Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Influencer Marketing › Influencer Measurement

Influencer ROI and Incrementality

How to judge influencer return: why earned media value misleads, cost per outcome, attribution limits, incrementality tests, and long-term effects.
Guideadvanced9 min readUpdated

What return means for creator work

Return on investment (ROI) compares the value created with the money spent. For creator work, that simple ratio hides three choices that decide the answer.

The first is which value. Sales revenue is the obvious one, but gross margin is the honest one, because a campaign that sells a lot at a discount can lose money. For subscription businesses or repeat purchases, the value of a new customer over time matters more than the first order, which connects creator return to customer economics covered in the LTV and CAC guide. The creator-acquired customer's lifetime value can be higher or lower than other channels', and is worth checking.

The second is which outcomes count. Link-tracked orders are the floor, captured with the methods in the KPIs and tracking links guide. Code redemptions add some. Branded search and direct traffic lifts suggest more. A test can estimate the total. Each step brings the estimate closer to the true effect and further from what a single report can prove. The influencer measurement hub sets out how these layers stack.

The third is the time horizon. Creator content can work immediately, but much of its effect on awareness and preference arrives over months, through the brand being remembered when the need arises. The IPA's work on advertising effectiveness has long argued that short-term measures understate the value of brand-building; creator programmes aimed at awareness should be judged with that in mind.

Earned media value and why to drop it

Earned media value (EMV) takes the impressions or engagements a creator post earned and multiplies them by a rate meant to represent what that attention would have cost as paid media. It became popular because it produces a large, single, currency-denominated number that makes creator work look profitable.

It should be retired, for several reasons:

  • It is not revenue. It measures the notional cost of attention, not what the attention produced. A campaign can have a high EMV and sell nothing.
  • The rates are arbitrary. Vendors use different values for a like, a comment or a view, and the same campaign produces very different EMVs depending on whose rates are used.
  • It rewards the wrong things. Posts with high engagement and no relevance score well; a creator with inflated engagement looks excellent.
  • It double-counts. Paid amplification impressions are sometimes included, valuing attention the brand paid for again.
  • Finance does not accept it. A finance team asked to fund a programme on its EMV will reasonably ask how much of it can be banked.

If a vendor reports EMV, ignore the total and use the underlying reach and engagement data, which are real measures. The headline number should be a cost per outcome, then an estimate of incremental return.

Cost per outcome

Cost per outcome divides the full cost of the creator work by the outcomes it produced. The most common version is cost per acquisition (CPA), the cost of each order or customer. Where the outcome is a new customer, it is a channel-level customer acquisition cost (CAC). The CPA calculator handles the arithmetic; the work is in deciding what goes on each side.

ElementIncludeOften missed
Creator feesAll fees for content and postingAgent commission, rush fees
ProductProduct sent to creators, at costShipping, replacements, extra samples
CommissionPayments on tracked salesCommission on orders later returned
Usage rightsLicence fees for paid and organic reuseLicence extensions
AmplificationPaid media behind creator contentSpend on tests that were stopped
Tools and agencyDiscovery tools, agency or platform feesThe share of annual tools used for this campaign
Team timeHours spent on selection, briefing, review, reportingUsually ignored entirely
OutcomesTracked orders, sign-ups or customersOrders through codes, sales outside the site

Calculate cost per outcome per creator and for the campaign. The per-creator figure decides who to re-book; the campaign figure decides whether to continue. Where revenue is tracked, return on ad spend can be calculated with the ROAS calculator, but remember that it uses the same tracked revenue and inherits its gaps. The same figures, read before a deal rather than after it, set the ceiling in the guide to negotiating creator fees and terms.

Compare cost per outcome with other channels on the same basis. Creator work judged on link-tracked orders alone against paid social judged on platform-reported conversions is not a fair comparison, because platforms count conversions links never see. Where an influencer agency or creators' agents are involved, the guide to working with creator agents and influencer agencies explains why the tracking data should stay in the brand's own analytics so this comparison remains possible.

Link-based Attribution credits an order to a creator when the buyer arrived through that creator's tracked link, usually within a time window and under the analytics tool's attribution model. The attribution models framework explains how models divide credit between touchpoints, and the choosing an attribution approach guide covers how a brand should pick one.

What links can show: that some identifiable people came from a creator's content and bought, how quickly, and with what basket. That is useful for comparing creators and content, because the gaps are roughly similar across creators of the same type.

What links cannot show:

  • People who did not click. Most viewers who are influenced do not tap the link; they search, type the address or buy elsewhere later.
  • Influence before the last click. A creator who introduced the brand gets no credit when the buyer later converts through a search ad.
  • Causation. A tracked order from an existing customer who was going to buy anyway is not an effect of the creator at all.

So links understate the reach of creator influence and overstate the causal effect among those who click. The net error can run either way. That is why tracked results are best described as an indicator and a floor for comparison, not as the return.

Incrementality tests for creator campaigns

Incrementality is the share of outcomes that would not have happened without the creator work. Measuring it means comparing a group exposed to creators with a similar group that was not. The incrementality testing guide covers test design in depth; creator campaigns lend themselves to a few practical forms.

Geographic splits. Run creators whose audiences are concentrated in some regions and not others, then compare sales trends in exposed and unexposed regions against their history. This works best for creators with local audiences and products sold across regions.

Time-based comparisons. Run the programme in alternating periods and compare. This is weaker, because seasonality and other activity change between periods, but it is sometimes the only option.

Holdouts in paid amplification. When creator content is amplified, platform lift-study tools or a held-out audience can measure the paid layer's incremental effect. It does not measure the organic posts, but it answers whether amplification earns its budget. The guide to influencers for a product launch keeps that paid layer separate in the budget, holding it back until the posts worth amplifying are known.

Matched creators. Compare sales among audience segments reached by creators with similar segments not reached, where the brand's own data allows it. This is approximate and needs care to avoid comparing different kinds of people.

Illustrative example — not a real result

A regional test for a beverage brand

Saltmarsh Drinks, an invented soft-drink brand, runs a creator campaign for 8 weeks in 4 northern regions, using creators with mostly local audiences, and holds out 4 comparable regions. Link-tracked online orders in the test regions come to 1,100. Retail sales data shows sales in test regions growing 6% faster than in holdout regions against the previous year, which the team estimates at about 9,000 extra units, many times the tracked figure. Total cost is £38,000 including fees, product and team time. Cost per incremental unit is about £4.20, against roughly £35 if only the tracked orders were counted.

Tests are noisy and need enough volume to detect an effect. They suit big questions, such as whether a creator programme lifts sales at all, better than small ones, such as which of two creators was better. Run one when the budget at stake justifies it, and use the result to calibrate how far above the tracked floor the true return probably sits for future campaigns.

Long-term and brand effects

Some of the value of creator work is not captured by any test run over a few weeks. Creators can change how a brand is perceived, make it familiar to people who will not need it for months, and produce content that keeps working as paid creative long after the campaign. The IPA's effectiveness work argues that brand effects build over longer periods than most campaign measurement windows, and creator programmes aimed at awareness are a case in point.

Ways to see those effects include brand tracking surveys that measure awareness and consideration among the target audience before and after a programme; branded search volume over months rather than weeks; the paid performance of licensed creator content over its whole licence; and repeat-purchase rates of customers acquired through creators. None gives a precise return. Together they show whether a programme is building something or only renting attention.

Be honest in reports about which effects are measured, which are estimated and which are assumed. A programme defended with a clear floor, a tested estimate and a stated view on long-term effects is far easier to fund than one defended with a single large number.

Because creator and partner programmes are often judged against each other, the influencer marketing versus affiliate marketing comparison explains how their costs, tracking and incrementality differ. To see how return sits alongside disclosure, contracts and usage rights in a running programme, the influencer programme audit scorecard scores each and ranks the fixes.

Common mistakes

  • Reporting EMV as return. It is a notional cost of attention, not value created, and it damages the programme's credibility.
  • Leaving costs out. Product, licences, amplification and team time belong in cost per outcome.
  • Treating tracked sales as the full return. Links miss most viewers and claim some sales that would have happened anyway.
  • Comparing channels on different bases. Creator link data against platform-reported paid conversions is not a fair test.
  • Testing too small. A test without enough volume will find nothing and be read as proof of no effect.
  • Using margin-blind revenue. Heavily discounted creator codes can produce revenue and lose money; judge return on margin.
  • Ignoring time. Judging an awareness programme on a few weeks of sales misses most of what it was for.

Frequently asked questions

Why is earned media value a poor measure?

It converts views and engagement into a notional amount of money by multiplying them by assumed media rates, so it describes what the attention might have cost to buy rather than what it produced. The rates are arbitrary and vary between vendors, the result is not revenue, and finance teams rarely accept it. Cost per outcome and incrementality answer the question it pretends to answer.

Is a holdout test realistic for a small programme?

Sometimes. A small programme can rarely split audiences finely, but it can often run creators in some regions or time periods and not others, and compare results. The test will be noisy, so it suits larger questions, such as whether creators lift sales at all, rather than fine comparisons between creators. Where a test is not feasible, be explicit that tracked results are a floor.

How should creator results be compared with paid social?

On the same basis: cost per outcome with all costs included, and ideally the same attribution view or test design. Creator costs should include fees, product, licences and management time; paid social costs should include creative production. Comparing creator link-tracked results with platform-reported paid results usually favours paid unfairly, because platforms count view-through conversions that creator links never see.

Terms used on this page

Tools to try

Frameworks

Where this page is used

Part of these learning paths

Also related

Sources

  1. Institute of Practitioners in Advertising — IPA

Related