Common ones — one tap to add:
Or add your own, with whatever range makes sense for it:
● raises · ● lowers · more dots = higher confidence
| Factor | Effect | Impact / SD | Strength | r | Conf. |
|---|
How your habits cluster, and how much of the day‑to‑day variation a few patterns capture.
Effect: whether more of this factor tends to raise or lower pain, from a multi‑factor regression (controls for the others).
Impact / SD: pain points (0–10 scale) change for a 1 standard‑deviation change in the factor — a fair way to compare units.
Strength: random‑forest permutation importance — how much the model relies on this factor, capturing non‑linear effects.
r: simple pairwise correlation with pain.
Conf.: our confidence, blending statistical significance, agreement between models, forest importance and how much data you have.
Same‑day analysis misses anything that hits with a delay (post‑exertional crashes, migraine build‑up). A separate pass checks each factor a few days back, after controlling for your own recent trend, so a slow drift isn't mistaken for a driver. A confound note appears when two factors move together closely enough that we can't confidently credit one over the other from observation alone.
Log ~15 days to unlock experiments.
Suggested test levels are capped to general safety limits (e.g. ≤400mg caffeine/day, ≥4h sleep/night) — not personalized medical advice. Check with a clinician before testing anything that feels wrong for you.
logged predicted next — tap a day to log or remove
Protected blocks with a clear start and end. The point isn't discipline — it's that switching tasks leaves "attention residue" that measurably degrades the next thing you do, and it's worse when you left the last thing unfinished (Leroy, 2009). Finish or park a task before you switch.
Consistency over time, not an unbroken chain. Monitoring genuinely helps — it's one of the better-replicated findings in the literature (Harkin et al., 2016, 138 studies) — but streak mechanics aren't, so this tracks your rate and doesn't punish a miss.
Shrink it until it's almost trivially easy. Motivation is the least reliable of the three things a behaviour needs, so lean on the other two — ability and a prompt (Fogg, 2009).
This becomes an if-then plan: if [cue], then I will [habit]. That specific form has a solid, replicated effect on follow-through where general intentions don't.
Identity framing tracks with habit strength noticeably better than outcome framing does.
Encrypted on this device with everything else. Never analysed, never scored, never turned into a factor — it's yours to write in, not more data about you.