Your Ops Metrics Are Clean and Your Best People Are Leaving

Here is a pattern worth recognising, because it repeats and because the reporting stack is structurally unable to show it.

Throughput is inside tolerance. Quality is inside tolerance. Cost per unit is flat or improving. Every operational metric a board reviews is clean, and has been clean for three consecutive quarters. In the same period, three of your strongest operators have resigned, and the exit interviews said something vague about growth opportunities.

Nothing in the operational data is wrong. The data is simply measuring the wrong layer.

Output metrics conceal effort

An operational metric records what came out of a process. It does not record what it cost the people inside that process to produce it. Those two things diverge, and they can diverge for a long time before the output moves, because good people absorb friction. That is what makes them good.

A capable operator encountering a broken handoff does not let the number drop. They work around it. They call someone. They stay late. They rebuild the report manually. The metric stays green, and the cost of keeping it green accumulates entirely inside one person, where no system can see it.

Then that person leaves, and the metric moves — not because the process changed, but because the compensating effort stopped.

Your best people are the reason the dashboard looks fine. They are also the ones paying for it.

Friction clusters, and clusters are findable

The useful property of process friction is that it is not evenly distributed. It concentrates. It concentrates at handoffs between teams, at systems that do not talk to each other, at approval steps owned by someone with no capacity, and at any point where the official process and the actual process diverged years ago and nobody updated the documentation.

Because it concentrates, it is findable — but only if you are collecting the signal. Friction lives in the experience of doing the work, which means the only sensor for it is the people doing the work.

A friction audit you can run this quarter

This does not require a platform, a consultant, or a budget. It requires asking a narrow question repeatedly and writing down where the answers cluster.

One question, asked monthly, to everyone in the operating function: What took you longer this month than it should have, and what was in the way?

Not a satisfaction survey. Not engagement. A specific, concrete, answerable question about the mechanics of the work. Free text. Two minutes.

Then do the only part that matters: tag every answer to a process step and a system, and look at the distribution rather than the content. You are not reading complaints. You are building a heat map. One team, one handoff, or one system will be carrying a disproportionate share of the answers, and it will be the same one next month.

Three months of that produces something no operational dashboard can: a ranked, evidenced list of where your organisation is spending effort that never appears in any metric.

What to do with it

Fix the top cluster, not the top complaint. The loudest individual answer is rarely the biggest cost. The densest cluster is.

Re-measure after the fix. If the cluster moves elsewhere, you relocated the friction rather than removing it. That is worth knowing.

Watch for silence. A team that stops answering has not run out of friction. It has concluded that answering does not change anything, which is a more serious finding than any individual complaint.

Read it alongside attrition risk. Sustained friction concentrated on your highest performers is a resignation you can see coming, one or two quarters out. That is the entire value of the exercise.