Category revenue told me the wrong thing every time: count independent winners instead

Made a mistake early in this research worth writing down, because it is the obvious way to read a market and it is backwards.

I ranked categories by total estimated revenue and got a list of places to avoid, presented as a list of opportunities.

Chatbots: about 273 million a month across the category. One product is roughly 267 million of it. Dating: about 126 million, two apps take 81. Language learning: about 61 million, one takes 48. Every one of those looks enormous and is, for somebody else.

Recounting by how many independent products clear a meaningful threshold inverts the list. AI music generation came out at 35 separate winners and only about 4.7 million a month in total, which reads as a bad market until you notice that no single product owns it. Weather and radar: 34 winners, under 7 million. Home pilates and yoga: 30 winners, under 3 million, no giant anywhere.

Small total plus many winners means the barrier is low and the ceiling is low. Huge total plus two winners means the barrier is a brand you cannot buy. Those are opposite situations and total revenue puts them in the same column.

Caveats, and they are not small: these are modelled revenue figures, not reported ones, and winner is whatever threshold you picked. Absolute values are useless. The count of independent products above a line is the part that survives, because a model that is wrong by a constant factor still counts heads correctly.