Stop Guessing: How to Measure Influencer Marketing

Zooming out Nowadays, we see and hear so many messages and so much information that we struggle to differentiate signal from noise and find shortcuts to valuable information. Trusted opinion leaders and creators help us narrow the myriad choices of brands and products, leveraging their audience's trust as a foundation. Brands and companies know this too, so they constantly seek ways to collaborate with creators to help promote their products. Many companies run collaborations, and many are successful – but how can you tell? Before taking the plunge, you probably have questions: How do I ensure my investment is well placed? How do I know it's working? \ Zooming in Imagine you begin exploring influencer marketing as a channel. You launch 1–3–5 collaborations, set up tracking, and see unclear results. There's a spike in organic traffic – but is it significant? And if you count only users who followed the tracking link, it looks like it wasn't worth it. How do competitors make it work? It still feels risky to scale. You may be feeling stuck. Everyone says it's a promising channel, but there isn't enough data or certainty to scale, and it's unclear how to move forward. \ Why read further and how Throughout my career, I have relied significantly on influencer marketing. I'm incredibly proud that we managed to measure every part of it, and I'm happy to share the approaches we used. This article discusses how to build and scale an influencer marketing channel so that you can measure it as precisely as possible and gather enough data to model further scaling. It is going to be a structured, long read. It is divided into independent sections so you can save time and focus on what interests you most. I've intentionally omitted some things to keep the article focused. Here are the sections: What we get from collaborations – from assets to effects How to structure sourcing and why How to launch the flights How to evaluate the results Let's begin! \ What we get from collaborations, from assets to effects We can look at collaboration results from two perspectives. From an effects perspective: Increase brand awareness (measured by brand awareness uplift) Build trust with users (an increase in one or several conversion metrics along the funnel) Bring sales (ROAS-positive sales or subscriptions directly from the collaboration) From a creative perspective: New creative ideas that resonate with the audience Reach and communicate with the influencer's audience Get video assets that can be repurposed on performance channels such as Facebook, Instagram, TikTok, etc. In some cases, my team and I aimed to get subscriptions from influencer content directly and make the channel at least ROAS-positive (or even ROI-positive). The launch and test plan were designed to achieve exactly that. \ How to structure sourcing and why The Niche Context is one of the most important factors when choosing influencers. When a user is in a specific context, they are more likely to receive, understand, and act on your message – there's a higher chance it will resonate. Take Scentbird as an example. As a fragrance subscription service, the most relevant context is influencers already talking about beauty. Then come adjacent niches like travel. The most far-flung would be comedy or sports. Imagine someone unwinding in the evening with funny content, and suddenly – a fragrance subscription ad. What?? It may seem obvious in theory, but when you start sourcing, you may be tempted to shift between topics or to prioritize audience size or geography over context. To build your sourcing and outreach strategy, create a grid of topics from most to least relevant, and work through them step by step in separate flights. Geography If applicable (depending on the product and available markets), it's helpful to select creators whose audiences are concentrated in specific geographies, especially for your first tests. When I say geography, I mean language – because with influencers, geography is controlled by the language they speak. Some languages are more widely spoken than others. For cleaner tests and measurement, it's worth choosing geographies with less widely distributed languages – Turkey, Germany, Korea, Poland – rather than France or Spain. Even though French, for example, is spoken across France, Côte d'Ivoire, and many other countries, product metrics (conversion rates, LTV, etc.) will vary significantly across those locations, making modeling more difficult and measurement more complex. If your product is a mobile app, there's an additional factor: device distribution between iOS and Android varies significantly by geography, which adds noise to your data. This is another strong argument for the approach above. The Audience Size Audience size is a critical factor, and it's actually possible to define it with precision. There are two conditions – one that sets an upper limit and one that sets a lower limit. The top benchmark. The biggest creators are usually too expensive for small and medium-sized companies. Working with micro-influencers with very small audiences, on the other hand, makes results hard to measure. Start by sourcing and reaching out to large and mid-tier influencers to understand the pricing landscape. The bottom benchmark. First, choose a key metric. It can be an actual conversion metric – like a subscription – if it happens relatively quickly. Or it can be a proxy metric that is high in the funnel, occurs frequently, and correlates strongly with the key conversion event. For this metric, observe the organic rate and its volatility over several weeks. In a specific geography, let's say the organic rate is (a ± b) on average, with 90% confidence, where a is the average daily organic rate and b is (1.64 × standard deviation) . For the result to read as a clear success, you want to see a spike of at least *(a+b)×2* within one day – something unambiguously above baseline, without needing to hire a specialist just to separate signal from noise. To get (a+b)×2 events per day, you need a certain number of impressions. Work backward: Conversion rate from click to install or website visit = m% Conversion rate from impression to click = n% Required impressions = ((a+b)×2) / m% / n% For example, your organic baseline is 40 installs/day with ±15 volatility. You need (40+15)×2 = 110 installs in one day to call it a clear signal. Your click-to-install rate is 10% and impression-to-click rate is 2%. So you need 110 / 10% / 2% = 55,000 impressions – meaning any influencer averaging fewer than ~55K views per video won't give you a readable result. From the influencer side, you typically know the average impressions on their last *x* videos (depending on the platform). Match your required impressions to the influencer's estimated impressions per post – assuming they make at least an average video. This gives you your minimum audience size threshold. Tracking Before you observe anything, you need a properly configured analytics and attribution system. A subscription tracking service is also helpful. The ideal setup is one where all data from attribution and subscription tracking flows into a single analytics platform. I know that sounds like a lot. But here's the reality: the fewer measurement systems you have, the slower you move, the less noise you can tolerate, and the more controlled your experiment setup needs to be. You cannot improve what you cannot measure. \ Observing Normal Once tracking is in place, the ideal next step is to observe the organic rate and product metrics for your target geography for at least two weeks under stable conditions – meaning no scaling of existing channels, and no new media buying started. I know this is one of the hardest things to ask for. In most companies, everything is happening simultaneously. But the cleaner your baseline, the easier it is to distinguish a real signal from background noise – which is exactly what you'll need to do when you evaluate results. \ How to launch the flights By this point, you've sourced, outreached, and collected pricing for several influencers in your chosen geography and niche. You've filtered for creators whose estimated impressions meet the threshold calculated above. Now compare them by estimated CPM. Proposed CPMs can vary by up to 10×, so add all options to a table and highlight the creators with the lowest CPM and the most affordable total price per flight. The result is your shortlist. If you want to measure results properly – and this is one of your first tests – the optimal approach is to launch different creators at different times within the same geography. If two influencers post simultaneously and you get a positive result, you won't be able to tell who drove it. The practical rule: run one collaboration at a time per geography, but run multiple geographies simultaneously. The mechanics of coordinating this across markets is a topic for another article. \ How to evaluate the results After each post, expect an organic spike on the day of the flight and for up to 7 days afterward. To evaluate each post, sum two things: Actions from users who followed the tracking link Actions from organic users above the baseline (total organic actions minus the usual organic rate calculated earlier) Also, track what share of users followed the tracking link – this is a key metric for future benchmarking. For the users who did follow the link, measure your primary product metrics. Then assume users who came without the link have the same metrics, and extrapolate accordingly. Run these calculations per influencer, and also calculate a blended result across all influencers in the flight. That gives you three reports: per creator, per geography, and per flight overall. About 7 days after the last post, evaluate the final results. The questions to answer: Which message resonated most with the audience? Did any creator in any geography significantly outperform the others – and why? Does any subset of influencers prove the channel works for you? If collaborations generally work, can this be a standalone channel, or does it only work as part of a broader marketing mix? Have these collaborations boosted other channels – performance, for example – by building brand awareness and trust? After several flights in the same geography, compare all funnel conversions and look for what changed. You may find that certain funnel steps improved significantly while others stayed flat. Brand awareness uplift can also be measured at this stage. \ Wrapping up Influencer marketing is often treated as a "feeling" channel – you run a few collaborations, see a vague signal, and either give up or scale blindly. Neither approach is necessary. The framework above is how you move from guessing to knowing: define the right geography and niche, calculate the minimum audience size from your own organic data, run flights cleanly one at a time, and measure both tracked and untracked users together. The specifics will vary by industry, product, and funnel – but the logic doesn't. Once you have a baseline, a measurement setup, and a few flights under your belt, you'll have enough data to model what scaling looks like and make the case internally. If you're at the stage of launching your first tests, start with tracking and baseline observation. Everything else follows from there.