The claim the app is willing to make
Everything on this site rests on one distinction. The app reports what lines up in your log. It does not report why, and it does not report what to do about it.
That can sound like a limitation being dressed up as a virtue. It's closer to the opposite. The narrower claim is the one that can be checked. The wider one would have to be taken on trust.
The five checks
Lagged correlation. Whether one logged thing lines up with another some days later, across a range of gaps rather than a single one.
Co-occurrence. Whether a tag and a level appear together more often than chance would explain, measured against how often the tag appears at all.
Sequence. What tends to be logged in the days leading up to a run of rough days.
Change point. A date where the typical level of a logged series moves and stays moved, rather than dipping and recovering.
Weekday cycle. Whether the log carries a weekly rhythm.
All five run on the phone, over your log only. There is no account, nothing is uploaded, and the analysis does not need a connection.
What a finding looks like
This looks like a real pattern.
After days when you logged more sleep than usual, your logged energy tends to be lower about 2 days later.
Drawn from 48 pairs of days, between March and August 2026.
Three parts, each doing a job. The first line is the confidence. The second is the association: what came before what, and how far apart. The third is the evidence: how many pairs of days, over what range of dates.
Read it backwards to see what each check contributed. The association came from a lagged correlation, the only one of the five that would have found a two-day gap. The count came from how many pairs of days in the log could be compared at that gap. The confidence came from how strongly and how consistently those pairs lined up.
Confidence, in words rather than numbers
A correlation coefficient with no context is an invitation to over-read. Confidence here is one of three sentences, and they are fixed.
- EmergingEarly signal. This might be a pattern, I need more to be sure.
- ProbableThis looks like a real pattern.
- StrongThis is a consistent pattern in your history.
They are never paraphrased and never tuned. Two findings at the same level read identically, whether they arrived in your second month or your tenth. That consistency is the point: the wording carries information only if it stops carrying anything else.
Why the day count is printed
A sentence drawn from 48 pairs of days and a sentence drawn from six are different sentences. Without their counts they look the same on a screen, and the thin one borrows credibility from the format.
Printing the count moves the judgement to you and hands you the material to make it. A finding with few days behind it is labelled as early rather than presented as true.
The problem with checking many things
The app checks a lot of pairings. Checking a lot of pairings is how false findings get made: run enough comparisons and some will line up by chance alone.
There's no way to make that disappear, so the design admits it instead. The confidence wording stays conservative, every claim arrives with its evidence, and findings lose their level when the log stops supporting them. Any finding can also be hidden for a month or marked as wrong.
Where these methods come from
None of the five checks was invented here. They are standard statistical methods, and the risk of running many of them at once is a standard problem with a literature of its own. The references are below so the claims on this page can be checked against something other than this page.
- Change pointKillick, R., Fearnhead, P. and Eckley, I. A. (2012). Optimal Detection of Changepoints with a Linear Computational Cost. Journal of the American Statistical Association, 107(500), 1590–1598.
- Checking many pairings at onceBenjamini, Y. and Hochberg, Y. (1995). Controlling the False Discovery Rate: a Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society, Series B, 57(1), 289–300.
- Lagged and cross-correlationBox, G. E. P., Jenkins, G. M. and Reinsel, G. C. Time Series Analysis: Forecasting and Control.
Citing the literature is not the same as claiming a paper endorses this app. None of these authors know it exists. They are here because a page making statistical claims should point at something outside itself.
What it refuses
It will not tell you why. It will not promise that a finding will appear, since that depends on your log and on how often the thing occurs. It will not describe itself as anything other than a journaling and reflection tool, and it is not therapy and it is not a mental health service.
A few weeks of logging is the honest floor before any of the five checks have enough to work on. Life Pattern Journal is not released yet. The waiting list on the front page sends one email now to confirm the address, then two more, ever: one when the beta opens and one when it launches.