What Predicts Retention

If the dashboard metrics do not predict who leaves, something must. These six do, and none arrives without somebody deciding to collect it.

The six

Turnover grouped by manager. The single most informative view available and the cheapest — you already have the data, it just is not grouped that way.

Whether people can speak up. A short repeated question rather than an engagement survey: can you raise a problem here without it costing you.

Whether trajectory has been discussed. Not whether a form was filled, whether the conversation happened and when. Different question per stage.

Time since a meaningful piece of feedback. Somebody who has not been told anything specific about their work in six months is disengaging whether or not they know it.

Internal movement rate. People who cannot move within an organisation move out of it, and this number is usually zero without anybody noticing.

And departure clustering. After a policy change, a reorganisation, a promotion round — the pattern is the finding.

What each costs

Turnover by manager: an afternoon a year. Reorganising data you hold.

Speaking-up question: two minutes per person per quarter, and it must be asked the same way every time or the comparison is worthless.

Trajectory conversations: ninety seconds per person, three times a year, recorded as a date rather than a form.

Feedback recency: nothing, if managers note the date.

Internal movement: a count.

Clustering: a chart with dates on it.

Total: perhaps two days a year for an organisation of a hundred, against replacement costs of 50–200% of salary per avoidable departure (SHRM).

Why they still lose to the dashboard

Because they produce a handful of numbers a year and a dashboard produces thousands.

Because each requires a person to do something rather than a system to emit something.

And because they are uncomfortable. Turnover by manager names people. A speaking-up score below half is a finding about the leadership, not about the workforce.

That discomfort is the actual obstacle, not the effort.

How to start

One metric, one quarter.

Turnover by manager, because it uses data you already have and requires nobody's cooperation.

If the departures cluster, you have found the thing. If they do not, you have eliminated the largest single factor and can move to the next.

Adding all six at once produces a project that stalls, and a stalled measurement programme is worse than none because it discredits the next attempt.

The short version

• Six predictors: turnover by manager, whether people can speak up, whether trajectory was discussed, feedback recency, internal movement, and departure clustering
• None arrives automatically, and all six together cost roughly two days a year for an organisation of a hundred
• Against replacement costs of 50–200% of salary per avoidable departure, that is not a close comparison
• They lose to dashboards because they produce few numbers, need a person, and are uncomfortable — the discomfort is the real obstacle
• Turnover by manager names people, and a low speaking-up score is a finding about leadership rather than workforce
• Start with one metric for one quarter, and use turnover by manager because it needs nobody's cooperation

For additional context on this topic, see Digital Camera World.