £45,000) / £45,000 = 0.33, or 33 per cent.
State whether the ratio is prospective, observational or causally estimated. Show the numerator and denominator, not only the percentage. The current Green Book stresses risk, uncertainty and options in public appraisal. It does not make this private calculation a government-approved business case, but it supports transparent sensitivity work.
Vary one uncertain input at a time and explain its evidence range. Add confidence intervals where the design supports them, plus practical uncertainty from returns, delayed sales, missing identifiers and cost allocation. Do not publish a precise point estimate when the method cannot justify it.
Set decision thresholds before examining the result. Name who may continue, restrict, pause or stop, and when the calculation will update. Finance, statistical evaluation, advertising, privacy or PECR and regulated-product reviewers should clear their own parts. A platform dashboard can inform that review, but it cannot serve as the counterfactual.
UK measurement must respect PECR and ICO guidance on consent for cookies, pixels and similar tracking. The ASA's CAP Code expects evidence for performance claims made in paid social. Record the consent basis, the lawful basis for personal data and who signs off any public claim.
Before you act
- Agree the business outcome and value definition with finance.
- Freeze and version the platform export with its configuration date.
- Pre-register the counterfactual design, comparison unit and success threshold.
- Collect written quotes for agency, media, creative, tooling and internal time.
- Label the estimate prospective, observational or causal before sharing.
- Record the named decision owner, review date and update trigger.
Common questions
Why does dividing platform-attributed value by media spend not give a causal ROI?
It answers a different question and omits agency work, creative, measurement, control and exit costs. The quotient is a media efficiency ratio, not a causal ROI.
What should be recorded when specifying the counterfactual?
Record eligibility, allocation, comparison group or time period, contamination, other interventions, missing data and assumptions. A qualified analyst should select a proportionate approach before launch. If no credible counterfactual is available, report observed and platform-attributed results without the word incremental.
What belongs in the full economic cost calculation?
Full economic cost equals agency, media, creative, measurement and internal time. It also includes specialist controls, correction, and renewal and exit allocation.
Every input needs currency, VAT status, unit, quantity and period. Each line also needs source date, inclusions, exclusions, confidence and owner. Finance should decide tax treatment and whether amounts are cash or economic cost.
