Look at the software you signed up for this year. Not the ones you use — the full list. The password manager will have it even if memory does not.
For most people the ratio is grim. A dozen trials, two survivors. And the two that survived were rarely the two that impressed most during the trial.
This is worth examining, because the industry has trained us to evaluate software at exactly the wrong moment.
Adoption is not retention
Adoption is easy to measure and easy to manufacture. A good landing page, a generous free tier, and a fifteen-minute setup will produce adoption. Most product teams optimize for it because it is the number that moves fastest.
Retention is different. Retention is what happens after the novelty is gone, when the tool has to compete not with its alternatives but with the way you already do things. That competition is much harder than it sounds, because the incumbent is usually a habit — and habits do not require a login.
The gap between the two is where most software dies. Not in evaluation, where it looked promising. In week three, quietly, without a decision ever being made.
The worst week is the real test
Here is the pattern worth noticing: tools are trialed during calm periods and abandoned during busy ones.
That is backwards, and it is understandable. You try a new note system on a Sunday afternoon, when you have the patience to import things and set up templates. It feels good. It is also the least informative possible test, because on a Sunday afternoon almost anything works.
Then a deadline arrives. Under time pressure, people do not choose the best tool. They fall back to whatever requires no thought — the app already open, the folder already there, the workflow that has never once failed them. Anything that demands a decision gets skipped. And a tool that gets skipped during the week that mattered rarely comes back afterward.
If you want to know whether software will last, introduce it during a bad week. It is an unpleasant test and a fast one.
Durable tools subtract
The tools that survive tend to share a structural property: they remove a step you were already taking, rather than adding one you now have to remember.
The distinction sounds small. In practice it decides everything.
A tool that subtracts gets easier to use the busier you are, because the pressure that would kill a new habit instead reinforces the shortcut. A tool that adds gets harder to use exactly when you most need it to work, because it competes for the attention you no longer have.
Most software marketing describes additions. New capability, new surface, new workspace. Genuine subtraction is harder to sell, because the pitch is unglamorous: this removes a thing you were doing anyway. But that is the category with the survival rate.
The honest exception
This principle has a real limit, and pretending otherwise would be too neat.
Some tools legitimately add a step, and are worth it. Version control adds ceremony to writing code. Double-entry bookkeeping adds work to recording a transaction. Both survive because the added step buys something you cannot get otherwise — recoverability, or accuracy under audit.
What separates those from the software that quietly disappears is that the added step is load-bearing. Skip it and something visibly breaks. Most productivity tools cannot make that claim. Skip them and the day proceeds normally, which is precisely why they get skipped.
So the sharper question is not whether a tool adds or subtracts. It is whether anything actually breaks when you stop using it. If nothing does, you have your answer about how long it will last.
Running a trial that predicts something
A few adjustments make evaluation more honest:
Trial during a deadline week, not a quiet one. The tool should have to earn its place against real pressure. A tool that only works when you are calm is a tool for calm days, which are not the days you need help.
Decide what it replaces before you start. If you cannot name the step it removes, it is probably an addition wearing the language of an improvement. Write the sentence down: this replaces ___. Difficulty finishing that sentence is itself the result.
Check the exit before the entrance. Whether you can leave with your work intact tells you more about a company’s confidence than its feature list does. This deserves its own treatment, and gets it in The real cost of switching.
Ignore the first week entirely. Everything is interesting in week one. The signal starts in week three, when the tool has to survive on usefulness rather than novelty.
Why the market produces so much of this
None of this is a story about bad intentions. It is mostly a story about incentives.
Software that is easy to start is easy to sell, and the metrics that fund the next round tend to reward starts. Retention shows up later, in cohorts, long after the decisions that shaped the product were made. A team can build something genuinely well-crafted and still be aiming at the wrong number.
Pricing shapes this too — a one-time purchase and a subscription reward completely different behavior from the people building the product. That mechanism is worth understanding on its own, and we look at it in Subscription is a design decision.
What this leaves you with
The useful reframe is small. Stop asking what a tool can do, and start asking what it still does for you on the worst day of the month.
Almost every tool performs well on a good day. That is the condition under which they are demonstrated, evaluated, and reviewed — and it is the condition that tells you the least. Capability is easy to verify and nearly worthless as a predictor. Survivability is harder to test and tells you almost everything.
The software that lasts is rarely the most impressive software you tried. It is the one that quietly stopped being a decision.