Writing
Do habit trackers actually work?
The honest answer is: partly, for reasons that have little to do with the features most habit apps compete on.
The part with evidence behind it
Self-monitoring — the plain act of recording what you did — is one of the better-supported techniques in the behaviour-change literature. It appears consistently in taxonomies of effective intervention components, and it shows up as a useful ingredient across domains as different as diet, physical activity and medication adherence.
The mechanism is not mysterious. Recording a behaviour makes it visible to you, and a lot of the behaviour people want to change is behaviour they are not accurately aware of. Most people cannot tell you how many days last month they actually went for the walk. The estimate and the count differ, usually in a flattering direction.
So: does the act of tracking help? On balance, the evidence says it plausibly does. That is a real finding and it is the one habit apps are implicitly citing when they say tracking works.
The part with much weaker evidence
What that literature does not establish is that streaks, badges, levels, or daily push notifications help. Those are engagement mechanics, and engagement is a metric that serves the app before it serves you.
There is a reasonable theoretical worry about them, too. When a behaviour is tied to an external reward — a number that goes up, a chain that must not break — the reward can become the reason for doing it. When the reward stops, so does the behaviour. This is contested territory in the research and the effect sizes are argued about, so it should be held as a caution rather than a proven mechanism. But it matches something most people have experienced directly: the streak breaks, and the habit that the streak was supposed to be protecting quietly ends with it.
The tell is that the streak was never the habit. Missing a Tuesday does not undo eleven weeks of running. It only undoes the number.
The question tracking usually fails to answer
Here is the gap. You track a habit for three months and you now know, precisely, that you meditated on 71 of 92 days. What you still do not know is whether it did anything.
That is the question you started with. Almost every tracker answers a different one — how consistent were you — and then presents consistency as though it were the outcome. It is not. It is the input, and an input is only worth sustaining if it produces something.
Answering the real question needs two things most apps do not do: recording an outcome as well as a behaviour, and then actually running the comparison between them.
What a useful answer looks like
It looks like a relationship with a number attached, and an admission when there is not one. "Your mood is higher on days following seven or more hours of sleep, r = 0.31 across 46 paired days" is a finding. "You are on a 12-day streak" is a scoreboard.
It also looks like a system willing to tell you nothing was found — because a system that always finds something is not measuring, it is generating. If an app has surfaced an insight every single week you have used it, that is evidence about the app rather than about you.
This is the thing Crescendo is built around: it correlates habits against sleep and mood, and reports a relationship only when it clears 20 paired days, a coefficient of 0.25 or stronger, and a p-value corrected for the number of comparisons run. Most weeks, early on, that means it reports nothing.
So should you use one
Probably, with clear eyes about which part is doing the work. Record the behaviour, because self-monitoring is the ingredient with support behind it. Record at least one outcome you care about, because otherwise you cannot ever answer the question you are asking. Treat the streak as decoration.
And give it enough time to say something. Twenty paired days is roughly the floor at which a personal correlation is worth reading at all, which in practice means about a month before any of this becomes interesting.