Habit tracking
A habit tracker that is allowed to tell you it did not work
Most habit trackers count. This one also checks whether the counting meant anything, and says so plainly when it did not.
What is Crescendo as a habit tracker?
Crescendo is a habit tracker that records daily habits alongside sleep and mood, then tests whether a habit is statistically related to how you feel. It reports a relationship only when it clears three thresholds at once: at least 20 paired days of data, a correlation of 0.25 or stronger, and a Bonferroni-corrected p-value of 0.00625 or lower. When nothing clears them, it says nothing was found.
How habit tracking works here
You define the habits you care about and tick them off as you go. That part is ordinary, and it should be — a tracker that makes logging clever makes logging a chore, and the data stops arriving within a fortnight.
What is different is what happens to the ticks afterwards. Alongside habits you log sleep and mood, and the app fixes its hypotheses in advance — sleep against training, meditation against the following day's mood — then tests them. Fixing them in advance is the part that matters: testing every pair of variables until one looks impressive is how you manufacture a finding out of noise.
Because several hypotheses are tested at once, the p-value threshold is corrected for that. Without the correction, running eight tests at the usual 0.05 makes at least one false positive more likely than not.
What happens when you miss a day
The streak bends rather than resetting to zero. A single missed day inside an otherwise consistent month is not evidence that the habit has collapsed, and a counter that says it is teaches people to stop opening the app after a slip — which is the exact moment the app was supposed to be useful.
The weighting is explained in full on the forgiving streaks page, including what it will and will not forgive.
Who this is for
People who have used a habit tracker before, kept it for six weeks, and stopped because the numbers never told them anything they did not already know. The trade is explicit: for the first three weeks the insights screen says, in effect, not yet — because twenty paired days is a real threshold and there is no honest way around it.
It is a poor fit if you want a tracker that always finds something encouraging to say. That product exists and is easier to enjoy.
How it differs from other habit trackers
Three ways, all of them consequences of the same decision. It can return a null result. It corrects for multiple comparisons. And it treats a missed day as a data point rather than a moral event.
The cost is that it is less immediately satisfying. Two of the hypotheses in the demo dataset are deliberately weak and are correctly rejected — which is how you can tell the filter is doing something rather than decorating a conclusion that was reached in advance.
At a glance
- Minimum data before a finding
- 20 paired days
- Correlation threshold
- |r| ≥ 0.25
- Corrected significance threshold
- p ≤ 0.00625 (Bonferroni)
- Missed days
- Weighted, not reset to zero
- Journal and mood data
- Never sent to an ad platform
- Export
- Your full history, on request
Common questions
- How long does it take to see anything useful?
- About three weeks of consistent logging. The correlation tests need at least 20 paired days before they will report anything, so until then the insights screen tells you it does not have enough data yet. Habit and streak tracking work from day one; it is the statistical findings that have to wait.
- How many habits should I start with?
- Between one and three. More than that and the daily logging becomes the habit you are actually practising, and it is usually the first one abandoned. You can add more once the first ones are running without effort.
- What happens if I break a streak?
- Nothing resets to zero. Streaks are weighted, so one missed day inside a consistent stretch reduces the figure slightly rather than erasing it. The reasoning and the exact weighting are set out on the forgiving streaks page.
- Is Crescendo free?
- There is a free tier that covers habit tracking, sleep and mood logging, and the correlation analysis. Current limits and prices are on the pricing page.
- Can I track more than one habit at a time?
- Yes, and the correlation analysis is designed for it — the Bonferroni correction exists precisely because several habits are being tested at once. Note that each additional hypothesis makes the significance threshold stricter, which is the honest consequence of asking more questions of the same data.
- Does my journal or mood data get used for advertising?
- No. Journal text never leaves your account — not to an ad platform, not into a training set, and not into our own analytics, where the event recorded when you write is a word count rather than the words. Mood and sleep are treated the same way under every consent setting.