A step counter detects the repeating acceleration pattern of walking using a three-axis accelerometer, then software decides which peaks count as steps.
Your tracker’s number is an inference, not a measurement. The sensor records raw motion and an algorithm decides what a step looks like — which is why two devices on the same walk so often disagree. Once you know what the hardware is actually doing, the quirks stop looking like malfunctions.
What Sensor Counts Your Steps?
A three-axis accelerometer does the counting. It measures acceleration along three directions and logs continuous changes in linear motion.
Fitbit’s documentation states that its devices use a three-axis accelerometer to count steps and also to derive movement frequency, duration, intensity, and patterns. A walking step produces a repeating acceleration signature; an isolated motion usually gets rejected or handled differently.
The software side is a signal filter plus a decision rule. The algorithm weighs repetition, timing, intensity, and movement pattern to judge whether a burst of movement is a step. Those detection rules are almost always proprietary, so published research describes consumer devices as inferring steps from tri-axial accelerometer data rather than measuring them directly.
Some devices add a gyroscope, GPS, barometer, or optical sensor. GPS is better at distance and route than at counting individual strides.
Pendulum Pedometers vs. Sensor-Based Counters
A traditional mechanical pedometer counts with a pendulum or weighted arm that swings on vertical movement. Electronic pedometers and phone apps replaced that hardware with motion sensors and software.
That difference explains a lot of the drift you see between devices. A pendulum responds to a physical bounce; an algorithm interprets a pattern. The second approach is far more flexible — and far more opinionated about what counts.
If you are comparing dedicated wrist units before buying, this breakdown of tested wrist step counters covers how each one handles everyday walking.
Why Phones and Watches Disagree
Each device runs its own algorithm on its own sensor placement, so totals rarely match.
- Placement matters most. Wrist-worn devices are sensitive to arm motion, and research reports hip-based devices can be more accurate than wrist-worn ones. Smartphone accuracy varies with where you carry the phone and what you are doing.
- A left-behind phone records nothing. Movement made while wearing only a watch is only counted if the watch stores it independently and syncs later.
- Settings get in the way. The operating system, app, model, sensor availability, permissions, and battery-saving settings all affect whether steps register or transfer.
- Sync can silently stall. Restrictive background settings may delay or block data from reaching your account.
Pick one designated source for your daily numbers and compare against that, not across devices.
| Condition | Effect on Count | Practical Fix |
|---|---|---|
| Slow or shuffling gait | Undercounts; motion too weak to classify | Accept the gap on slow days |
| Pushing a stroller or cart | Undercounts; arms stay still | Carry the phone in a pocket |
| Holding a handrail | Undercounts; arm anchored | Wear the device on the hip if supported |
| Vigorous arm movement | Overcounts; false step signature | Keep wrist motion natural |
| Driving or riding | Overcounts on rough roads | Review and trim obvious spikes |
| Phone in a bag or on a desk | No steps recorded | Keep it on your body |
| Wrist vs. hip placement | Wrist generally less accurate | Stick to one position every day |
How Accurate Are Step Counters, Really?
Accuracy collapses when you leave the treadmill. In controlled walking tests, devices look excellent; in free-living conditions they drift considerably.
An expert statement in the British Journal of Sports Medicine reports mean absolute percentage errors of roughly −0.7% to 5.2% during controlled treadmill walking, against 17.5% to 22.8% during free-living conditions for one wrist-worn tracker — a jump worth knowing about.
A 2026 study comparing consumer devices with research-grade thigh accelerometers found systematic underestimation: consumer devices averaged 10,414 steps against 12,608 reference steps. Smartwatches underestimated less than smartphones in that comparison.
Earlier controlled research found smartphone-app differences ranging from −6.7% to 6.2% versus direct observation, while tested wearable devices ranged from −22.7% to −1.5%. Age and walking speed matter too: one study found some devices undercounted in older age groups.
Treat the daily total as an estimate for tracking trends, not a medical measurement. Step counts do not replace clinical-grade activity monitoring or medical advice.
Sources
- Fitbit / Google Health. “How Fitbit tracks your activity and sleep” Documents the three-axis accelerometer as the step-counting sensor and the derived movement metrics.
- British Journal of Sports Medicine. “Accuracy of wrist-worn wearables during controlled and free-living activity” Supports the controlled-versus-free-living error ranges and the wrist-versus-hip accuracy comparison.
- JAMA. “Accuracy of smartphone applications and wearable devices for tracking physical activity data” Source for the smartphone-app and wearable-device error ranges versus direct observation.