Costs and pricing
Part of How to structure and compare influencer agency quotes
Influencer-agency ROI from full economic cost and defensible incremental contribution
Calculate influencer-agency ROI only from full economic cost and defensible incremental contribution, with attribution and uncertainty kept visible.
Influencer agencies return on investment cannot be read from a platform dashboard. Calculate it only with a full economic cost, a defined contribution measure and credible evidence of what would have happened without the activity. Otherwise report observed or attributed outcomes without a causal label.
What to take away
- Influencer agency ROI needs full economic cost, a defined contribution measure and credible counterfactual evidence.
- Platform attribution and causal effect are separate evidence classes, and moving between them requires evidence.
- Gross sales and contribution are not interchangeable, so define value before subtracting cost.
- Choose the evaluation design before publication and keep its feasibility cost in the budget.
- If the counterfactual fails, retire the causal figure rather than downgrading the caveat.
Define the decision and population
Write down the campaign, England audience, eligible orders or other outcomes, comparison period, currency and decision. Fix the source systems, query versions, timezone, exclusions, refunds and data cut-off. Decide whether the analysis concerns a creator, content item, campaign or customer cohort. Mixing units can create a plausible but meaningless ratio.
Four Evidence Classes
Evidence class
- Delivery
- Approved post published
- Platform attribution
- Named window and model
- Operational outcomes
- Eligible net orders
- Causal effect
- Against a counterfactual
What it shows
- Delivery
- Platform attribution
- Operational outcomes
- Causal effect
Movement between these classes requires evidence. A tagged order may be associated with a post, but association alone does not show what the buyer gained because the post existed.
Build the denominator completely
Full economic cost includes agency service, software, creator cash, separately valued products or services, affiliate commission, paid media, production, travel, accessibility, specialist review, staff time, correction and exit. Add irrecoverable VAT only after case-specific advice. Do not deduct rebates or credits until finance has evidence.
HMRC's VAT reclaim guidance connects recovery to registration, business use and records. Its online-platform income page identifies gifts and services as possible creator income. Neither page determines the treatment of a particular campaign. They support recording the transaction before qualified accounting and tax review.
Define value before subtracting cost
Use buyer-approved blank fields:
observed contribution = eligible revenue - refunds - variable fulfilment cost - other specified deductions
Gross sales and contribution are not interchangeable. State whether returns have matured, whether sales taxes are excluded and which costs vary with the event.
For causal analysis:
incremental contribution = estimated contribution with the campaign - estimated contribution under the counterfactual
ROI = (incremental contribution - full economic cost) / full economic cost
The counterfactual might use a randomised holdout or another design the evaluation owner accepts. It must address selection, concurrent promotions, seasonality, spillover, missing tracking and non-compliance. Do not choose a method because it produces a positive result.
The government's 2026 Green Book distinguishes appraisal, monitoring and evaluation. Its companion impact-evaluation quality guidance says observed change alone cannot establish that an intervention caused the change and points to the counterfactual. These are public-sector analytical sources with UK scope, not commercial benchmarks.
Choose the evaluation design before publication and budget its feasibility cost. Record the unit assigned or exposed, allocation method, intended comparison, sample constraints, interference risks and analyst decision rules.
If creators or audiences move between conditions, document the likely contamination. If the buyer cannot preserve a comparison, the correct output may be a descriptive contribution report rather than an incremental estimate. Agree that boundary while a different design is still possible, not after attractive dashboard results arrive.
Report uncertainty and decide
Publish the estimate with method, eligible population, period, exclusions and uncertainty interval where the design supports one. Run sensitivities for refund maturation, contribution assumptions, missing events, media cost, staff time and attribution leakage. Keep platform-reported and causal results in separate columns.
Set the decision rule before seeing the outcome: [buyer-defined threshold and action]. Name who can continue, pause or redesign. If the counterfactual fails, retire the causal figure rather than downgrading the caveat in small print. A transparent no-estimate conclusion is a valid result.
Before you act
- Define the decision, population, period, currency and data cut-off.
- Build the denominator from agency, software, creator and production costs.
- Agree the contribution measure and counterfactual before publication.
- Keep platform-reported and causal results in separate columns.
- Set the decision rule before seeing the outcome.
- Report uncertainty and run sensitivities on key assumptions.
Common questions
What must a buyer have before labelling a result as ROI?
The buyer needs a full economic cost, a defined contribution measure and credible evidence of what would have happened without the activity. Without these, report observed or attributed outcomes without a causal label.
Why is association between a tagged order and a post not enough?
A tagged order may be associated with a post, but association alone does not show what the buyer gained because the post existed. The four classes are delivery evidence, platform attribution, reconciled operational outcomes and estimated causal effect. Movement between them requires evidence.
What should happen if the counterfactual fails?
Retire the causal figure rather than downgrading the caveat in small print. A transparent no-estimate conclusion is a valid result. If the buyer cannot preserve a comparison, the correct output may be a descriptive contribution report rather than an incremental estimate.
