Middle management is the layer most often described as overhead. Much of what it does is invisible precisely because it succeeds, and the cost surfaces only when it is removed.
Translation runs in both directions
Strategy arrives from above as direction rather than instruction. Someone has to convert it into specific work for specific people in a specific week.
Information travelling upward requires the same conversion in reverse: individual problems aggregated into patterns senior leaders can act on.
Neither conversion is recorded anywhere. It appears in the output of the layers above and below, which is why the layer doing it looks like a pass-through.
Buffering absorbs variation
Priorities change, urgent requests arrive and resources are reallocated. A manager who passes every change straight through leaves their team unable to finish anything.
Effective managers absorb a proportion of that churn, resequencing quietly and shielding the team from changes that will be reversed within days.
This is genuine work with a real cost, and it registers only as an absence of disruption, which no measurement system captures.
Conflict is resolved before it escalates
Most disagreements between teams, over priority, scope, or who owns a problem, are settled at manager level without reaching an executive.
Removing that level does not remove the disagreements. It routes them upward, where they consume far more expensive attention and take longer to resolve.
Flat organisations that report exceptional efficiency often have senior people spending much of their week on coordination that a middle layer previously handled.
Judgement about people accumulates locally
Who is struggling, who is ready for more, who works well with whom: this knowledge is built through daily proximity and is not transferable through a system.
Performance processes attempt to formalise it, and they capture a fraction. The rest lives with the manager and disappears when the role does.
Which is why reorganisations that reshuffle reporting lines frequently produce a period of poor allocation decisions that nobody predicted.
Why the layer gets cut anyway
Middle management costs are easy to identify and its outputs are not, which makes it the obvious target whenever headcount reduction is required.
The savings are immediate and visible; the consequences are delayed, diffuse and attributed to other causes when they arrive.
Organisations that cut and then rebuild the layer under a different name are common, and the rebuilding is rarely recognised as a reversal.