If your daily total looks wrong, the useful question isn't "is the watch broken?" It's "which source produced these steps, and what was I doing at the time?" Fitness watch step count accuracy depends on both. A watch can behave well on a steady walk and still log odd numbers during slow shuffling, a gym session, or a drive. This guide gives you a short routine to identify the source, run a controlled walking check, log the non-walking activity around it, and save a record that support, or you later, can reproduce.
Confirm where the step number comes from
Before you test anything, write down what you're looking at. Step counts are easy to misjudge if you compare numbers that were never meant to match.
- Which device and model. Note the exact watch or band, the wrist, and how it was worn.
- Which screen. A watch face, the companion app, and any aggregating health app can each show a "daily steps" figure.
- Which interval. Compare the same clock window, not "today so far" on one screen against "yesterday" on another.
- Whether data has synced. An unsynced watch and an updated app will disagree for ordinary reasons.
If an aggregating app shows a larger total than your watch, don't conclude it's double-counting. It can legitimately exceed either device's total when it merges non-overlapping time intervals, for example watch steps from the morning and phone steps from the afternoon when you left the watch at home. If the app offers an interval-level or per-source breakdown, look at which source supplied each block. Overlapping intervals from two sources are a lead, not proof of duplication: the app may apply a priority rule that keeps only one source for those minutes.
Menu paths and settings differ by model and app version, so none are listed here. If your manufacturer documents step counting for your exact model, check that current documentation before changing settings.
Choose a short walking reference
A reference is only useful if it's more trustworthy than the thing you're checking. The simplest one is a manual count of a short walk.
A 2020 peer-reviewed study in PeerJ (criterion validity of a wrist-worn consumer tracker and an ankle-worn research monitor at various walking speeds) shows a design worth borrowing. Thirty participants walked on a treadmill at six speeds from 1.6 to 5.6 km/h while being video recorded, and steps were counted manually as the reference. The findings are specific to that setup: one wrist-worn consumer band, controlled treadmill walking. In that study:
- The median difference from the manual count ranged from −49.5 steps at 1.6 km/h to −1 step at 4.0 km/h, so the band tended to undercount.
- Mean absolute percentage error stayed under the authors' 3% threshold at 3.2–4.8 km/h (0.61–1.27%).
- Errors were larger at the slowest speed than in the middle of the range.
Don't carry those numbers over to another device or treat them as a general error rate. What transfers is the lesson: walking speed matters, so a single test at one pace tells you little. A broader systematic review of commercial wearables covering literature up to May 2019 is a reasonable place to read about validity and reliability across devices and populations.
A practical home version:
- Pick a short, flat route or a set treadmill speed, and decide the pace beforehand.
- Count your steps by hand, or have someone count or film the walk. Aim for a modest number, such as 100 steps, so counting stays reliable.
- Note the watch's step total immediately before and after, viewed on the same screen both times.
- Repeat at a slower and a more typical pace, because speed changed the result in the study above.
Judge the result by the pattern. One mismatch could be a counting slip. A repeated gap at the same pace, with the same wrist position, is a signal worth recording.
Record other activity that interferes
Steps are inferred from motion sensing, so arm movement that isn't walking is a plausible source of confusion. The validation study above doesn't test these situations, so treat anything below as a hypothesis to check, not a verdict.
Log, with times (for a vehicle, read the count only when safely stopped, or as a passenger):
- Activities where your arms move but your feet don't, such as cooking, driving, or typing.
- Activities where your feet move without normal arm swing, such as pushing a cart or carrying something.
- Gym or home workouts, with start and end times, since stepping on a treadmill, rowing, or lifting all load the wrist differently.
- Any time you wore the device on the other wrist or took it off.
Illustrative example (invented numbers, not a test): Your watch shows 6,200 steps by 6 p.m. An aggregating app shows 7,000. Its source list shows 800 phone steps between 12:00 and 12:45, when the watch was charging. The two sources cover different time blocks, so the higher total is expected. If both sources reported steps across the same 12:00–12:45 block, you'd have an overlap to investigate, including how the app treats it.
Save a reproducible issue
A good report lets someone else see the same thing. Keep one note per incident:
| Item | What to write |
|---|---|
| Device and wear | Model, wrist, band position |
| Source screen | Watch face, companion app, or aggregating app |
| Interval | Exact start and end times |
| Reference | Manual count, number of steps, pace or treadmill speed |
| Watch reading | Before and after totals, sync status |
| Context | Non-walking activity in or near the window |
| Repeat | Whether it happened again under the same conditions |
Accept the check as resolved when your repeats agree with the reference across paces, or when a difference lines up with a logged non-walking interval or a source gap. If a repeatable gap remains in a clean walking test, attach the table to your support request, along with screenshots of any interval-level source data, so support doesn't have to guess which number you meant.
Whatever device you use, the same routine applies: name the source, fix the interval, test against a manual count, and log what else you were doing.
