Measuring influencer marketing ROI: which number survives the follow-up question
The report is on the table, the curves point upward, and someone in the room asks: "And what did that get us?" What decides the answer was settled long before the report: what the return of this campaign should be recognizable by.
In a survey by the US advertisers' association ANA from the summer of 2026, 67 percent of 78 marketing leaders rated measurement as the hardest step in influencer marketing, harder than selection, negotiation or approval.
What gets measured is what is easy to collect. In the Influencer Marketing Hub Benchmark Report 2026, a survey of more than 600 participants, the brands that are raising their budget sharply name brand awareness as a success metric at 89 percent, engagement at 51 percent, and content quality at 39 percent. Conversions are named by 35 percent, attributed sales by 25 percent. In the same survey, 87.5 percent of those who answered this question expect a higher influencer budget for the current year, and 72.2 percent even expect an increase of more than half. More money in a channel whose success is judged mostly by awareness and reaction makes the open question more expensive, not clearer.
The short answer
The return of an influencer campaign can be measured only if, before the start, it is settled what change it is supposed to produce. A figure becomes defensible through two measures: the one that belongs to the chosen job — net reach in the target group, or attributed actions — and a comparison that shows what would have happened without the campaign. Anyone who defines both only after the start can no longer collect them for this campaign.
Before you go further. Write down the change this campaign is supposed to produce. If that sentence is missing, the rest of this page is the reason why.
Why the report looks good and the question of return still stays open
Engagement is an honest metric for what it measures. It shows how many people reacted to a post. It does not show whether those people belong to the target group, whether they knew the brand beforehand, or whether their behavior changed afterwards. A campaign can achieve excellent figures because a creator has a loyal community that reacts to every one of their posts — including the one that shows a product they will never buy.
The reverse is also possible: a campaign with unremarkable figures that was commercially successful because it reached exactly the two thousand people who were on the verge of a purchase decision. In the report, the effective campaign looks worse than the ineffective one. As long as the judgment hangs on the reaction, the decision about the next budget is made on the wrong basis.
The decision before the brief
Before a metric makes sense, the campaign needs a job. The jobs exclude one another.
- Awareness in a group that does not yet know the brand. The decisive measure is net reach in exactly that group, not the sum of follower counts.
- Trust in a group that knows the brand and is hesitating. What counts here are signals beyond the post: brand-related searches, direct visits, questions to customer service.
- Action in a group that is ready to buy or to sign up. Here the campaign needs a trace that can be assigned — its own link, its own code, or its own page — and a comparison group without the campaign.
Anyone who assigns all three jobs at once gets a report that shows something about each of them and an answer about none.
Five metrics and what each one answers
Each of these metrics is usable as long as it is clear what it does not do. It becomes a problem when a metric is asked to answer a question it was not built for.
- Reach, impressions. Answers how many chances to be noticed were created. Does not answer whether anyone looked, or whether it was the same people over and over.
- Engagement rate. Answers how strongly a post moves its audience to react. Does not answer whether the reaction came from the target group, or whether it led to anything.
- Net reach in the target group. Answers how many of the right people were actually reached. Does not answer whether they acted.
- Attributed actions and sales. Answer what commercial contribution can be measured. Do not answer what would have happened without the campaign.
- Brand-related searches, direct visits. Answer whether attention continued past the single post. Do not answer how much of it belongs to one particular creator.
For thresholds above which an engagement rate counts as good, there is no sound public basis. The figures in circulation come from data sets whose composition is not disclosed. For one's own assessment, comparison with one's own earlier campaigns says more than an industry figure.
A question that can be answered today. Take the report from your last campaign and look in it for the measure that was fixed before the start. If you find none, that is already the result of this section — and the starting point for the next campaign.
Where budget disappears without showing up in the report
Budget disappears in two ways. The first is inauthentic reach. A study by CHEQ together with Prof. Roberto Cavazos of the University of Baltimore put the share of global influencer spend lost to fraud in 2019 at 15 percent — about 1.3 billion US dollars on about 8.5 billion US dollars of annual spend. The authors explicitly call this amount a cautious calculation. The separate checks they build it from diverge widely: among mid-reach influencers about 20 percent of followers were under suspicion, in a sample of 10,000 influencer accounts 25 percent counted as fraudulent in some form, and in individual brand campaigns the figures lay between 32 and 78 percent. The figure is several years old and says nothing about how high the share is today. It says the effect is large enough to check before booking, not after.
Inside the industry this is not a fringe case. The Benchmark Report of the Influencer Marketing Hub, already cited, counts the fraud and quality problems that occurred in the respondents' own programs. Inauthentic or automated followers account for 56.5 percent of all reported problems, inauthentic comments and purchased engagement together for 20.8 percent. Only 10.9 percent of respondents selected none of the risks they were asked about. What is counted is what occurred, not what someone is worried about. Whether a check follows before booking, the report does not say.
The second way the loss arises involves no fraud at all
It happens even when every creator booked has real reach. When several creators reach the same people, the report adds their reach figures together. Reality does not. The same person sees the campaign three times, the brand pays three times, and the number of people reached does not rise.
A higher contact frequency can be wanted. Then it is a decision. Left unchecked, it is a by-product: the brand pays for reach and receives frequency. How large this overlap typically is cannot seriously be put into a figure; the ranges in circulation come from vendor material with worked examples rather than from measurements. It can still be checked once the audience data of the individual creators is available.
Creator selection decides whether anything measurable comes of it at all
Both ways of losing budget have the same origin: a selection made by prominence and gut feeling. Anyone judging creators one by one checks follower count, the look of the profile, and a few posts. What stays out of sight is the make-up of the audience and how far it overlaps with the other candidates.
Checking both is not a debrief. It is a precondition. Anyone who finds after the campaign that three out of five creators serve the same community can explain the result, but can no longer change it. Anyone who finds that beforehand swaps two names and reaches more people on the same budget.
All 78 participants in the ANA survey rated creator selection as important, 86 percent as very important. 97 percent rated the prior check as important, and 45 percent at the same time as difficult. A step that practically everyone considers important is hard to carry out for almost half of them.
Manual review or systematic assessment
The difference between the two ways is not the effort. It is when the reason is given. In a manual review it comes after the decision. In a systematic assessment it comes before.
Manual review
- Basis: experience, a visual check, individual posts
- Traceability: the reason is produced afterwards
- Overlap across several creators: rarely checked
- Limit: works only up to a few dozen candidates
Systematic assessment
- Basis: uniform criteria across all candidates
- Traceability: the criteria are fixed before the decision
- Overlap across several creators: can be checked once audience data is available
- Limit: replaces neither the fit of the content nor the definition of the goal
Have an existing creator list reviewed. If you already have a selection in house, bring it to the first conversation. The review checks audience overlap between candidates, whether the audience matches the job fixed before the brief, and which of the seven X7 dimensions are missing from the shortlist.
Bring your shortlistHow we set up selection at Mashin
We work with X7, our system for assessing creators. According to Mashin's own description, X7 uses more than 100 specialized agents and judges candidates along seven dimensions instead of a follower count: Character, Omnichannel, Narrative, Tenets, Engagement, Niches and Topics. Each dimension is collected in the same way for every candidate, so that in the end names are compared, not impressions. On wearemashin.com/en/x7 those seven names are the published key aspects of a brand's image for influencer selection. The page does not say what each dimension checks. A demonstration is requested through the contact form on that page.
We are present in ten countries: Argentina, Bolivia, Chile, Colombia, Guatemala, Mexico, Paraguay, Spain, Uruguay and the United States. With that come more than 150 specialists. This is what decides whether we can judge a local market from our own observation or only from data.
No system takes over the definition of the goal. That stays the brand's decision, and it is made before the brief. X7 makes the selection comparable. It does not make it correct.
Six steps to reporting that survives follow-up questions
Step 1 Set the job
Awareness, trust, or action. One of them, in writing, before the brief.The temptation is to name all three, because each of them sounds right on its own. A campaign that is meant to build awareness, convince, and sell at the same time cannot afterwards be judged against any of the three jobs. The sentence belongs in the brief, not in the report.
Step 2 Name the measure that proves it
One lead metric for each job. Everything else is context.A lead metric does not mean it is the only one collected. It means it is settled beforehand which figure answers the question when the result is ambiguous. Without that decision, the metric chosen after the campaign is the one that looks best.
Step 3 Prepare the comparison
A time span, a region, or a group without the campaign.Without a comparison, every change remains a claim, because nobody knows what would have happened anyway. The comparison does not have to be elaborate: a region left out, or the same period the year before, is enough, provided it was fixed beforehand.
Step 4 Check the selection before it is fixed
Make-up of the audience, plausibility of the reach, overlap between the candidates.Once contracts are signed, this check is worthless, because its result no longer changes anything. Before that, it decides whether anything measurable comes of the campaign at all. It therefore belongs in the schedule, not in the debrief.
Step 5 Lay down the trace
Own links, own codes, own pages. Nothing can be assigned after the fact.An assignment that is searched for only after the campaign is no longer there to be found. Each creator needs something of their own, reachable only through them. It is the least expensive step on the whole list, and the one most often forgotten.
Step 6 After the campaign, ask the same question as before
Did the measure named before the start change?This question is the whole difference between a report and a result. It can be answered only if the five steps before it were taken. If the answer is negative, that too is a usable result for the next campaign.
- Job fixed in writing (one of: awareness, trust, action)?
- Lead metric named for that job?
- Comparison prepared (time, region, or group without the campaign)?
- Selection checked before contracts?
- Trace per creator in place?
- Same question ready for after the campaign?
Common questions
What is a good engagement rate in influencer marketing?
There is no sound general answer to that. What tells you something is the comparison with your own earlier campaigns and with the other candidates in the same selection. Next check: compare those campaigns before you treat any industry figure as a bar.
How do I recognize inauthentic reach before booking?
Through oddities in the relationship of reach, reactions and comments, through the make-up of the audience, and through sudden jumps in growth with no visible cause. A study by CHEQ put the loss from fraud for 2019 at 15 percent of influencer spend and calls the sum calculated from it a cautious calculation. Next check: look at the audience before you book.
Is influencer marketing still worth it?
There is no blanket answer. In an industry survey of more than 600 participants, 87.5 percent of those who answered expect spend to rise, and 72.2 percent expect an increase of more than half. That is a statement of intention, not proof that it works. Next check: fix the job of the campaign before you treat rising spend as a reason to buy.
Why does measurement fail in practice?
Because the question is asked too late. In the ANA survey, 67 percent of 78 participants named measurement as the hardest step, because the yardstick is missing. Next check: name the measure before the brief, not after the report.
How do I compare influencer marketing with paid social?
Only on a measure both channels can supply, as a rule attributed actions or net reach in the target group. Next check: pick that shared measure before you open the comparison.
Sources: ANA, Influencer Marketing: Reducing Waste and Optimizing Investment; CHEQ with Prof. Roberto Cavazos, The Economic Cost of Bad Actors on the Internet; Influencer Marketing Hub, Benchmark Report 2026.
If the selection is still open, both decisions can still be made: who gets booked, and by what their contribution should be recognizable.
Contact us