Notes / 18 January 2026

Weekday versus weekend usage baselines

A weekly average session length is a convenience for people who do not want two charts. It is also how a Sunday football rights spike becomes “the product is growing”. In the Usage Cohort Atlas we split Monday–Friday from Saturday–Sunday before we compute P50 or P75. Bank holidays in England, Scotland, and Northern Ireland are annotated; we do not pretend the United Kingdom shares a single leisure calendar.

Habit products (parking, banking, travel status) usually show a weekday ridge and a weekend dip. Binge products (catch-up video, long-form audio) often invert that. Mixing them in a peer set, then quoting one weekly mean, is how “category engagement” becomes a number nobody could reconstruct.

The split is not moral. It is mechanical. If your peer set is four catch-up apps, weekend P75 is the interesting band. If your peer set is utilities, weekday P50 is the band. The atlas one-pager states which band you are using and why the other is in an appendix.

Teams sometimes want a “seasonally adjusted week”. We rarely allow it in week six. Adjustment models hide the rights weekend you should be explaining in a sentence. Explain it. Then, if you still need a smoother series for an operations dashboard, keep that series off the board pack.