What affects step count accuracy more — the device or how I walk?

How you walk matters more than which device you own. At normal walking speed (0.9–1.3 m/s), accuracy differences between Garmin, Apple Watch, and Fitbit are smaller than the error caused by your walking speed, arm swing, and device placement. Research shows walking speed is the single biggest factor — at slow speeds (<0.9 m/s), even the best devices can miss up to 74% of steps.

Why does my wearable miss steps when I push a shopping cart or stroller?

Wrist-worn devices detect arm swing as a proxy for walking — they don't actually count footfalls. When your arms are stationary (pushing a cart, holding handrails, hands in pockets), devices can miss 35–95% of your steps. Ankle-worn trackers perform significantly better in these situations because they detect actual leg movement.

What causes phantom steps on my fitness tracker?

Phantom steps occur when your arms move without walking — cooking, gesturing while talking, driving on rough roads, or desk work. Research shows Oura Ring and Samsung Galaxy Watch have the highest phantom step rates, adding 500–3,500+ false steps per day in some conditions. Garmin's 10-step bout threshold filters phantom steps more effectively than most brands.

Does age affect step count accuracy?

Yes. A 2025 meta-analysis found Apple Watch MAPE was 4.3% for users under 40 versus 10.9% for users 40 and older — a 2.5x accuracy gap. Older adults experience compounding effects: slower gait speed, shorter stride length, and reduced arm swing all reduce accuracy. Ankle placement can help.

Which device placement is most accurate for step counting?

Hip placement is most accurate (~0.4–5% error), followed by ankle (~2–6% error), then wrist (~5–25% error). Ring/finger placement shows the highest error (~10–50%+). The same Fitbit worn at ankle achieved 5.9% error at slow speeds versus 48–75% error when worn on the wrist — a 10x difference from placement alone.

Are wearable step counts accurate for people with Parkinson's or stroke?

Consumer wearables are significantly less reliable for neurological conditions. Research shows stroke patients' step detection rates of 11–30% and Parkinson's patients' rates of 20–47%. Standard algorithms are trained on 'normal' gait patterns and struggle with asymmetric or shuffling gaits. Clinical-grade devices are necessary for accurate tracking.

Why do lab studies show better step count accuracy than real-world use?

Lab conditions (treadmill, controlled speed) produce consistent accelerometer signals with no confounders, yielding ~3–8% MAPE. Real-world use involves mixed activities, variable speeds, and phantom step triggers, resulting in >10–25% MAPE. A study showing 2% accuracy on a treadmill won't translate to your actual workday.

What Actually Affects Your Wearable's Step Count Accuracy — Ranked by Impact

Q: Why does my wearable miss steps when I push a shopping cart or stroller?

Wrist-worn devices detect arm swing as a proxy for walking — they don't actually count footfalls. When your arms are stationary (pushing a cart, holding handrails, hands in pockets), devices can miss 35–95% of your steps. Ankle-worn trackers perform significantly better in these situations because they detect actual leg movement.

Q: What causes phantom steps on my fitness tracker?

Phantom steps occur when your arms move without walking — cooking, gesturing while talking, driving on rough roads, or desk work. Research shows Oura Ring and Samsung Galaxy Watch have the highest phantom step rates, adding 500–3,500+ false steps per day in some conditions. Garmin's 10-step bout threshold filters phantom steps more effectively than most brands.

Q: Does age affect step count accuracy?

Yes. A 2025 meta-analysis found Apple Watch MAPE was 4.3% for users under 40 versus 10.9% for users 40 and older — a 2.5x accuracy gap. Older adults experience compounding effects: slower gait speed, shorter stride length, and reduced arm swing all reduce accuracy. Ankle placement can help.

Q: Which device placement is most accurate for step counting?

Hip placement is most accurate (~0.4–5% error), followed by ankle (~2–6% error), then wrist (~5–25% error). Ring/finger placement shows the highest error (~10–50%+). The same Fitbit worn at ankle achieved 5.9% error at slow speeds versus 48–75% error when worn on the wrist — a 10x difference from placement alone.

Q: Are wearable step counts accurate for people with Parkinson's or stroke?

Consumer wearables are significantly less reliable for neurological conditions. Research shows stroke patients' step detection rates of 11–30% and Parkinson's patients' rates of 20–47%. Standard algorithms are trained on 'normal' gait patterns and struggle with asymmetric or shuffling gaits. Clinical-grade devices are necessary for accurate tracking.

Q: Why do lab studies show better step count accuracy than real-world use?

Lab conditions (treadmill, controlled speed) produce consistent accelerometer signals with no confounders, yielding ~3–8% MAPE. Real-world use involves mixed activities, variable speeds, and phantom step triggers, resulting in >10–25% MAPE. A study showing 2% accuracy on a treadmill won't translate to your actual workday.

Q: Does walking surface affect step count accuracy?

No. Garmin validated step counting across lawn, gravel, asphalt, linoleum, and tile with minimal accuracy differences. Surface type is essentially a non-factor for step counting accuracy.

Ryan - Kygo Health
Feb 22
7 min read

Updated: Mar 22

Last Updated: February 22, 2026

Smartwatch displaying "10,000" steps with a green checkmark, surrounded by fitness icons: clouds, heart, graph, shoe, person on treadmill, and walking dog. Representing effects on step count accuracy.

Your walking speed, arm swing, and where you wear your device matter more than which brand you own. At normal walking speed (0.9–1.3 m/s), the accuracy difference between Garmin, Apple Watch, and Fitbit is smaller than the error introduced by your own behavior and body.

We analyzed peer-reviewed studies from 2020–2026 to rank every factor that affects step count accuracy. All claims below are sourced from published research unless marked with a caution emoji for consumer testing.

What we found: if you're a healthy adult walking at a normal pace, device choice barely matters. But if you're elderly, recovering from surgery, or pushing a stroller daily — your accuracy challenges are different, and fixable.

High Impact Factors — These Matter More Than Which Device You Own

Walking Speed

This is the single biggest factor affecting step count accuracy. Every device struggles at slow speeds — it's not a brand problem, it's a physics problem.

Slow walking produces weaker, less rhythmic accelerometer signals that are harder to distinguish from background noise. The research is clear on the magnitude.

Speed	Typical Accuracy	What This Looks Like
<0.5 m/s	<50%	Shuffling, very elderly gait, post-surgical — most steps missed
0.5–0.9 m/s	50–80%	Slow casual walking, window shopping — significant undercounting
0.9–1.3 m/s	>90%	Normal walking pace — all devices perform acceptably
1.3–1.8 m/s	>95%	Brisk walking — sweet spot for wrist-worn accuracy
>1.8 m/s	>95–99%	Jogging/running — highest cadence = clearest signal

At speeds below 0.9 m/s, even the best devices can miss up to 74% of steps. At normal pace, Garmin, Apple, and Fitbit are all within a few percent of each other.

If you walk slowly and accuracy matters: Ankle-worn trackers dramatically outperform wrist-worn devices at slow speeds.

Sources: Feehan et al. (2020); Choe & Kang (2025); Sensors (2025)

Wear Location

Where the sensor sits on your body changes accuracy more than which sensor you buy.

Placement	Typical Error	Why
Hip	~0.4–5% MAPE	Closest to center of mass; detects trunk movement directly. Research gold standard (ActiGraph, ActivPAL).
Ankle	~2–6% MAPE	Detects actual leg movement. Best option for slow walkers.
Wrist	~5–25% MAPE	Detects arm swing as a proxy for walking. What 95%+ of consumers use.
Finger (ring)	~10–50%+ MAPE	Detects hand movement. Not designed for steps — useful for sleep/HRV.

The gap between placements is significant: wrist-to-hip accuracy difference is approximately 30% in young adults and nearly 50% in elderly subjects.

A Fitbit worn at ankle achieved 5.9% error at 0.4 m/s. The same Fitbit on wrist at the same speed showed 48–75% error. Same algorithm, same hardware — placement alone caused a 10x accuracy difference.

Wrist remains the consumer standard for convenience. Nobody wants to wear a research accelerometer on their hip all day. For healthy adults at normal walking speed, wrist devices are accurate enough.

Sources: Roos et al. (2020); Garmin validity review (2020); Johnston et al. (2021)

Arm Swing

This is the fundamental limitation of all wrist-worn step counting. Your device doesn't count footfalls — it detects repetitive arm motion and infers that you're walking.

When your arms move but you're NOT walking → phantom steps (overcounting)

Activity	Overcounting Magnitude
Animated gestures / talking with hands	+10–15%
Cooking (chopping, stirring, mixing)	+15–25%
Cleaning / scrubbing	+10–20%
Clapping / drumming	+20–35%
Driving on rough roads	+500–3,500 phantom steps/day (Samsung, Oura worst)

When you're walking but your arms are STILL → missed steps (undercounting)

Activity	Undercounting Magnitude
Pushing a shopping cart	−35% to −60%
Pushing a stroller	−40% to −70%
Carrying grocery bags (both hands)	−50% to −80%
Hands in pockets	−35% to −65%
Holding handrails (stairs, treadmill)	−60% to −95%
Using a walker / mobility aid	−70% to −95%

Some devices handle this better than others. Garmin's 10-step bout threshold filters phantom steps more effectively than most brands. In a December 2025 consumer test, Garmin FR970, COROS APEX 4, and Apple Watch Ultra 2 all tracked approximately 5,000 steps accurately from a pocket — suggesting some devices can detect leg motion without wrist swing.

Sources: Android Central (2025) ⚠️; Kristiansson et al. (2023) — Oura phantom step data

Moderate Impact Factors

Age

Your age affects step count accuracy even with the same device, speed, and conditions.

Age Group	Apple Watch MAPE	Source
Under 40	4.3%	Choe & Kang (2025)
40 and older	10.9%	Choe & Kang (2025)

Older adults experience compounding effects: slower gait speed + shorter stride length + reduced arm swing = triple hit to accuracy.

Delobelle et al. (2024) found Fitbit's stepping bout detection dropped off at cadences exceeding 120 steps/min specifically in older adults.

For users over 60 where accuracy matters for clinical tracking, ankle placement helps significantly.

Sources: Choe & Kang (2025); Delobelle et al. (2024)

Gait Pathology

If you have a neurological condition affecting your gait, consumer wearables are significantly less reliable.

Condition	Step Detection Rate
Stroke (hemiparetic gait)	11–30% of steps detected
Parkinson's disease	20–47% of steps detected
Multiple sclerosis	Highly variable

Standard step-counting algorithms are trained on "normal" gait patterns. Asymmetric, shuffling, or irregular gaits produce accelerometer signals that don't match expected templates. This is a fundamental algorithmic limitation across all consumer wearables — not a specific device problem.

Sources: Sensors (2025); Johnston et al. (2021)

Lab vs. Real World

Every device looks better in a study than in your daily life.

Setting	Typical MAPE	Why
Laboratory (treadmill, controlled)	~3–8%	Consistent speed, clear walking signal, no confounders
Free-living (your actual day)	>10–25%	Mixed activities, variable speed, phantom step triggers everywhere

This matters when reading research. A study showing 2% MAPE on a treadmill doesn't mean you'll see 2% accuracy during your workday. Always check whether a study tested free-living accuracy, not just lab conditions.

Sources: O'Driscoll et al. (2024); Giurgiu et al. (2023)

BMI

BMI doesn't directly affect your device's accelerometer. But obesity alters gait biomechanics — wider stance, shorter stride, different arm swing pattern — which indirectly reduces step detection accuracy. The device isn't measuring BMI; it's failing to recognize an atypical gait pattern.

Sources: Scataglini et al. (2025)

Low Impact Factors

Surface Type

Garmin validated step counting across lawn, gravel, asphalt, linoleum, and tile with minimal accuracy differences. Surface type is essentially a non-factor.

Sources: Garmin validity review (2020)

Dominant Hand

No significant accuracy impact from wearing a device on your dominant vs. non-dominant wrist.

Sources: Modave et al. (2017)

Step Count Accuracy: Bias by Condition

Condition	What Happens	How Much	Most Affected
Slow walking (<0.9 m/s)	Underestimates	Up to 74% of steps missed	All wrist/hip devices
Normal walking (0.9–1.3 m/s)	Near-accurate	<5% error	All devices fine
Free-living (mixed day)	Overestimates	+10–35% above actual	Wrist-worn devices
Stationary (desk, driving)	Phantom steps	500–3,500+/day	Oura, Samsung, Polar
Arms still while walking	Underestimates	−35% to −95% missed	All wrist-worn devices

Why Step Count Accuracy Matters for Food-Biometric Patterns

If you're trying to understand how nutrition affects your activity levels, recovery, or energy, the accuracy of your step data matters. When measurement error is high, correlations between food intake and activity patterns become harder to detect reliably.

This is one reason we built Kygo Health to integrate with multiple wearable platforms. Understanding your device's accuracy profile helps you interpret the patterns we surface in your data. A device that systematically undercounts during your daily stroller walks produces different data than one adding phantom steps at your desk.

Download free on iOS or join the fully free Android beta.

For detailed device-by-device accuracy data, see our companion post: Which Wearable Has the Most Accurate Step Count?.

To compare devices visually, use our free Step Count Accuracy Comparison Tool.

Key Takeaways

If you walk at a normal pace and swing your arms normally, any major brand device (Garmin, Apple, Fitbit) is accurate enough for daily tracking. Device choice barely matters.
If you're slow, elderly, or push a cart/stroller daily, your step counts are likely significantly undercounted regardless of device. Ankle placement is the best fix.
If you get phantom steps at your desk, Garmin's 10-step bout threshold filters these best. Oura Ring and Samsung Galaxy Watch show the highest phantom step rates in testing.
If you have a neurological gait condition, consumer wearables may miss 50–90% of your steps. Clinical-grade devices are necessary.
Don't compare your step count to someone else's. Their gait, speed, arm swing, age, and device placement create a completely different accuracy profile.

Ready to correlate your activity data with nutrition patterns? Kygo Health integrates with Oura, Apple Health, Garmin, and Fitbit — use whichever device fits your accuracy needs and see how food choices connect to your biometric data.

Download free on iOS or Android — fully free to join.

Disclaimer: Kygo Health LLC is a personal data aggregation and insights platform designed for informational purposes only. The information provided does not constitute medical advice, diagnosis, or treatment. Always consult a licensed healthcare provider with any questions regarding medical conditions.

Have sources or data that should be included here? Reach out at Ryan@kygo.app.

Sources

Choe S & Kang M (2025). Physiological Measurement. DOI: 10.1088/1361-6579/adca82 — 56 studies, 270 effect sizes
Feehan LM, et al. (2020). PeerJ. DOI: 10.7717/peerj.9381
Roos L, et al. (2020). Int J Environ Res Public Health, 17(20), 7123. DOI: 10.3390/ijerph17207123
Garmin Validity Review (2020). PMC. DOI: 10.3390/ijerph17134269
Johnston W, et al. (2021). Br J Sports Med, 55(14), 780-793.
O'Driscoll R, et al. (2024). Sports Medicine. DOI: 10.1007/s40279-024-02077-2
Giurgiu M, et al. (2023). Technologies, 11(1), 29. DOI: 10.3390/technologies11010029
Kristiansson E, et al. (2023). BMC Medical Research Methodology, 23, 50. DOI: 10.1186/s12874-023-01868-x
Delobelle J, et al. (2024). Digital Health, 10, 20552076241262710. DOI: 10.1177/20552076241262710
Scataglini S, et al. (2025). Int J Obes, 49(4), 541-553. DOI: 10.1038/s41366-024-01659-4
Sensors (2025). Sensors, 25(18), 5657 — Step counting in neurological conditions
Modave F, et al. (2017). JMIR mHealth, 5(6), e88. DOI: 10.2196/mhealth.7870
⚠️ Android Central (December 2025). 10-watch step test — pocket tracking data