The short answer: a streak has no denominator. It counts the days you succeeded and never the chances you had, so a quiet week and a week you avoided five opportunities produce identical numbers. And because its only two states are intact and broken, a single lapse deletes months of real progress — which is a property of the counter, not of your habit.
Streaks are the default in habit tracking for a good reason: they are legible. One number, going up, and you know instantly whether you are winning. Nothing else in behaviour change is that easy to read at a glance.
The trouble is that legibility is bought by throwing almost everything away. Here is what a streak actually knows about you: how many days in a row you ticked a box. Here is what it does not know: how many chances you had, how many you turned down, whether the streak is going up because you are trying or because the bar is low, and whether today was hard or trivial.
Problem one: there is no denominator
This is the root of it, and the rest follows.
Say your habit is “start a conversation with someone new.” On Tuesday you were at a desk all day and spoke to nobody. On Wednesday you were at a conference, had perhaps fifteen openings, and took none of them.
A streak records those two days identically.
Both are blank. Both break the run. But only one of them is a problem, and it is the one you would never have guessed from the tracker. Tuesday was circumstance. Wednesday was avoidance, and it is the only day of the two worth talking about.
Any measure that cannot separate those is not measuring the habit. It is measuring the calendar.
Problem two: the punishment is disproportionate
Sixty days of genuine effort. One evening you get home at midnight from a hospital visit. The counter reads zero.
No sensible account of behaviour change says those sixty days stopped existing. The reset is not a fact about your habit; it is a rule someone wrote into an app, and it is far harsher than the evidence justifies.
Lally, van Jaarsveld, Potts & Wardle (2010), How are habits formed: Modelling habit formation in the real world, European Journal of Social Psychology 40(6), 998–1009. Ninety-six participants tracked a daily behaviour for twelve weeks. Automaticity took a median of 66 days, ranging from 18 to 254 — and missing a single opportunity to perform the behaviour did not materially affect the habit-formation process.
Read that last clause again, because it is the whole argument. The one event a streak treats as catastrophic is the event the research found to be close to irrelevant. Streaks are not modelling how habits form. They are modelling how slot machines hold attention.
Problem three: it teaches you to protect the number
Give someone a fragile counter and they will start defending it, and defending a counter is not the same activity as building a habit.
You know the behaviour if you have ever held a streak: the 11:58pm token effort. The one-minute lesson that keeps the number alive. Choosing the trivial version of the habit on a hard day, not because it is what you needed, but because the counter does not distinguish between them.
The metric has quietly replaced the goal. And the app cannot tell, because from where it is standing a ticked box is a ticked box.
Problem four: the freeze is an admission
Almost every serious streak app now sells or rations some form of repair — a freeze, a shield, a restore.
Think about what that concedes. The mechanic is too punishing to leave running, so it needs an escape hatch. Rather than fix the measure, the fix is sold back to you at the exact moment you are most anxious about losing something you worked for.
A number that has to be insured against is a badly designed number.
What to count instead
Keep the part that works — one legible number — and give it a denominator.
Log two things rather than one: the opportunity, and whether you took it. Your headline figure becomes actions divided by opportunities. That is an action rate, and it behaves the way a streak does not:
- A quiet day is not a failure. No opportunities means nothing added to either side. Your rate is unchanged, because nothing happened, which is the honest result.
- An avoided day counts properly. Fifteen chances and none taken moves the number hard. It should.
- One bad day is survivable. Miss one chance in thirty and you are at 97%. There is nothing to reset, so there is nothing to protect, so there is no reason to fake it at 11:58pm.
- You find out the problem was never scarcity. Most people who start counting the denominator discover their supply of opportunities was always fine. The gap is between noticing one and acting on it.
That last point is usually the uncomfortable one. It was for me: when I first counted both columns for a conversation habit, I was acting on about twenty per cent of what came up. That number is the reason this app exists, and the whole story is here.
When a streak is genuinely fine
To be fair to the mechanic: if a habit is entirely within your control and truly meant to happen every single day — medication, flossing, making the bed — then opportunities and days are the same thing, the denominator is always one, and a streak is a rate in disguise. Use it.
The mechanic fails when the habit depends on circumstances arising. Speaking up in a meeting. Starting a conversation. Choosing water over a drink. Applying for a job. In every one of those, the number of chances varies day to day, and a measure that ignores the chances cannot tell you anything reliable.