Ask the wrong question and you will get a confident, well-formatted, entirely useless answer. This is not a failure of the data. It is a failure of the question, and the two are indistinguishable on a slide. The chart looks equally authoritative either way; only the person who chose the question knows whether it is worth anything, and that person is rarely the one presenting it.
Here is the reversal, straight from BeanBreaker Intelligence’s How The World Talks About Coffee, August 2026. Sort the study by count, and Coffee Shop wins outright: 598 accounts, the single largest category in the entire survey, the obvious headline, the one you would put on the poster. Now sort the same study by median audience, and Coffee Shop collapses to near the bottom of the pile – a median of 1,865 followers apiece. Same accounts. Same dataset. Same afternoon’s work. Two rankings that could not disagree more violently if they tried. One was quietly answering the question “how many of these exist,” the other “how far does each one actually reach,” and a leaderboard that blurs the two together is selling you a map of a country it has never once visited.
Coffee, of all products, should have taught us this distinction already, because we drink it every day. Volume and strength are not the same axis and never have been. A bucket of drip and a single ristretto are both, unarguably, coffee – but one is large and weak and the other is tiny and concentrated, and only a fool orders by the litre and expects intensity. Coffee Shops are the bucket: vast, everywhere, ubiquitous, and individually thin. The concentrated shots are sitting in categories you would never think to weigh, precisely because there are so few of them that the count-based ranking buries them at the bottom of the page.
Look at what those shots are pulling. Sixteen Education accounts carry a median of 10,615 followers, out-reaching the typical coffee shop by five and a half times. Seventeen Bakeries run a median of 8,905, nearly five times the typical shop. That is a rounding error’s worth of accounts – sixteen, seventeen – each one a ristretto, quietly out-pulling a category more than thirty times its size. And the pattern does not stop there. Twenty Podcast and Media accounts post a median of 10,260. Twenty-four Public Figures sit at 9,167. Creators and Bloggers land at 8,193. None of these will ever be the biggest category by headcount. Every single one of them dwarfs the biggest category by reach.
Which tells you something the count-sorted chart is structurally incapable of admitting: size and power are simply different assets that happen to share a leaderboard. The thing that has the most of something is almost never the thing that reaches the furthest with it, and the entity quietly controlling the second axis while everyone stares at the first has always gone home with the better afternoon. Headcount is a census. Reach is an army. You do not win the war by having the most villages, and you do not win attention by having the most accounts. The census tells you who is present. It says nothing whatsoever about who is heard, and it is being heard that the whole leaderboard was supposed to measure.
This is the entire game of attention, and it is played in every arena that has ever mattered. Count the troops or measure the artillery. Tally the seats or weigh the votes that actually move. Add up the shops on the high street or find the sixteen accounts that genuinely reach a room. The question you choose is not a tidy preparatory step you complete before the real analysis begins. The question you choose is the analysis. Everything that comes after it is just arithmetic wearing the costume of insight.
So the next time somebody hands you the largest category on the chart and calls it the winner, ask them – gently, and in front of everyone – which question they were answering. Then watch how fast the biggest number in the room discovers it was never once the strongest.
