Troubleshooting confusing stats

When charts disagree with intuition, verify definitions before assuming a bug.

Start with definitions

Before assuming a bug, verify that you understand what each metric measures. The most common source of stat confusion is expecting one metric to behave like another:

  • Words read uses forward-first accounting — re-reading earlier sentences may not increase it. See Words read and study credit.
  • Reading time excludes idle periods — leaving the reader open without interaction does not accumulate time. See Reading time and idle pauses.
  • Quiz accuracy reflects attempts in the selected time range only. A high all-time accuracy may coexist with a low recent accuracy if you hit harder material today.
  • Vocabulary counts show per-language totals. Switching languages changes the vocabulary picture entirely.

Time zones

Daily stats buckets depend on your study timezone setting. The day boundary follows your configured timezone, not UTC, and the app uses a 3:00 local-time boundary for study days. If a session that should appear on one day shows up on a neighboring day, check:

  • Your study timezone in Settings → Goals
  • Whether the session crossed midnight in your timezone
  • Whether your device clock is accurate

If midnight boundaries look odd even after verifying your timezone, capture the date range and expected vs actual bucket when reporting.

Delayed or missing updates

Reading stats are flushed while you move forward through a story and again when the visit ends. Quiz stats appear after a quiz is saved, and familiarity totals update after a status change is saved. If stats seem stale:

  • Advance into a new sentence to generate reading credit. Tapping a word by itself resets the idle timer but does not add words-read credit.
  • Wait briefly and refresh the dashboard. Reading events are batched during a visit.

Report issues

If something still looks inconsistent after verifying definitions and timezone:

  1. Note the specific chart and date range where the values seem wrong
  2. Capture what you expected to see and what is actually shown
  3. Mention your study language and timezone setting

Support can investigate the underlying data with this context. Without it, a report like "my words read is wrong" is hard to act on.