Strategic freemium — the death of "give it all away"

The market repriced free while nobody announced it

Freemium works fine. Unlimited freemium is what stopped working. In the last two years Slack capped free workspaces at 90 days of message history, Notion cut its free block allowance and then reworked it again, HubSpot pared back what its free CRM tier actually does, and Calendly moved features that used to be free behind a paid line. Four companies with almost nothing in common except a growth model, all pulling the same lever in the same window.

That is not a coincidence, and it is not a coordinated retreat. It is a repricing of what “free” is allowed to cost the business that offers it. The 2018 version of freemium, give away nearly everything and monetize the sliver of users who need admin controls or SSO, worked because the inputs were cheap and the funnel math forgave a lot. Both of those conditions changed. The companies that noticed early are tightening now, calmly, before the numbers force a panic. The ones that didn’t will tighten later, badly, under duress.

Why unlimited worked in 2018 and stopped working

The unlimited free tier was a bet on marginal cost. Storage was trending toward zero, compute was cheap, and the only real expense of an idle free user was a row in a database. If a free account cost you a fraction of a cent a month and one in 20 converted, you gave away everything and let volume do the work. Slack’s free tier was a growth engine precisely because a free workspace cost Slack almost nothing and seeded the paid expansion inside the same company.

Two things broke the math. First, AI features have a real marginal cost. A free user running inference against your model is not a free row in a database; they are a bill from your GPU provider that arrives whether or not they ever upgrade. Notion AI, HubSpot’s AI tooling, and every “ask your workspace” feature shipped since 2023 turned the cheapest user into a metered one. This is the same shift that pulls AI-native gross margins well below classic SaaS. Second, the funnel got more expensive at the top. Organic and paid acquisition both cost more than they did, which means each free user has to carry more expected value to justify the cost of getting them in the door. When acquisition was cheap and the product was cheap to run, you could afford a fat free tier as a top-of-funnel machine. When both got expensive, the fat free tier became a subsidy the business was funding out of its own margin.

Unlimited freemium was arbitrage on cheap inputs, and the inputs stopped being cheap.

Most teams can’t name which of the four models they run

Freemium isn’t one strategy. It’s four, and most teams can’t say out loud which one they’re running, which is why they tighten the wrong lever.

The first is freemium as acquisition, where the free tier exists to get the product into a company so it can spread. Slack’s original model. Here the free tier’s job is reach, and you gate on the things that only matter once the product is load-bearing: history, integrations, admin.

The second is freemium as trial, where free is a permanent evaluation and the expectation is that serious use converts. Calendly leans this way. You gate on the features that separate a curious user from a committed one, and you make the committed use case obviously worth paying for.

The third is freemium as network, where free users are the product for paid users because they create the graph everyone else pays to reach. This is where free has to stay genuinely open, because the day you throttle the free side you starve the paid side. Tighten here and you break the thing you’re monetizing.

The fourth is freemium as land-and-expand, where free seeds a single team inside an org and revenue comes from expansion across seats and functions. HubSpot’s free CRM is this. You keep the individual experience whole and gate on the things a growing organization needs: seats, automation, reporting.

The tightening story looks reckless if you assume every company runs the same model. It doesn’t. Slack tightened an acquisition tier by capping a feature (history) that doesn’t slow adoption but does make the product painful to rely on for free forever. That’s a precise cut, not a blunt one. The mistake is a network-model company copying an acquisition-model company’s playbook and throttling the free side that its paid side depends on.

The upgrade line is a design decision, not a pricing one

The question a tightening move has to answer is not “what can we charge for” but “at what moment does a user cross from evaluating us to depending on us.” That moment is where the line goes. Everything before it stays free. Everything after it is where upgrade lives. This is the same instinct behind reading product-qualified signals as a sequence of adoption rather than a single conversion event.

Slack drew the line at time. A free workspace works fine for a project; it becomes untenable once your team’s institutional memory lives in it and the memory expires at 90 days. Dependence is the trigger. Calendly drew it at feature depth, where one-off scheduling is free but the workflows a business runs its calendar on are not. HubSpot drew it at organizational scale. In each case the free tier still does the job it advertises. What changed is that the moment of real reliance now has a price attached, and that moment is chosen deliberately rather than left as the accidental edge of the free tier.

The failure mode is drawing the line at frustration instead of dependence. When the free tier gets worse at the thing it was supposed to do well, users don’t upgrade; they leave, and they tell people why. The line has to sit at the point where staying free is a choice against your own interest, not at the point where the free product is simply annoying.

How to tighten without a revolt

The mechanics matter as much as the strategy. A tightening move that’s strategically right can still detonate if it’s executed like an ambush.

Grandfather the users who are already dependent. The people most likely to churn loudly are the ones who built their workflow on the old terms, and they cost almost nothing to keep whole. Announce the change before it lands, with a clear reason that isn’t insulting. “We’re changing this to keep building the product” is fine; “we’ve updated our plans” while quietly deleting a feature is how you earn a front-page thread. Move features up a tier, don’t delete them. A user who sees where their feature went upgrades; a user whose feature vanished leaves. And tighten one lever at a time, so that when the numbers move you know which cut did it.

The companies tightening now are not abandoning freemium. They are correcting a version of it that only ever penciled out on inputs that no longer exist. Free was always a bet on marginal cost, and that bet has been repriced. The teams that redraw the line on purpose keep the growth engine. The teams that wait until the margin forces their hand redraw it in a panic, at the point of maximum frustration, and pay for it in the exact users they most wanted to keep.