Keyboard cadence as a fatigue signal
Typing speed is a weak signal in isolation and a strong one across a session. The drop from minute 5 to minute 35 is what tells you something, not the absolute number.
- Absolute typing speed is not a fatigue signal; the drop within a session is.
- Error rate moves with strain more reliably than WPM does.
- The clearest patterns show up across early, middle, and late thirds of long sessions.
Forget how fast you type
Productivity tools that read keyboard activity tend to stop at WPM and treat the number as a verdict on focus. That is the wrong unit of analysis. People type at very different speeds, and even the same person varies wildly by task — a code review block looks nothing like a writing block looks nothing like a Slack thread.
The useful question is not how fast you type. It is *how fast you type relative to the start of your own session.* That is where the signal lives.
WPM as a session-shape signal
Sarenica records WPM per minute and rolls it up into early, middle, and late thirds of every session. The interesting number is the slope, not the level.
A clean session looks roughly flat across the three thirds, or even climbs slightly. A fatigue-prone session shows a clear decline by the last third. That is the shape the weekly report keys off when it claims "30-45 minute sessions look more demanding than 15-30 minute ones." It is reading the slope; it just does not show it to you that way.
KPM catches what WPM misses
KPM (keystrokes per minute) sees what WPM cannot: code, shortcuts, navigation, anything that does not produce words. It moves more with task type than with strain, but the *burst pattern* across a session is a useful tell.
Irregular KPM bursts followed by long quiet stretches usually means task-switching — which itself correlates with fragmented focus, even when the typing speed in each burst looks fine.
Error rate, the cleanest signal of all
Error rate is the percentage of keystrokes that result in a backspace or correction within a short window. It does not care about your normal speed, your task, or your typing style. It rises with strain. And once it rises, it rarely drops back down within the same session.
That is what makes it the most reliable keyboard signal. When error rate, eye-closed seconds, and head-pose drift all climb together in the late third of a session, the weekly report has its strongest case for "shorter blocks here." Three independent signals agreeing on direction. That is what *probable* looks like.
- WPM drift → session-shape problems.
- KPM burst patterns → fragmented work.
- Error-rate climb → cleanest in-session strain signal.
Where the weekly report uses it
Keyboard signals do not appear as their own section in the weekly report. They feed into productive density and focus drift upstream. That is why a report can call out "your 30-45 minute blocks lose productivity" without ever mentioning typing — the typing pattern is one of the inputs that finding rests on.
If you want to see the underlying pattern, the in-app session view breaks out the early/mid/late thirds for any saved session. That is the right place to verify a weekly finding feels real.
