Measuring the Manager
Organisations measure what people produce and rarely measure the person deciding what they produce. Given the size of the effect, that is the largest gap in most measurement programmes.
Four measures
Turnover among their reports, over two or three years, compared to peers with similar roles. The cheapest and most informative. For a related reference, see self-reporting bias.
Whether their people can speak up. A single repeated question, asked identically each quarter, reported at team level and never attributed to individuals.
Internal movement out of the team. People moving to other teams is healthy; people never moving is a signal, and people only ever leaving the company is a stronger one.
And whether development conversations happened. A date, not a form.
All four use data you can get without a new system.
Why upward feedback usually fails
Anonymity is not credible in a team of six. People know it, and they answer accordingly.
It is annual, which makes it an event rather than a channel.
It is used for evaluation, so the manager treats it as a threat and the reports treat it as risky. The same conflict as the performance review doing four jobs: feedback and judgement in one instrument.
And nothing visibly happens afterwards, which teaches people that participating costs something and returns nothing.
What works better
Team-level metrics rather than opinion surveys. Turnover, movement and speaking-up scores are harder to game and do not require anybody to accuse anybody.
Comparison across managers rather than against a standard. An absolute score means little; the same measure across twelve teams means a great deal.
And a stated response. If a signal appears and nothing changes, the next round of data is worse — not because the situation worsened, because people stopped answering honestly.
The trap
Ranking managers on a composite score.
It creates exactly the incentive that destroys the data, the same mechanism as everywhere else on this site: the measure becomes a target, the target gets managed, and the number stops describing anything.
Use the measures to find where to look, not to produce a league table. A manager whose turnover clusters is somebody to have a conversation with, and the conversation is the intervention.
What this is not
Not a case for removing managers. The evidence points the other way — the problem is untrained managers, not management.
And not a performance management instrument. These measures tell you where support is needed, and using them for dismissal converts a diagnostic into a threat within one cycle.
The short version
- Four measures: turnover among reports, whether their people can speak up, internal movement out of the team, and whether development conversations happened
- All four use data available without a new system
- Upward feedback fails because anonymity is not credible in small teams, it is annual, it is used for evaluation, and nothing visibly follows
- Team-level metrics are harder to game and require nobody to accuse anybody; compare across managers rather than to a standard
- Ranking on a composite score creates the incentive that destroys the data
- Use the measures to find where to look, and treat the conversation as the intervention rather than the consequence
For additional context on this topic, see Canva.