Diary Tags and how patterns are found in them

Overview

A Diary Tag is a thing you record about a day that no sensor can see: a medication taken, a stiff back, three coffees, a walk outside. Grio’s built-in tags come in three recording shapes, and the shape is chosen for one reason: it matches how people actually remember the thing.

How it’s calculated

Three shapes, chosen by recall

This favours consistent, familiar answers over finer-looking data. The default templates deliberately do not ask for specialist estimates such as caffeine in milligrams, or invent precision for a routine whose duration nobody tracks. Caffeine is counted in drinks, not milligrams. Meditation is Yes/No, not minutes toward ten.

You can still customise a recurring tag when your own mental model differs.

Shape changes keep history readable

A recording definition is date-effective. Editing a tag’s shape takes effect on the local date of the edit; earlier dates keep the definition that governed them. Grio never coerces an old counter, Yes/No, or scale value into a different meaning. If the effective date already has an entry and the shape changes, that one entry is deleted in the same transaction, after you confirm.

Derived analysis respects that boundary: a pattern ignores observations from before the current shape revision became effective, because combining incompatible definitions would produce a number about nothing.

How patterns are found

A Diary Tag Condition Pattern pairs each tag observation against the following day’s condition response, over a bounded history.

• Numeric tags stay continuous: each recorded amount is plotted against the next day’s condition. Grio does not bin values into high/low groups, does not infer a recommended range, and does not present the association as a cause or a medical effect. Fixed bins and clustering were both rejected because they manufacture thresholds that shift as your records change.
• An explicitly stored zero is an observation: an absent entry or a null value is missing, and is never imputed as zero.
• Only the exact following day is paired: Days without a condition response stay as unpaired evidence rather than being dropped or interpolated.
• The shape is summarised, not fitted: the server derives at most four x-ordered mean anchors. Four anchors can express as many as two turns (a U-shape or an inverted U), and the app draws a smooth path through them without adding anchors, fitting coefficients, or extending the line beyond the first and last observed summaries. Raw daily observations stay in the foreground.

Missions mirror into tags, one way

A mission revision may name one linked diary tag. Each completion or progress change of its occurrences writes a mirrored entry into that tag for the same local date:

• Completion writes Yes to a Yes/No tag.
• Numeric progress writes the value to a counter tag, which is also the only shape an allowance mission may link, because “coffee at most two” mirrors cups used.
• Expiry writes nothing. An empty line is not a lapse.

Nothing flows back. A diary edit never completes, progresses, or reverts a mission. A mirrored entry remembers which occurrence created it, so a revert removes only an entry the mission still owns. Once you edit that entry yourself, the mission leaves it alone from then on.

Edge cases and fallbacks

• A hidden tag makes a mission’s link dormant rather than severed. Unhiding resumes mirroring.
• A tag may be linked by at most one current mission, and a mission links at most one tag.
• Units are not compared: a mission’s unit is a fixed enum and a tag’s unit is free text you typed, so the picker shows the tag’s unit and the choice is yours.
• Scale and to-do tags do not link: a subjective reading and a single occurrence are not what a mission result means.

How to read your score

• Read a pattern as “worth investigating”, never as “proven”: a bounded personal history is exactly the kind of evidence that produces convincing coincidences.
• Watch the sample count: it is shown in the row summary and in detail for a reason.
• The shape carries more than the direction: an inverted U (better up to a point, worse beyond it) is a real and common finding that a single correlation number would erase.

What to do with it

• Record fewer tags, more consistently: Ten tags kept for three weeks produce nothing; three tags kept for three months produce a readable history.
• Record the zeros: “No coffee today” is an observation. Leaving it blank is not.
• Use the shape the question deserves: If you will not reliably recall the use Yes/No. The data you actually enter beats the data you meant to.

Troubleshooting & FAQs

Does a pattern mean the tag caused the change?

No. Grio describes an observed association over a bounded history and is explicit that it is not a cause or a medical effect. A monotonic association is only described when the evidence supports it, and the raw observations stay visible so you can judge it yourself.

I changed a tag from minutes to Yes/No. What happened to my history?

Earlier dates keep the definition that governed them, and the new shape applies from the edit date onward. Old values are never coerced into a new meaning. If the effective date already had an entry, that single entry is deleted after you confirm. Pattern analysis then ignores observations from before the current shape took effect.

Why does completing a mission also write a diary entry?

Because a mission may name one linked diary tag, and its completions and progress changes mirror into that tag for the same date. The flow is strictly one-way: the mission is the authority for the action and the diary reflects it. Editing the diary entry never completes, progresses, or reverts a mission.