Segmentation: Choosing the Cut
By the end of this lesson, you should be able to: pick a segmentation from what could have changed rather than from what's available, explain why slicing every dimension always finds something, and judge whether a cut is worth running.
What segmentation is
Segmentation means splitting a metric by a dimension, like country, platform, plan or acquisition channel, to see where a change came from.
Any metric can be split many ways, and each split answers a different question. A cut by country asks "did this happen in one place?" A cut by app version asks "did a release cause this?" A cut by channel asks "did the people arriving change?"
The cut you choose decides which of those questions you're answering. So choosing it is the most important step in the analysis, not a detail.
A real example: Spotify's revenue per subscriber
Spotify's average revenue per Premium subscriber (ARPU) was €4.69 in 2024 and €4.63 in 2025, down 1%. On its own, a 1% drop doesn't say much. Spotify's annual report splits it into three causes.
| Cause | Effect on Premium ARPU |
|---|---|
| Exchange rates: revenue earned in other currencies was worth fewer euros | −€0.17 |
| Product and market mix: changes in which plans and countries subscribers are on | −€0.14 |
| Price increases | +€0.25 |
| Total change | −€0.06 |
Each line points to a different dimension you could cut by, and each leads to a different conclusion.
- Cut by currency, and the biggest part of the drop is an exchange-rate effect. Nothing about the product changed.
- Cut by plan, and you'd look at the Family and Duo plans. A Family plan counts up to six Premium subscribers on one subscription, so each person brings in less, even when the household pays more.
- Cut by country, and you'd look at markets where Spotify sets lower prices to match local spending power.
- Cut by price change, and you see the one lever Spotify pulled on purpose. It added €0.25.
None of these splits is wrong. And at the same time, Premium revenue grew 11%, because the number of subscribers grew much faster than ARPU fell. A team looking only at ARPU, cut the wrong way, could easily spend a quarter fixing a problem that doesn't exist.