Skip to main content
Log inGet started

What we rank people on

This list is an allow-list. A metric is not something members are ranked on unless it appears below, so anything this product measures in future starts out non-competable and has to be added here deliberately.

Registry competable-metrics-2026-09-04

Competable

Every one of these is a count or a movement your own history produced, measured against your own starting point. None of them is a reading taken off your body.

  • streak

    days in a row the member kept their streak going

    Used on: leaderboard, challenge

  • most_improved

    normalised points gained across the period, so a low starting point is no penalty

    Used on: leaderboard

  • improvement

    movement across the window against the member own starting point

    Used on: challenge

  • consistency

    days inside the window the member showed up for

    Used on: challenge

  • pct_goal

    percent of the member own goal reached each day

    Used on: challenge

  • band_days

    days spent in the member own target band

    Used on: challenge

  • pct_personal_goal

    percent of each side own readiness goal, so two sides compete against themselves

    Used on: head_to_head

  • mean_score

    average normalised readiness score across the named window

    Used on: segment

  • check_in_days

    days checked in across the named window

    Used on: segment

Never competable

Named here rather than left off the list above, because an absence reads as an oversight and this one is a decision.

  • reserve

    The slow metric reads an accumulated pattern, and the study behind it separates a state produced by taking in too little energy from a state produced by sustained demand. A wearable cannot tell those two apart, so the number can move in the flattering direction for a member who is eating too little and resting. A ranking pays for whatever it ranks, and an employer-paid product must never pay for that. It stays off every board, and a derived or aggregated version of it is the same metric under another name.

    Also kept out of: sharing with an audience, any peer-facing payload, comparison against another member

  • training_load

    A load figure is not a measurement in any shared unit. It is a position on a scale its own vendor invented, capped where that vendor chose and curved how that vendor chose, so two members on two devices hold numbers that are not the same quantity and cannot be ordered against each other. A board built on it would not rank effort, it would rank whose hardware scores generously. Normalising it away does not fix that: the normalisation would be this product inventing the comparison the providers refused to make. It stays off every board, and a derived or aggregated version of it is the same metric under another name.

    Also kept out of: sharing with an audience, any peer-facing payload, comparison against another member

  • strain

    Strain is one vendor's name for their own load scale, and every argument against ranking a load figure applies to it unchanged: the number is a position on a scale that vendor invented rather than an amount of anything, so ordering two members by it ranks their hardware. It is registered under its vendor name as well as the canonical one because a key arriving from that provider carries their word for it and would otherwise be graded against nothing.

    Also kept out of: sharing with an audience, any peer-facing payload, comparison against another member

The list is enforced where a board is built, not where it is drawn. A request that names an excluded metric is refused by the API with the reason above, whoever sends it and whatever screen it came from.