The advice in our device guide was to wear a device for a week and compare it with what you already know. Here is how to do that properly, using the same ideas the validation studies use, scaled down to a kitchen table. It will not measure accuracy. It will tell you whether a device is worth believing for you.
What a week can and cannot tell you
The studies that establish accuracy compare a device with a reference. For sleep, that reference is polysomnography, the laboratory recording.2,3 Your own diary is not that. It is a record of what you remember and noticed. So the test below checks whether a device is plausible and consistent with your experience. It does not check whether it is correct. Keep those two questions separate.
Why not just check the correlation?
It is tempting to ask whether the two sets of numbers “go up and down together.” A classic 1986 paper on comparing measurement methods argued that correlation is the wrong tool for this and is misleading, and proposed a simple graphical alternative instead.1 The intuition: two things can rise and fall together while one of them is always an hour off. What you want to know is how far apart they are, and whether that gap is steady or wanders.
The week
The published framework for testing sleep trackers includes a discrepancy analysis, Bland-Altman plots and an epoch-by-epoch comparison.2 A home version of each:
- Before you look at the device each morning, write down when you think you fell asleep, when you woke, and any time you were awake in the night for more than a few minutes. Writing first keeps the device from colouring your account.
- Then record what the device says for the same things.
- Make a difference column. Device minus diary, in minutes, for each night. That is the discrepancy analysis.
- Look at the pattern, not the average. If the device is always twenty minutes late to say you fell asleep, that is a steady offset you can learn to read. If the gap swings from five minutes to ninety, no single correction will rescue it. That is the idea behind graphing the disagreement instead of correlating the numbers: the size and spread of the gap is the result.1
- Check the minutes, not just the totals. The epoch-by-epoch idea is to compare moment by moment instead of night by night.2 At home, that means one question: on the night you remember lying awake from 3 to 4, does the device show it?
The night worth waiting for
In a laboratory comparison of seven consumer sleep trackers with polysomnography across three nights, in 34 healthy young adults, the devices were very good at detecting sleep (sensitivity at least 0.93 for every device) and much weaker at detecting wake (specificity between 0.18 and 0.54). The authors also found the devices generally performed worse on nights with poorer or disrupted sleep, and that sleep-stage readings were inconsistent.3
Translate that: a device may be better at confirming you were asleep than at noticing you were awake, and a bad night is exactly where it is most likely to slip. So the most informative night in your week is a restless one you remember clearly. If you can, do not only test the nights that went well.
Add a second device, if you have one
Wearing two at once, as the earlier guide suggested, gives a second comparison. The same caution applies: agreement between two devices is not proof that either is right. Disagreement, however, is useful. In a 2026 comparison of ten wearables against an ECG reference, heart-rate accuracy was strongly device-specific, with mean absolute errors from 4.5 bpm for the best to 9–14 bpm for several others.4 Different devices genuinely do differ, so a gap between yours is information.
For heart rate, compare resting readings. Error during activity ran about 30% higher than at rest in one large study, so a mismatch during a hard set tells you less.5
Reading your result
- Steady, small gaps: the device is consistent with you. Trust its trends.
- Steady, large gaps: usable for trends, not for absolute times. Know the offset.
- Wandering gaps: treat the nightly numbers as rough. Look at weekly patterns only.
- It misses the nights you were awake: expected from the research, and worth remembering whenever a score looks flattering.
The takeaway
- Write your own account first, then read the device. Do it for seven nights.
- Look at the size and steadiness of the gap, not whether the numbers rise and fall together.
- Make sure one of the nights is a disrupted one; that is where devices are most likely to slip.
- Compare resting heart rate, not exercise heart rate.
- A diary is not a gold standard. This tests plausibility, not accuracy.
Sources
The studies and standards behind the claims above, so you can read them yourself. Listing a source is not an endorsement of any device it tested.
- Bland JM, Altman DG (1986). The Lancet 1(8476), 307–310, doi 10.1016/S0140-6736(86)90837-8 — statistical methods for assessing agreement between two methods of clinical measurement.
- Menghini L, Cellini N, Goldstone A, Baker FC, de Zambotti M (2021). Sleep 44(2), zsaa170, doi 10.1093/sleep/zsaa170 — a standardized framework for testing sleep-tracker performance.
- Chinoy ED, Cuellar JA, Huwa KE, et al. (2021). Sleep 44(5), zsaa291, doi 10.1093/sleep/zsaa291 — seven consumer sleep trackers vs polysomnography, n=34, three nights.
- Gielen J, Van Oost CN, Debard G, et al. (2026). JMIR Formative Research, e85186, doi 10.2196/85186 — ten commercial wearables vs ECG-derived heart rate, n=45.
- Bent B, Goldstein BA, Kibbe WA, Dunn JP (2020). NPJ Digital Medicine 3, 18, doi 10.1038/s41746-020-0226-6 — six wearables vs ECG patch, n=53.
This article is general wellness education, not medical advice. SoluneFit is a wellness tool, not a medical device. It does not diagnose or treat any condition. Talk to a clinician about health concerns.