DATA QUALITY
A better efficiency score or better sensor coverage?
An efficiency value can change because the recording changed. Compare the amount and location of measured data before comparing the score.
Ask what the average includes
A summary heart rate may be an average of the samples that exist, rather than a measurement covering every second of the activity. Check the chart for gaps and ask which portion the average represents. A missing value is not the same as a zero-bpm sample.
The placement of a gap matters. Missing data spread across a steady outing poses a different interpretation problem from losing the entire faster finish. Coverage percentage alone does not tell you whether the remaining data represent the full activity.
An illustrative missing-half example
Assume a 40-minute run at a constant 3 m/s. Its first 20 minutes average 140 bpm and its last 20 average 160 bpm. With complete, equally weighted data, the overall average is 150 bpm.
| Available HR samples | Recorded mean | HR cost using whole-run speed |
|---|---|---|
| Both halves | 150 bpm | 50.00 bpm/(m/s) |
| Only the first half | 140 bpm | 46.67 bpm/(m/s) |
| Only the second half | 160 bpm | 53.33 bpm/(m/s) |
The missing-first-half and missing-second-half versions describe neither a changed route nor changed running. They show what happens when a whole-run speed is paired with incomplete heart-rate coverage. The lower value is not evidence of a better effort.
Distinguish absent, measured and reconstructed data
An empty interval, a sensor reading and a model-generated value are three different things. Keep those labels visible. Filling a gap can help a display, but it does not turn the filled values into observations.
Likewise, more samples do not automatically mean more accurate samples. Inspect outliers and the source as well as duration. If a recording changes devices halfway through, note that boundary when comparing its early and late sections.
Use a narrower conclusion when needed
Runome's detailed metrics depend on the available activity data. If the heart-rate stream is incomplete, you can still compare distance and time on an appropriate basis. You may also study a clearly identified shared segment, while leaving the whole-run physiological interpretation unresolved.
Read the missing-metric troubleshooting guide to inspect the import path. The heart-rate cost calculation shows exactly where the mean enters the ratio. Keep a note of recording limitations so a future change in coverage does not masquerade as a training trend.