The quiet contraction
Something has been happening across software for about three years, and because it happens one product at a time, it rarely gets discussed as a single trend.
Free tiers are shrinking.
Not disappearing overnight. Shrinking. A limit appears where there was none. A feature moves up a tier. The generous plan you signed up for gets grandfathered, then eventually not. The pricing page stops leading with the free option.
Each announcement is easy to dismiss as one company's pricing decision. Together they describe a structural shift in how software gets funded, and it changes what "free" is worth planning around.
What actually happened
Three examples, spanning very different categories.
Heroku removed its free tier entirely in November 2022. For over a decade the free dyno was where an enormous number of developers deployed their first application. It went away completely, and the ecosystem of "Heroku alternatives with a free tier" articles that followed is a reasonable measure of how many people had built on it.
Docker has tightened repeatedly rather than cutting once. Free accounts saw private repositories reduced, Scout repositories cut from three to one, Build Cloud minutes removed, and pull rate limits imposed on both anonymous and free authenticated users. Paid tiers rose in the same period. No single change was dramatic. The direction was consistent.
Evernote reduced free accounts to 50 notes and a single notebook, and later restructured its paid plans entirely. For a product whose whole proposition was "put everything in it", a 50-note ceiling is not a limit so much as a different product.
These are not three companies making the same mistake. They are three companies responding to the same conditions.
Why now
The money changed
For roughly a decade, capital was cheap and growth was rewarded over profitability. In that environment a free tier is straightforwardly rational: you are buying users with investor money, and the market pays you for having them.
When rates rose, that trade stopped working. The same free tier that read as an investment now reads as a cost with no path attached. Nothing about the product changed. What changed was the price of the money behind it.
Infrastructure stopped being nearly free
Compute and storage got cheaper per unit for years, which is what made generous free tiers affordable in the first place. That curve flattened, and then AI features arrived and pushed it the other way.
Inference costs real money per request in a way that serving a web page does not. A free tier that includes anything AI-shaped has a variable cost attached to every free user, and variable costs are what kill free plans. This is roughly why the tightening accelerated at the same time everything acquired an AI feature.
The conversion maths were always thin
Free tiers are usually justified as marketing: some percentage converts, and their revenue covers everyone else. That works when the percentage is high enough and the cost per free user is low enough.
For most products the percentage is low. Combine that with rising per-user cost and the free tier stops being a funnel and becomes a subsidy, paid for by the customers who did convert. Once that framing takes hold internally, the free tier is on borrowed time.
The pattern, in the order it usually happens
Reductions almost never arrive as "we are cutting the free plan". They arrive in a sequence.
- A new limit that does not affect you. Framed as preventing abuse. Set high enough that almost nobody notices.
- A tier appears between free and the old first paid tier. Cheaper than the paid plan, so it looks like a concession. It is where the features that used to be free now live.
- The limit tightens. Described as a clarification, or an alignment with actual usage.
- Grandfathering. Existing users keep the old terms. This is genuinely generous and it is also the clearest possible signal, because it means the company has decided the old terms were a mistake.
- Grandfathering ends.
The step worth noticing is four. When you are told you will keep your existing plan, you have been told what the company thinks of that plan. The clock has started.
How to tell which free tiers will survive
Not all of them are at risk. The question is not how generous a free tier is, it is why it exists. Three kinds, with very different life expectancies.
The demonstration
Free exists because the paid product is obviously better and you cannot tell without trying. A generous free tier here is a sales cost with a clear return, and it survives, because cutting it would reduce paid conversions.
Signals: the paid upgrade is about scale or capability rather than removing an artificial restriction, and the free tier is genuinely usable for its stated scope.
The moat
Free exists to keep people inside an ecosystem where the money is made elsewhere: hardware, an enterprise contract, a platform. This survives as long as the ecosystem does, and can be extremely stable, because the free tier is not expected to pay for itself.
Signals: the company sells something substantial that is not this product.
The unresolved
Free exists because the company has not decided how it makes money yet. This is the fragile one, and by far the most common.
Signals: no visible revenue model, no paid tier at all, or a paid tier that seems arbitrary.
An unresolved free tier is not sinister. It is unfinished. And unfinished resolves eventually, into a paywall, into advertising, or into shutdown. We covered the same question from the app side in how free apps actually make money.
The cliff, and why it costs more than it looks
The other half of this is what happens at the boundary.
Free tiers rarely taper. They stop, and the first paid tier is often a large step rather than a small one. Research on cloud providers found a substantial share of users ended up paying more than they would have elsewhere once they crossed that line, and the reason is not carelessness. It is that by the time you hit the cliff, moving costs more than staying.
You have configuration, data, integrations and habits invested. The provider knows this. The pricing reflects it.
Which produces the practical rule: the moment to evaluate a paid tier is before you adopt the free one. Once you are past the cliff you are not choosing between providers any more, you are choosing between paying and migrating, and migrating always looks worse from the inside.
The one case where free is safe
There is a category this article has been quietly ignoring, and it deserves a mention because it is the exception that explains the rule.
Software you run yourself is not subject to any of this. An open source tool installed on your own machine has no free tier to shrink, because there is no company deciding what free means. The pricing risk moves elsewhere, into whether the project stays maintained, but the specific failure described here cannot happen to it.
That is not a recommendation to replace everything with self-hosted alternatives. For most people, most of the time, the convenience of a managed service is worth the pricing risk, and pretending otherwise is how you end up maintaining a server instead of doing your actual work.
It is a useful reference point, though. When you are deciding how much to build on a free tier, the honest comparison is not free versus paid. It is free-that-can-change versus paid-that-is-predictable versus yours-but-you-maintain-it. Each has a cost. The mistake is treating the first as though it has none.
What this means for you
Assume every free tier will get smaller. Not out of cynicism, but because the pressures above have not gone anywhere. Plan for the version one step less generous than today's.
Know why it is free before you commit. Demonstration and moat are reasonably safe. Unresolved is not. Thirty seconds on the pricing page usually tells you which you are looking at.
Check the exit before you build on it. Export function, data format, and whether export still works after downgrading to free. This single item decides how much a pricing change can cost you, and it is easiest to check on day one. More in what happens to your data when an app shuts down.
When a cut is announced, export first and decide second. Export tools during a transition are built quickly, run under load, and are frequently the first thing to break.
Separate "good" from "free". When a free tier ends, the honest question is whether you would have paid for this from the start. Often the answer is no, and the announcement has done you a favour by forcing a decision you had been avoiding. That is the same audit as subscription creep, arriving on someone else's schedule.
Where we sit, including the uncomfortable part
We build apps, so this is a disclosure rather than commentary from outside.
PackPilot and PawDex are both free on Google Play. No advertising, no in-app purchases, no subscription.
By the framework above, that puts us squarely in the unresolved category, which is the fragile one. Skipping that would be convenient and would make the rest of this article worthless.
The honest position: the intended path is a paid tier once each app has enough users to support one, not advertising. Advertising would mean optimising for time on screen, and both apps are built to be finished with quickly. That is a plan, not a guarantee, and plans change under financial pressure.
What you can check today rather than take on trust: neither app contains an advertising SDK, which is why both permission lists are short, and neither has engagement mechanics. When a paid tier does arrive, the free tier will not drop below what it does today, because the argument of this whole piece is that doing so is a broken promise rather than a pricing decision.
Hold us to that. It is written down and dated.
The short version
- Free tiers are shrinking across software, one product at a time, for the same three reasons: expensive capital, rising per-user cost, thin conversion.
- The sequence is predictable. Grandfathering is the signal, not the reprieve.
- Ask why a free tier exists. Demonstration and moat survive. Unresolved does not.
- Evaluate the paid tier before adopting the free one, because the cliff is where the leverage sits.
- Check the export before you build. It decides what a pricing change can cost you.
Related Reading
- How Free Apps Actually Make Money covers the same question on the mobile side.
- Subscription Creep: How to Audit What You Are Paying For is what to do once the charges start.
- What Happens to Your Data When an App Shuts Down covers exit cost in detail.
- How to Choose an App You Will Still Use in a Year puts the funding question first for a reason.
- The Lifetime License Comeback covers why one-time purchases are emerging as the sustainable alternative to shrinking free tiers.
- How to Find Hidden Subscriptions covers what to do once a few of these have crept in.
Sources and further reading: Docker Hub policy changes, Evernote free account limits, cloud free tier comparison.
Frequently Asked Questions
Three pressures at once. Cheap capital dried up, so growth funded by losses stopped being rewarded. Infrastructure costs rose, particularly anything touching AI inference. And most free users never convert, so a large free base became a cost centre rather than a marketing channel.
Heroku removed its free tier entirely in November 2022. Docker has repeatedly tightened free limits, cutting private repositories and Scout repositories and imposing pull rate limits. Evernote reduced free accounts to 50 notes and a single notebook. The pattern spans categories rather than being specific to developer tools.
Only if you know why it exists. A free tier that serves a clear commercial purpose, such as demonstrating a product whose paid version is obviously better, tends to survive. A free tier that exists because the company has not decided how to make money yet does not.
The point where a service's free allowance ends and pricing begins, often as a steep jump rather than a gradual increase. Research on cloud providers found a large share of users paid more than they would have elsewhere once they crossed it, because migrating after the fact costs more than choosing correctly at the start.
Watch for grandfathering language, new limits described as clarifications, a paid tier appearing between free and the previous first paid tier, and pricing pages that stop showing the free option prominently. Each usually precedes a reduction.
Export your data first, before deciding anything. Then work out whether you were using it because it was good or because it was free. Those lead to different answers, and the second is more common than people expect.