Precision Is Not Accuracy

A tight group of shots far from the bullseye is precise and inaccurate. Most workforce measurement looks exactly like that. For a related reference, see further details.

Hours are counted to the minute. Badge swipes are timestamped. Activity scores update continuously. And the things that predict whether somebody stays and does good work are measured rarely or not at all.

What gets measured precisely

Presence. Days in the office, hours logged, times of arrival.

Activity. Keystrokes, application time, idle detection, message counts.

Output volume. Tickets closed, calls made, lines shipped.

All three are easy to capture, produce clean numbers, and update without anybody deciding anything. That is why they exist, not because anybody established they predict something.

What predicts the outcome

Manager quality. Perceptyx research puts the increase in turnover risk from poor management at four times, and reports that 24% of employees say they work for the worst manager they have had — a figure unchanged since 2023. (Vendor research; the direction is widely corroborated, the exact multiple is theirs.)

Whether people can speak up. Psychological safety in the specific sense, not general morale.

Whether the work has visible trajectory, which differs by career stage.

None of these appears on a productivity dashboard, and all three are measurable — just not automatically.

The evidence on presence specifically

33%lower turnover under hybrid working, with no productivity dropStanford study published in Nature — peer-reviewed

Gallup reports that among remote-capable US employees, 53% work hybrid, 27% fully remote and 20% fully on-site.

And strict return-to-office mandates are associated with roughly 14% higher attrition, concentrated among senior staff. (Industry analysis; treat the figure as indicative.)

Presence is measured most precisely and predicts least. That is the whole argument of this section in three numbers.

Why the pattern persists

Because automatic beats informative. A number that arrives without effort will be used; one that requires a conversation will not.

Because it is defensible. "Attendance was down" is a fact somebody can point at. "The manager is not good" is a judgement somebody has to make and own.

And because the tooling supplies it. Software measures what is easy to instrument, and the availability of a metric becomes the reason for it.

What to do instead

Keep the precise numbers and stop treating them as the answer. Attendance data is useful for scheduling and useless for performance.

Add the slow measurements. Turnover by manager, exit reasons recorded consistently, whether people raise problems before they become resignations.

And check the correlation yourself. Your own turnover data will tell you which of your metrics moves with people leaving, and it is frequently none of the ones on the dashboard.

The short version

For additional context on this topic, see Fstoppers.