How Discoveries work

How Discoveries finds patterns

• When it runs: automatically about every 12 hours in the background, and right away when you tap Scan now on the Discoveries tab.
• What it compares: days. For each pair of trackers, Nomos counts the days you logged both, only one, or neither. It then checks whether one makes the other more likely than usual. That ratio is the lift.
• Same day vs. the next day: it checks two timings. Do X and Y tend to happen on the same day? Does X on one day predict Y the next day? A next-day pattern is a stronger hint about cause, because X came first.
• Guarding against chance: with many trackers there are many possible pairs, and some will line up by luck. Nomos corrects for this (Benjamini–Hochberg false discovery rate, cut-off 15%), so a card needs real evidence, not just one striking coincidence.
• Quiet stretches are skipped: if nothing at all was logged for more than 7 days in a row, those days are left out so a holiday doesn't distort the result. Ordinary 1–3 day gaps still count.
• Minimum data: comparisons between trackers need at least 12 tracked days. A same-day pattern needs the two to coincide on at least 4 days. A next-day pattern needs at least 3 occurrences.
• "Symptom" trackers: Nomos treats a tracker as a symptom or outcome if its name contains one of these words (whole words, plurals and simple endings like "Headaches" or "Stressed" count): headache, migraine, fatigue, insomnia, pain, reflux, stress, anxiety, cramps, soreness, hangover, nausea, rash, bloating, tired, cold, flu, symptom, flare. "Painting" or "Fluids" don't count. Mood ratings are not treated as symptoms. A symptom is never shown as the cause of a habit.
• Hidden and archived trackers: left out of Discoveries entirely, including from the baseline. Hiding a tracker removes its cards on the next scan.
• Dismissed cards stay dismissed: cards you dismiss with ✕ stay dismissed after future scans. Cards that stop meeting the statistical thresholds disappear on their own.

Reading a Discovery card

Each card in your Discoveries feed is constructed with eight key elements:

1. A badge: the card type showing the statistical relationship (detailed below).
2. A title: A ➔ B indicates a directional trigger, while A + B represents things that co-occur together.
3. A plain sentence: clear explanation summarizing the finding with key numbers.
4. Outcome probability comparison: how often B happened with A versus without A. For example, "With Coffee 62% · Without Coffee 31%" means B showed up on twice as many days.
5. Lift: how much more likely B is with A, as a percentage. +100% means twice as likely.
6. Odds ratio: on applicable cards, another way to express relationship strength (1.0× means no relationship).
7. "Test as N-of-1 trial" button: available wherever a 2-tracker directional trial makes sense.
8. ✕ button: to dismiss cards you do not wish to track further.

At the top of the feed, you can filter by All Discoveries, Target Goals, or by individual tracker chips.

Card types

N-of-1 personal trials

A personal trial (an "N-of-1" scientific study: one person, you) tests whether a suspected trigger genuinely influences an outcome for you.

How to start: on a Discovery card that links two trackers, tap Test as N-of-1 trial. The button appears on time-lagged, high-confidence, and exploratory cards. It is not shown on routine stacks, symptom co-occurrences, mediated confounders, or attribute-level drivers.

How it runs:

1. Phase 1, Elimination (7 days): avoid the trigger completely and continue logging the outcome. This establishes your clean baseline rate.
2. Phase 2, Reintroduction (7 days): resume the trigger and continue logging. Does the outcome increase or return?

The trial card displays "Day N of …", the number of events recorded in each phase, and days tracked. You can tap Advance to Phase 2 early if needed: the transition date adjusts to that moment and Phase 2 will still run for a full 7 days. When complete, tap Finalize trial verdict. You may cancel an active trial at any time.

What makes a result fair: both phases share identical duration; logs from before you started the trial are excluded; outcome rates are computed per day and displayed normalized per week.

What trial results mean

During a trial, Nomos tracks your progress, and once finalized, evaluates event rates across both phases to determine your verdict:

The wording is deliberate: a personal trial provides supporting evidence, not clinical proof.

Tips for better discoveries

• Log consistently for at least 2–3 weeks: Discoveries require a minimum of 12 tracked days before running comparisons.
• Log symptoms when they happen and when they don't: Statistical patterns rely on contrast between active and inactive days.
• Name symptom trackers clearly: Use descriptive names (such as "Headache", not "H") so the engine can identify symptoms and avoid showing them as causes.
• Track baseline lifestyle anchors: Add a Sleep, Alcohol, or Stress tracker; Nomos uses these to catch and warn about mediated confounders.
• Mark target goals: Set key attributes as "Treat as a target goal" to unlock Target Goal Driver insights.
• Use Hide from Insights: If a tracker only introduces noise, hide it from Insights to remove it from discovery baselines.

Where the analysis happens

Every calculation—including statistical scans, Benjamini–Hochberg corrections, N-of-1 trials, and verdicts—runs entirely on your phone. Nomos requires no user account, connects to no cloud servers, includes zero analytics SDKs, and declares no network permissions in its Android manifest. Your personal data cannot leave your device.

Opening this online help page is an ordinary web browser visit that transmits none of your local logs or app state.

Frequently Asked Questions