How to measure influencer marketing: KPIs, attribution and limitations

1. In short: choose the metric by the goal, and expect each number to show only part of the effect.

Measuring influencer marketing starts with the goal. Reach and views show how many people were reached and how often the content was seen. Engagement shows how the audience reacted, and clicks and purchases show what it did next. Changes in awareness, attitude and preference are a goal of their own with their own metrics [1, 2].

Nor do the numbers mean the same thing on every platform. Each platform decides for itself what a view is, how impressions are counted and what counts as engagement [3]. Meta counts repeat views, YouTube has counted a view as soon as the video starts to play since 24 August 2026, and TikTok counts one view per impression without replays in its ad reports [4, 5, 6].

Sales are easiest to follow with discount codes and links, but both have known gaps. Codes spread beyond the creator’s followers, and someone who would have bought at full price anyway can get a discount [7]. Codes and last-click also give the credit to what closes the sale, not to what influenced it [3]. Statistics the creator sends as a screenshot cannot be checked by the company itself, and the platforms’ own numbers are rarely audited by anyone outside [3, 7].

The rest of the article goes deeper. Section 2 pairs goals with metrics, and section 3 shows what the platforms count as a view and as reach. Section 4 covers discount codes, links and attribution models. Section 5 is about whose numbers you can trust, and section 6 about what verified views are. Section 7 gathers what should be settled before the campaign, and section 8 describes the sources.

2. Each goal has its own metrics, and a metric for the wrong goal says little.

The industry body IAB Sverige splits the effect of influencer marketing into five areas. Reach is how many people have been reached, and views how many times the content has been consumed. Engagement is interactions such as likes, comments and shares, and conversions are for example traffic or purchases. Brand effect is changes in awareness, attitude and preference [1]. The US advertisers’ association ANA published guidelines in 2022 with three groups of metrics: awareness, engagement and conversion. The guidelines are endorsed by the industry body Media Rating Council [2]. Figure 1 pairs the goals with the metrics.

Mediemyndigheten, the Swedish Agency for the Media, describes how campaigns are followed up in practice: once the campaign is over, it is measured on key figures such as reach and views, but also on engagement rate [8]. IAB Sverige recommends following the results continuously while the campaign runs and comparing them with relevant benchmarks [1]. An awareness campaign judged by its sales, or a sales campaign judged by its views, is therefore measured against the wrong goal.

Brand effect is the hardest to measure in a short time. IAB Sverige writes that longer partnerships give better opportunities to measure real brand effect, by comparing results before and after the campaign periods [1]. A degree thesis from Örebro University on hotels notes that it can be hard to measure how much “buzz” a partnership creates around the brand [11].

3. A view means different things on different platforms, and each platform decides for itself what counts.

There is no shared definition of a view. IAB in the US writes in a review from January 2026 that each major platform decides for itself what a view is, how impressions are counted, what counts as engagement and which data can be retrieved through the platform’s developer interface [3]. The definitions also change over time. Figure 2 sums up the definitions that could be read on the platforms’ own pages.

TikTok

TikTok’s help centre for advertisers defines several metrics. A video view is counted each time the video starts to play, once per impression, and replays are not counted. There are also separate metrics for views of at least 2 and at least 6 seconds [6]. Reach is the number of unique accounts that have seen the ad at least once, and TikTok writes that reach is estimated and should not be taken as exact [12]. These are metrics for ads. The help page with TikTok’s definitions for ordinary posts could not be read when the article was written, so we do not state how TikTok counts a view on an ordinary post.

The same platform can have several definitions. For its Creator Rewards programme, TikTok counts “qualified views”: the watch time must be at least five seconds, several views from the same account count as one, and views of ads, paid promotions and sponsored content do not count. Sweden is not among the countries where TikTok says the programme is available [13]. The example shows that the word “views” in a report always needs a source and a definition.

Instagram and Facebook

In November 2024 Meta introduced the “Views” metric, which states how many times a reel or video has been played. Repeat views are included [4]. In Instagram’s developer documentation, reach is the number of unique accounts that have seen a reel at least once, unlike views, which can include several views by the same account. Here too, reach is estimated [9].

YouTube

YouTube first changed the count for Shorts. From 31 March 2025, a view is counted when a Short starts to play or replay [15]. Since 24 August 2026, views in every format are counted as soon as the video starts to play, including long videos and live streams. Earnings in YouTube’s Partner Program are still based on “engaged views”. YouTube also writes that it may slow down, freeze or change a count and discard low-quality playbacks [5].

The ad industry’s standard

The ad industry has a standard for when an ad counts as viewable. Under the Media Rating Council’s 2015 guidelines for desktop ads, at least 50 per cent of a video ad’s pixels must be visible for at least two continuous seconds [14]. The standard applies to ads and is not the same as the platforms’ view metrics in Figure 2.

4. Discount codes and links make sales measurable, but they give both too much and too little credit.

Discount codes

With their own discount code or a tracked link per creator, the company can see which purchases can be tied to the partnership [11]. In a bachelor’s thesis from the Stockholm School of Economics in 2020, five retail companies that work with influencer codes were interviewed, along with ten customers who had used such a code. All the companies said that discount codes were in principle the only form of partnership in which they could measure influencer marketing [7]. The thesis is small and qualitative, but it describes the mechanisms clearly.

Codes have two known gaps, pointing in opposite directions. The first is that the code spreads. According to the thesis, codes spread both through the creator’s reach and through customers sharing them with others, and if a code spreads widely there is a risk that customers who already meant to buy at full price make a discounted purchase [7]. Purchases the partnership did not cause are then credited to it, and the company pays the discount. The second gap is the reverse: someone who sees the video but buys without the code does not show up in the code’s numbers. We have found no research on how large the leakage usually is.

Links and UTM parameters

A link with UTM parameters tells web analytics where a visitor came from, with fields for source, medium and campaign [10]. A separate link per creator keeps the traffic from different creators apart. Even so, a link only measures the people who click.

Attribution models

When a purchase has several touchpoints, the attribution model decides which channel gets the credit. With last click in Google Analytics, the last channel the customer clicked through from gets the whole value, and direct traffic is ignored. With data-driven attribution, the credit is spread based on data for each conversion [16]. Example: someone who sees a creator’s video, later searches for the brand and then buys is credited to the search under last click. IAB in the US writes that creator performance today is mostly evaluated with last-click links, promo codes and affiliate tracking, and that those systems reward what closes the sale, not what influenced it [3].

View-through goes the other way. In TikTok’s ad tools it counts purchases made after someone has seen an ad without clicking, within a set time window, and TikTok recommends a window of one day [17]. The model applies to ads. It captures purchases that last click misses, but a purchase after a view is not necessarily a purchase because of the view. IAB in the US argues that the industry needs to move beyond last click, towards multi-touch measurement and measuring what the campaign actually added, known as incrementality [3].

5. Screenshots cannot be checked, and the platforms’ own numbers are rarely audited by anyone outside.

The numbers in a campaign report often come from several places. IAB in the US describes measurement as a hybrid system of the platforms’ own dashboards, creators’ self-reporting, third-party tools, affiliate tracking and brand lift studies. Platform data is supplemented with screenshots and numbers the creators report themselves, and the whole is often pieced together by hand in spreadsheets [3].

The same shows in the Swedish thesis from the Stockholm School of Economics. The companies could not see for themselves how far a post had spread, and depended on the creator sending the statistics from their account [7]. A screenshot shows a number at one moment, and the company has no way of its own to check it against the platform. The platforms can also change numbers afterwards [5].

The platforms’ own numbers have a different limitation. According to IAB in the US, delivery metrics are largely self-reported by the platforms, independent third-party auditing is still limited, and fake followers, bots and engagement manipulation persist [3]. The platforms work against this themselves. TikTok writes that fake likes, followers and other inflated signals are removed when detected, and YouTube that low-quality playbacks can be discarded [5, 18]. In the Netherlands, the consumer authority ACM in 2021 forced an online store that had bought 98,000 followers and 27,000 likes to stop, because it had misled consumers [19].

6. Verified views are the platform’s own number, read directly instead of reported by the creator.

By verified views we mean, in this article, views read directly from the platform through its developer interface, an API, instead of being reported by the creator, for example with a screenshot. On TikTok such access is given through permissions the creator approves, for example read access to the creator’s public videos [20]. Each video then has a field with its number of views [21].

That solves part of the problems in section 5. The number comes from the platform and not from an image, it can be read off again, and the company does not have to rely on the creator’s own report. But it is still the platform’s own number. TikTok’s documentation describes the field only as the number of views of the video, without saying how the views are counted or how often the number is updated [21]. The definition is the platform’s, and the number is not audited by any independent party [3]. Verified views are therefore a safer source than a screenshot, but not an independent measurement.

7. Settle the goal, metric, source and read-off day before the campaign starts.

Part of the uncertainty can be reduced by settling the measurement in advance. Start with one goal per campaign and one main metric per goal, following the split in Figure 1 [1]. Then write down which platform’s number counts, which of the platform’s metrics it is, and whether the number is read directly from the platform or reported by the creator. Also decide on which day the number is read off, since the platforms can change numbers afterwards [5].

If sales are to be measured, give each creator their own code and their own link with UTM parameters [10], and decide in advance how purchases without a code and shared codes are to be weighed. Expect codes and last click to underestimate the creator’s contribution in some cases and overestimate it in others [3, 7]. Follow up continuously while the campaign runs and compare with relevant benchmarks. If the goal is the brand, plan a measurement before and after the campaign [1].

8. About this article: firm definitions for ads, a thinner basis for ordinary posts and for code leakage.

The article was compiled in October 2026 from published sources, read on 15 October 2026. The definitions come from the platforms’ own help pages and documentation. Every source is tagged by what it is. “TikTok” marks TikTok’s own pages. “Estimate” here marks sources that are not independent: the platforms’ own documentation of their products, such as Meta’s and Google’s help pages, and the industry bodies IAB Sverige, IAB in the US and ANA. IAB in the US counts platforms among its members, and ANA developed its guidelines together with agencies and the major platforms. “Independent” marks authorities, research, measurement standards and the press.

What is not covered

The article gives no target values for engagement rate, cost per view or return, because the figures we found come from vendors with no method given. It does not cover how Swedish law views bought followers, likes or views, because that needs a legal assessment. IAB Sverige’s guides that sit behind a form have not been read.

Limitations

TikTok’s definitions for the statistics on ordinary posts could not be read in TikTok’s help centre, so the article does not state them. We have found no peer-reviewed research on how large a share of discount-code purchases comes from shared codes. The theses from the Stockholm School of Economics and Örebro University are small, qualitative student works and illustrate mechanisms, not the market as a whole. IAB’s review and ANA’s guidelines concern the US market. The column on what the metrics do not show in Figure 1 is our judgement, and we have not measured how many people search for the topic.