The first said the most expensive problems are the ones you can’t see. The second gave that invisibility a name — process debt — and showed how it compounds while no one is looking. The third explained why the most widely used process view (case-centric) structurally hides parts of the truth. The fourth warned that AI agents will faithfully execute whatever process you hand them, at machine speed, blind spots included. One thread runs underneath all four: you cannot manage what you cannot measure, and most operations are measuring the wrong things. This post is the constructive close.
Why dashboards default to lagging metrics
Three forces pull every operational dashboard toward the same shape:
- Lagging metrics are easier to count. Revenue, throughput, on-time rate, DSO — these are outcomes. They sit in one system, in one column, with a well-understood definition.
- They’re comfortable to report. A number that describes something already finished can be explained. A number that predicts something that hasn’t happened yet has to be defended.
- They’re tied to incentives. Quarterly targets are outcome targets. What gets rewarded gets measured. What gets measured is what’s already happened.
The result: every operations leader is looking at a dashboard that tells them exactly what went wrong, exactly one quarter too late to prevent it.
Motion versus health
The distinction that matters isn’t leading versus lagging. It’s motion versus health.
Motion metrics describe activity. How many invoices processed, orders shipped, tickets closed. They rise when the business is busy and fall when it isn’t. They tell you the operation is running.
Health metrics describe the condition of the process running underneath the motion. They can hold steady for months while motion looks perfect — and then move, sharply, weeks before anything in the motion layer starts to break.
Most dashboards measure motion. The five metrics below measure health.
1. Exception rate trend
Not exception count. Not exception percentage. The trend line of exceptions as a share of total transactions, tracked week over week.
Absolute exception rates are noisy — a 5% exception rate might be excellent in one process and disastrous in another. But a rising trend, in any process, means one thing: the standard path is losing ground to workarounds. This is process debt accumulating in real time. By the time it shows up in cost or cycle time, the debt has already compounded.
2. Rework loop frequency
The number of times the same activity fires more than once for the same case, per week.
Rework is the tax the operation pays on process debt. In most organisations it’s absorbed into “normal throughput” — the hidden factory — because no line item captures it. Object-centric process mining makes it visible for the first time. A rising rework frequency almost always precedes a rising cost per transaction by one to two quarters. That’s the warning window.
3. Handoff wait time — the tail, not the average
Every process has handoffs. Every handoff has a distribution of waiting times. The average waiting time is a comforting number. The P90 and P99 are the honest ones.
Value dies in the tail. A P50 handoff of two hours and a P99 of eleven days is not a healthy process — it’s a process where the exceptions are quietly destroying customer experience and working capital. Averages hide it. Tail metrics don’t.
4. Cycle-time variance
Not average cycle time. The spread between the fastest and slowest 10% of completions.
Two operations with identical average cycle times can be in radically different health. One is running consistently. The other is running well most of the time and catastrophically some of the time — and the catastrophes are what your customers remember, what auditors flag, and what agents will amplify.
Growing variance is the earliest available signal that a process is losing coherence.
5. Automation-ready ratio
The percentage of transactions that ran end-to-end on the standard path, with no manual workaround, no exception handling, and no human bridge across an integration gap.
This is the metric that ties everything together. It measures what percentage of the process an AI agent could safely run today — which is the same as measuring how much process debt has been paid down. A rising ratio means the operation is getting healthier and more automatable at the same time. A falling ratio means agent-readiness is going backwards, no matter how many pilots are in flight.
How to shift the dashboard
Three moves, in order:
- Add, don’t replace. Keep the existing motion dashboard. Add a health dashboard alongside it. Executives compare them; the gap is where the conversations start.
- Track trends, not absolutes. Every health metric on this list is a trend metric. The absolute value tells you where you are. The trend tells you where you’re going.
- Assign an owner. Motion metrics have owners already. Health metrics rarely do. Until someone is accountable for the trend, the trend won’t move.
The thread, pulled tight
Everything this month came down to visibility. Invisible costs. Invisible debt. Invisible distortion in the process map. Invisible risk carried into agent deployment.
Better measurement isn’t a nice-to-have. It’s the precondition for everything else being fixable.
That’s what process intelligence exists to do — and it’s what Sparkhs’s Celonis Certified consultants help operations leaders put in place.