Google Analytics is free, has more features than you will ever use, and is backed by the largest advertising company on earth. Plausible is a two-person web analytics tool with fewer features — and it charges money.
By common sense the story should end right there. Instead, Plausible makes millions a year and is doing fine.
It did not win anywhere on the feature table. The whole product is one sentence: we don't feed your visitors' data to an ad machine. Google cannot say that sentence — not because it wouldn't like to, but because saying it would mean shutting down its main business.
The same structure, four times
Tally vs Typeform. Typeform bills by responses received — the more successful your form, the more you pay: a fine levied on your success. Tally inverted it: unlimited forms free, money only for advanced features. Could Typeform copy that pricing? Sure — and its revenue would crater the same quarter. So it doesn't, and watches Tally grow inside a "saturated" market.
TypingMind vs ChatGPT Plus. A $20 monthly subscription makes light users subsidize heavy ones. TypingMind sells a one-time client where you bring your own API key and pay for what you use. Will OpenAI drop subscriptions for metered billing? Subscription revenue is its cash-flow floor. It can't move.
Plausible vs Google Analytics. Covered above: the price of free is that your visitors are the raw material. The position "pay money, own your data" is nailed shut for Google by its own model — permanently vacant.
The lock-in pattern in language-learning tools. The incumbents in this space retain users by trapping data: word lists that won't export, learning history lost on downgrade. Which makes "free one-click export to Anki" a wedge — the only way an incumbent can match it is to give up the lock-in it lives on. The reading product I'm building takes exactly this route, so this essay is not a spectator's summary; it is a page from my own playbook.
Why this wedge cannot be closed
Put the two kinds of attack side by side.
Attack the features: you ship something good, the incumbent's product team looks at it once, and next quarter they ship the same thing. They have more people and wider channels; your reason-to-switch existed for ninety days.
Attack the business model: your way of charging strikes at their way of charging, and every response available to them runs through self-harm. Google cannot stop monetizing data; Typeform cannot stop billing per response; OpenAI will not unbundle its subscription. The incumbent isn't unwilling to close the gap — its revenue structure institutionally forbids it.
A feature advantage has a half-life of one quarter. A business-model wedge has the half-life of the incumbent's entire revenue structure.
How to find yours
The incumbent's pricing page is the target map. More precisely:
Find the point where users' complaints overlap with the incumbent's revenue. Complaints come in two kinds: things it hasn't fixed yet (it will — stay away), and things it profits from. The second kind never gets fixed. Billing by response, data that won't export, free-in-exchange-for-surveillance — each has been cursed at for years, and each is alive and well, because each is a revenue pillar.
Where to find the cursing? The incumbent's low-score Trustpilot pages, Reddit threads titled "X alternative", the comment sections of cancel-subscription tutorials. The requirements document is already written; it's lying there.
The honest boundary
Of the four cases, three are wedges in pricing structure (free-vs-paid, the billing unit, subscription-vs-buyout) and one in data lock-in. The more distant form — attacking a growth model with something like "we refuse to build hoarding features" — has no solid case yet. So the reliable range of this essay is: look at how it charges and how it retains, and find the thing it cannot change inside those. Beyond that, I'm waiting for evidence like everyone else.
One precondition, easy to miss: a wedge pries open a proven market — the incumbents' existence is the demand validation. On empty ground there is no incumbent, hence no business model to attack; that is a different game with a different risk structure.
Take one sentence with you: don't ask "can I build a feature they don't have" — ask "which part of how they make money do users hate, and which of those can they never change." That seam is where your whole company gets to live.