Insights is where you ask a direct question of your own data: does alcohol actually affect my recovery? Do late meals cost me sleep? Does running make me sleep better or worse?
Pick one thing you do and one metric it might change, and NOYU looks for a link.
What you can compare
Things you do come from your journal and your workouts. Any journal item you log, and activity types like running, strength or cycling. Some items support thresholds, so you can ask about three or more drinks rather than any alcohol at all.
Things that might change are your daily scores (Sleep, Readiness, Strain), measured metrics like HRV, sleeping heart rate and sleep duration, and anything you rate yourself such as mood or energy.
A few metrics are deliberately excluded. Weight and VO2 max move over months, so a link found inside a 90-day window would be noise. Training load is left out because it is already derived from your activity, so comparing the two would mostly measure arithmetic.
What NOYU needs before it will answer
At least five days where you did the thing, and five where you did not.
Below that, you get "keep logging" rather than a shaky answer. This is the same floor other trackers use, and it exists because with three days either side, one bad night can invent a pattern that is not there.
The comparison runs over the last 90 days. Longer windows sound better but drag in seasonal change, training blocks and life events, which makes it harder to tell what caused what.
How the findings are worded
NOYU reports a direction and a confidence, deliberately without percentages.
Confidence runs from preliminary to suggestive to clear. Reaching "clear" needs both a large effect and plenty of days in each group. With only a handful of days in a group, nothing rises above "preliminary" no matter how big the difference looks.
Where the two groups differ only slightly, NOYU reports no difference, rather than presenting noise as a small effect.
You will not see "alcohol reduces your recovery by 23%". That number would be false precision on a sample of a few dozen days from one person, and it is the single most common way this kind of feature misleads people.
Delayed effects
Some things do not show up the same day, so NOYU checks whether a link appears with a one or two day lag.
Overnight metrics are also lined up correctly with the day before. When you ask about late meals and sleep, Tuesday's late dinner is compared with the night that followed it, not with Tuesday morning's score.
Things NOYU already expects
Some relationships are true by definition rather than discovered. Running raises your Strain because that is what Strain measures.
These are badged Expected rather than presented as a finding. Selling you an obvious mechanical relationship as an insight into your health would be a cheap trick, and it would make the genuine findings harder to trust.
What a finding is not
A pattern in your own data is a pattern in your own data. There is no control group, and life does not hold still.
If your worst recovery days are also your drinking days, they may also be your late nights, your stressful weeks and your travel days. NOYU describes what it saw; it does not claim to have isolated a cause.
This is exactly what experiments are for. A strong pattern is a good reason to run one: change one thing deliberately for a fixed window and see what moves. See How experiments work.
Where to find it
Insights lives at its own place in the app, separate from the chat. Open it directly to see your patterns - it is the home for this analysis, and the chat will happily talk through what it shows you.
