If you are trying to work out why your step count data sources disagree, start with a different question: which source wrote each number? A health app that gathers steps from a phone, a watch, and other connected apps is an aggregator. Its daily total can match neither the phone nor the watch alone, and that does not by itself mean anything is broken. Before judging whether a count is right, find out which devices and apps wrote data, and for which stretches of the day.
This article gives you a way to trace the chain from device to summary, a same-interval comparison you can repeat, and a checklist for spotting real duplicates or gaps. Menu names differ between phones and apps, so treat the steps as a map and confirm exact wording in your app's own help pages. Where a feature is not present in your app, skip that step.
The path a step takes to the daily total
Think of each step number as passing through these stages:
| Stage | What to identify | Where to look |
|---|---|---|
| 1. Origin | Phone, watch, band, or a third-party app | The steps view's list of sources or devices, if your app has one |
| 2. Permission | Whether that source may write steps to the health app | The health app's access settings and the source app's own permission screen |
| 3. Priority | Where the source sits when several provide the same data type | A source-ordering screen, if your app offers one |
| 4. Entry type | Sensor data or a manual entry | The list of individual records behind the total |
| 5. Display | Daily total versus the records beneath it | Summary card versus the detailed data list |
A daily total is a stage-5 summary. Questions about accuracy belong at stage 1 and in the entries beneath the total. Some apps rank sources and use the top-ranked one when two provide the same data; others combine or show them separately. Check your app's documentation for which applies, because this article cannot tell you how your specific app resolves overlapping records.
Check sources, permissions, and priority
Work through these in order:
- List every source that writes steps. Note the phone, any watch or band, and every app or accessory you have connected over time.
- Check what each source is allowed to do. An app installed months ago may still be writing data, and a source you expect may have no permission at all.
- Note the priority order, if your app has one. A recently connected source may be placed above an older one. Read the order and understand it first; do not use it as a lever to make a total look right.
- Look for manual entries. A suspicious block of steps at a round time may have been typed in rather than sensed.
- Check that the device was worn and carried. A watch left on a desk contributes nothing, and a phone left in a bag may not travel with you.
Why the daily total can exceed both device totals
Many people see a combined figure larger than either device's own number and assume steps are counted twice. That inference is unsafe from the total alone.
Suppose the phone was in your pocket from 7:00 to 12:00, and the watch was on your wrist from 12:00 to 18:00 while the phone sat on a charger. Each device saw only part of the day. A merged view covering both intervals legitimately beats either single-device total, because the intervals do not overlap. Likewise, if you read a total on the watch in one place and in the health app in another, they may be built from different source sets.
The reverse needs care too. Two sources with stored entries for the same minutes are not proof that both were counted; the app may prefer one source. What settles it is interval-level provenance: which source wrote each entry, what start and end times it covers, and where the sources rank. Sources that alternate in time suggest normal coverage. Entries from two sources over the same minutes are a cue to check how your app handles priority, not a verdict.
Same-interval comparison (illustrative example)
This is a teaching example with made-up numbers, not a recorded test.
Input. You pick a Saturday morning from 9:00 to 10:00. You wore the watch and carried the phone. In the detailed steps data for the day, you read the entries inside that hour by source.
| Source | Entries in 9:00–10:00 | Steps |
|---|---|---|
| Phone | 3 | 1,850 |
| Watch | 3 | 1,910 |
Judgment. Both sources cover the same hour, so this is a like-for-like comparison. The 60-step gap is about 3% of the larger figure. Ask whether it has an obvious cause: did the phone sit on a counter for ten minutes while you walked around with the watch? Was your arm occupied carrying groceries? Those are hypotheses, and the log alone cannot confirm them.
Output. Record the hour, both counts, the difference, and your hypothesis. Then check where each source ranks and whether the daily total seems to reflect one source for this hour or something else. Repeat on a second day. A single hour tells you little; a pattern across several matched hours tells you more.
Matching numbers show that two devices agree with each other, not that either is correct. For that you need a reference count, such as counting steps by hand over a short walk.
What published validation does and doesn't tell you
App documentation describes how data is collected and ordered; it is not an independent accuracy test. One example of a reference-based design is a 2018 treadmill validation study (also at Springer Nature Link). It tested one wrist-worn device against steps manually counted from video, across several walking speeds, in adults on a treadmill.
- It is a result for the specific device tested at that time. Do not transfer it to a different or newer device, or treat it as a benchmark; this article reports no error figures from it.
- It was a treadmill protocol, not a full day of mixed activity, stroller pushing, or varied phone placement.
- It says nothing about how a phone's own step counter performs, or about which record reached your summary on a given day.
- Agreement between devices is not proof of truth. Only a reference count speaks to correctness.
A systematic review of commercial wearables covering steps, energy expenditure, and heart rate offers broader context on how validity varies by measure and setting. Use it to temper confidence, not to rank devices.
Duplicate and missing-data checklist
Keep a small log for each day you investigate:
| Clue | What to record | What it suggests |
|---|---|---|
| Two sources, same minutes | Source names, start and end times, counts | Possible overlap; check priority behavior before concluding anything |
| Sources alternate in time | Hand-off times | Normal non-overlapping coverage; a higher total is plausible |
| Long gap with no entries | Start and end of gap | Device not worn or carried, permission off, or source not syncing |
| Round-number block | Time, count, source label | Possibly a manual entry |
| Unfamiliar source | App or device name | Check its permission and ranking |
| Recently added source | Date added | May have changed which source is preferred |
Acceptance criteria. You can call the chain verified when: (1) every source writing steps is listed and recognized; (2) each has the permission you intend; (3) for at least two matched intervals, you can explain differences by coverage or placement; (4) you have found no unexplained manual entries; and (5) the daily total is consistent with the non-overlapping intervals from your sources, or any overlap is explained by priority.
If something still looks wrong, keep the existing entries. Do not delete records or remove a source just to lower a total. Review the source list and ordering as your app's documentation describes, note any change you make, and compare the next day's matched intervals.
