Field note 01 · Wearable data

Before You Trust a Wearable Trend, Check Its Source

A daily graph can look complete even when its underlying sources changed. This checklist helps you establish what was actually recorded before interpreting a trend.

By Eugene Blago · Updated 10 October 2026 · Educational information

The first question is not “what does the score mean?”

Ask which device recorded the value, when it recorded it and whether that same source contributed on the comparison days. Nutrinaut can display available signals from compatible wearables and connected health platforms, but an app cannot recover a measurement that the source never supplied. A missing value is not a normal value and should not be silently filled in as one.

Apple Health, for example, lets people inspect and change the priority of data sources for a health category. A manually entered value, a watch and another connected app may all be present. The visible total can therefore change for a reason unrelated to the body. This is a data-quality check, not a claim that one brand is more accurate than another.

Run this four-part check

Check	What to look for	What to do if it fails
Source	Which device or app supplied sleep, heart-rate or activity data on each day?	Do not compare unlike sources as if nothing changed. Note the switch.
Coverage	Was the device worn, charged and permitted to share the relevant category?	Mark the day incomplete. Do not turn absence into zero sleep or zero activity.
Timing	Were measurements taken at roughly comparable times and in the same time zone?	Record travel, late syncs and unusual timing before interpreting a change.
Context	Did illness, travel, training, alcohol, stress or a manual edit coincide with the change?	Keep that note alongside the number. It does not prove a cause.

A transparent worked example

Illustrative data, not a Nutrinaut user test: a person sees 7.4 hours of sleep on Monday, no sleep value on Tuesday and 6.6 hours on Wednesday. The person changed their data source after Monday and did not wear the new device on Tuesday. The average of the two recorded nights is (7.4 + 6.6) ÷ 2 = 7.0 hours. Calling this a “three-night average” would hide the missing night; calling Tuesday “zero hours” would be plainly wrong.

The useful output is “two of three nights recorded; source changed; mean of recorded nights 7.0 hours.” It is not a verdict that sleep or recovery improved. Consumer-wearable validation studies vary by device, outcome and testing method, so even complete data deserve careful interpretation.

Where Nutrinaut fits

Use Nutrinaut to put available sleep, activity, training and optional lifestyle information in context. If an insight conflicts with how you feel, begin with the source check and missing-data check above. Do not use a wearable summary to diagnose illness, clear an injury or override concerning symptoms. For an individual medical or training decision, use an appropriately qualified professional.

Read the sleep and caffeine guide or open the seven-night sleep planner.

When two readiness scores disagree

Use the Apple Watch and Garmin comparison checklist to compare the original scales, reading times and available records. The downloadable worksheet keeps known differences separate from unanswered questions.

The watch missed a night. What can you still use?

A missing sleep record does not make every other observation useless. It does change which questions the record can answer. Keep a recorded walk as a recorded walk; do not silently turn it into evidence of last night’s sleep or complete recovery.

Our Nutrinaut storage review found different source labels and combinations of available fields. We did not verify the current implementation of every historical mode, so we do not translate internal labels into promises about today’s app. The practical check below uses visible sources and event dates instead. Each scenario is illustrative, not a description of an identified user.

Scenario A: a current sleep record is present

Known: the record gives a sleep interval ending this morning, with a named source. The displayed event date is current, rather than merely the page’s refresh time.

Still unknown: whether the entire night was captured accurately and how the device defined sleep. A complete-looking record is not independent validation of the measurement.

Reasonable comparison: other nights from the same source and definition, with similar coverage. Keep a device or algorithm change visible instead of treating the full series as automatically consistent.

Your check: open the original interval, inspect start and end times, and look for gaps or duplicate sources. When travel is involved, check the time zone and the date assigned to the night.

Scenario B: only a last-known sleep value is visible

Known: a sleep value exists, but its underlying event ended on an earlier date. A dashboard may be current while the observation it displays is not.

Unknown: last night’s sleep. Carrying an earlier value forward does not measure a new night, even if that value looks plausible.

Reasonable comparison: compare that earlier observation with other observations from its actual period. Do not count repeated displays of it as several separate measured nights.

Your check: distinguish event time, import time and the time you viewed the screen. Mark last night as unknown until a suitable record arrives. If a late sync provides it, revise the record with the actual night, not the arrival date.

Scenario C: activity is present but sleep is absent

Known: steps, a workout or another activity observation was recorded. That evidence can support a description of the recorded activity within its own limits.

Unknown: sleep duration for the missing night. Movement does not supply the absent measurement, and a summary score does not restore the missing detail.

Reasonable comparison: activity against comparable activity records. Keep sleep-based comparisons separate and label their missing coverage.

Your check: inspect the sleep source’s records, sharing permissions and sync status. Do not delete valid activity to make a dashboard look consistent, and do not enter zero hours just because sleep was not received.

A short decision path

Can you name the source? If not, mark source unverified and inspect the original health category.
Does the event date match the period you want to discuss? If not, treat it as historical information, not today’s observation.
Is the relevant record present and complete enough for the question? If not, describe only the part you can verify.
Does the comparison use compatible definitions and sources? If not, annotate the change or keep separate summaries.

In Apple Health, the Data Sources & Access controls help identify contributing sources and their priority. Review the entries before changing settings so that you preserve an explanation for a difference. The same principle applies elsewhere, although the exact controls vary by product.

If you record your own recollection of a missed night, label it as self-reported rather than sensor-measured. It can add context without becoming an equivalent replacement for a device observation. Keep that distinction when a weekly average or AI summary is produced.

Before acting on a generated interpretation, use the AI evidence card. The question is not whether a screen can display a number; it is whether the available record supports the particular conclusion you want to draw.

A record you can compare

Make a provenance card before changing settings

“Synced today” answers a different question from “measured today.” Keep three times separate: when the observation happened, when it arrived in the destination, and when you viewed it. A delayed import can be a new arrival of an old observation. A refresh is not a second measurement.

A late-sync example

Invented timeline, not a device test: a sleep interval ends on Monday at 07:00. It reaches a connected health app on Tuesday at 10:00. You open a dashboard on Wednesday at 09:00.

The sleep observation still belongs to the interval ending Monday. Tuesday is the import date, not an additional night; Wednesday is the viewing date. If a weekly summary counts those three dates as three nights, its denominator is wrong even before anyone interprets the sleep duration.

Copy this card for a disputed reading

Question: Which specific trend or difference am I checking?
Value and definition: Number, unit and exact category: for example, sleep duration rather than time in bed.
Original source: Device or app name, plus whether the entry is measured, calculated or self-reported.
Observation interval: Start, end, date and time zone where shown. Write “not shown” instead of guessing.
Arrival and view times: Sync/import time if available, and the time you checked the screen. Keep both separate from the interval.
Coverage and changes: Missing records, duplicate intervals, a source switch, changed permissions or a manual edit.
Comparison decision: Comparable; comparable only with a note; or not comparable yet. State the unresolved detail.

On iPhone, Apple’s documented route is Health → Browse → a category and subcategory → Data Sources & Access. It shows contributing sources for that data type and allows their priority to be changed. Record the existing order before editing it. Do not change several permissions, source priorities and device settings at once if you want to understand which change mattered. Apple’s current instructions.

If the original source and Nutrinaut differ

Compare one category and one interval, not two headline scores. If the original source has no matching record, resolve that gap first. If it has a matching record but the destination does not, keep a source-side and destination-side note with the times above; the public Nutrinaut Connections & Data Guide explains the supported paths and their limits. Do not delete a valid observation or enter a substitute just to make screens agree.

A useful support report can name the app/device, category, relevant dates, time zone and the expected versus displayed behaviour without publishing a personal health history. Share sensitive screenshots only through an appropriate private support channel.

Is this a daily summary or a workout?

A list labelled “activity” can contain two different kinds of record: a summary that accumulates across the day and an individual exercise session. Both can be useful, but their time fields do not necessarily answer the same question. Establish the record type before adding minutes or comparing days.

Daily activity summary

Question: what activity has been accumulated within this reporting window?

The window might run from local midnight to the latest refresh. Its length describes the period covered, not continuous exercise. A cumulative total can already include activity from a separately recorded session.

Individual workout

Question: how long did this particular session last?

Check its start, end and duration definition. Elapsed session time, active time and moving time are different measures; breaks may be included in one and excluded from another. Compare like definitions rather than relying on the word “minutes.”

18 hours of coverage does not mean 18 hours of exercise

Illustrative example, not a user record: on a day without a clock change, a daily summary is refreshed at 18:00 and covers the 1,080 minutes since midnight. A separate workout lasted 45 minutes. The coverage window is not a claim of 1,080 active minutes. Adding the two gives 1,125, but that is not a meaningful workout-duration total.

Keep the 45-minute session as a session. Keep the daily summary as a summary, using the activity measures it actually reports. Do not add their calorie or activity totals either unless you have verified that their definitions are compatible and the same activity is not being counted twice.

Before combining activity records

Name the record type: daily cumulative summary, individual workout or an unresolved category.
Name the time measure: reporting-window length, elapsed session time, active time or moving time.
Check overlap: does the daily total already contain the separately listed session?
Keep unknowns visible: if the source does not explain a field, do not turn it into exercise minutes.

For a session-level check, read Before You Trust a Workout Calorie Estimate, Check the Minutes. For midnight, travel and clock-change questions, use One Night, Two Dates. Resolve the record’s meaning first; only then decide whether it belongs in the comparison.

Sources and scope

Sources support the general method or limitation, not the accuracy of an individual Nutrinaut result. The worked numbers on this page are labelled examples, not user data or clinical advice.

Found a factual or arithmetic error? Send the page URL to [email protected]. See our corrections policy.