A week with two blank days
Imagine a device records 6,000 steps on each of five days, then has two blank days. Dividing the recorded total of 30,000 by seven gives about 4,286 steps per calendar day. Dividing by five gives 6,000 per recorded day.
Neither calculation recovers what happened on the blank days. They summarize different quantities. A label that simply says “daily average” hides that choice.
| Calculation | Result | What it actually summarizes |
|---|---|---|
| 30,000 divided by 7 | About 4,286 | Recorded steps per calendar day |
| 30,000 divided by 5 | 6,000 | Steps per recorded day |
The reason for missingness changes the interpretation
If someone removes a watch for swimming, missing periods may contain activity the device cannot see. If they wear it mainly on days they feel well, recorded days may not represent the whole week. A charging schedule can create a recurring blind spot without any change in behavior.
These examples do not identify the direction of bias for every device. They show why the absence of a number needs an explanation.
A useful comparison needs coverage too
When comparing weeks, it helps to know recorded days, wear time, and the rule for a valid day. A rise from five to seven recorded days can change a total even when the person's routine does not change.
The FDA's digital-health measurement guidance treats missing data and usability as part of study planning. The same question is useful for a personal dashboard: was the change in the person, the recording coverage, or both? If coverage is unknown, a small difference may not support a confident interpretation.
