Measuring social media marketing without fooling yourself: which numbers tell you about sales

The short answer: track three things. How much content was published, how many people saw it and how sales developed over the same period. Everything else is tooling data that helps you make better videos but does not tell you whether the work pays off.

The problem with measuring social media marketing is that the platforms surface by default the numbers that look good. Likes, follower count, saves and engagement rate are all easy to pull up and completely detached from whether any money came in. Meanwhile the numbers that actually move show up in a different system, and often with a delay.

The confusion is expensive, because real decisions get made on the wrong numbers. Production gets stopped at the three-month mark because the follower curve is flat, even though brand searches may have doubled over the same period.

A view is not a result, it is raw material

We report views to clients, more than seven million a month across all partners combined. The individual numbers are big: Sneak.fi collected 9.8 million views in the first 30 days and Traveller more than 500,000 views from its first two posts.

These are real numbers, but they are not results. They are volume. A view tells you that distribution works and the algorithm is showing the content to the right people. It is a precondition for results, not a result in itself. If volume is zero, there is no point looking at anything else. If volume is big but sales do not move in three months, the fault is in the message or the offer, not in distribution.

This is how we split the metrics in practice:

Vanity: follower count, likes, number of comments, engagement rate out of context, views of a single video.

Raw material: views, accumulated watch time, profile visits, link clicks, share of new viewers.

Business: monthly sales development, number of new customers, direct traffic to the website, number of brand searches, quote requests, bookings, app downloads.

Engagement rate deserves a separate mention. It is a good tool for the creator, because it tells you which hook worked and which did not. It is not a reporting metric for the management team. With the Props app, engagement was ten times that of paid advertising, but that number only took on meaning once 80,000 app downloads were placed next to it.

The business metrics worth tracking

Three numbers are enough for most consumer brands.

The first is total marketing efficiency, meaning total revenue divided by total marketing spend. It is a crude number, but it does not lie, because it does not divide the credit between channels. If the number improves when short-form video production starts and other costs stay the same, the work is working.

The second is direct traffic and brand searches. When people see you on social media often enough, they do not click the link. They search for your brand by name on Google a week later or type the address straight in. Search Console shows the development of brand searches for free, and it is an honest indicator of whether awareness is growing. Separate brand searches from product searches, because they move for different reasons.

The third is the question "where did you hear about us" on the order form or at the checkout. It is imprecise, people misremember and some do not answer. Still, it is the only metric that asks the buyer directly, and as a trend it is usable. When the share of TikTok or Instagram in the answers grows month after month, that means something, even if any single answer is imprecise.

Why attribution fails in short-form video

Attribution was built for a world where a person sees an ad, clicks and buys. Short-form video does not work like that. The viewer sees the video a few times over a couple of weeks, never clicks once, and then buys when the need arises.

When the same purchase shows up in both the Meta and the TikTok report, and Google counts the brand search as its own achievement on top, the sum of the conversions reported by the platforms is bigger than actual sales. Cookie consents and browser tracking prevention cut further. So the platform's own report is a good tool for optimising ads and a poor tool for deciding which channel deserves the budget.

A practical example: with Hans Välimäki we made 41 short videos and they gathered 3.2 million views. Table bookings rose during the first two months. A restaurant has no click path to follow. Nobody watches a video and clicks themselves into a table. The only way to read the result was to look at the booking book before and after.

The same applies to most consumer brands. A single paid distribution campaign can produce a measurable result, like the 8x return on ad spend of Sneak.fi's Black Friday campaign, but the effect of the organic side only shows in the fact that the campaign worked that well in the first place. The audience had been warmed up in advance.

Three ways to measure when there is no click

The switch-off test. Switch off a channel or paid distribution for two to four weeks and see what happens to sales. It is a crude and expensive test, but it is the only one that directly answers the question "what happens if we stop". Do it in a quiet period, not in peak season.

Before and after with a fixed time window. Take three months before the start and three months after it. Compare sales, direct traffic and brand searches. Account for seasonality by also comparing with the same period the previous year.

A baseline per channel. Record the starting point before anything launches. Many companies start production without recording where they started, and six months later nobody can say whether anything improved. With OnePlus Finland organic visibility grew by 650%, and that number only exists because the baseline was known. The first video collected 300% more views than the ones made by the previous partner, which was directly comparable on the same account.

Over what period to read the numbers

A single video tells you nothing. The return on short-form video is distributed unevenly, and one post in ten can bring most of the visibility. If you look at one video, you are reading noise.

A four-week period tells you whether distribution works. A three-month period tells you whether sales move. A six-month period tells you whether the contract pays off. Decisions made on a shorter cycle than that are guesses, and that is one reason why a one-off purchase and ongoing work cost different things.

Posting frequency affects how fast data builds up. Eight videos a month produce a readable sample faster than two. This is one reason why pricing should be worked out per month rather than per video, and the topic is covered in more detail in the article how much social media marketing costs.

When not to start building a set of metrics

There are situations where time spent on measuring is time taken away from doing.

If you publish fewer than five videos a month, the sample is too small. Build a production rhythm first, then measure. A dashboard with three data points is not a dashboard.

If you make a handful of sales a month, the swings in sales are down to randomness, not marketing. The law of large numbers does not work on small numbers. In that situation qualitative feedback is more useful than a spreadsheet.

If you are not prepared to stop anything based on the numbers, do not measure. The only purpose of a set of metrics is to steer a decision. If the decision has already been made and the numbers are being collected to justify it afterwards, that is reporting theatre.

If you are only just starting, put your energy into the first posts. How to start making short videos is a different question from how to measure them, and it comes first.

What to demand from a partner

If you buy the work from outside, measurement belongs in the contract. Ask in advance which numbers are reported, how often and what happens if they do not move. An agency that only reports follower count and likes has chosen metrics that are easy to improve without anything changing.

Also ask who owns the accounts and the data. If the partnership ends, the historical data has to stay with you. These questions are covered in the article how to choose a marketing partner.

On the paid distribution side measurement is more straightforward than in organic, because the budget and the result sit in the same tool. The ROAS reported by the platform alone is still not enough for decision-making, and there is more on that on the page paid advertising for businesses.

Start from the baseline

If you want a set of metrics that holds up for six months, start by recording the current state today. Monthly revenue, direct traffic, number of brand searches, plus followers and views per channel. One spreadsheet is enough. Update it once a month on the same day so the numbers are comparable.

We are happy to go through this with you and at the same time look at where, in your case, results should be read from at all. Book a time on the social media marketing service page or tell us your current situation, and we will say directly whether short-form video is the right channel for you or not.