Why your KPIs look good while your business still feels out of control.

The dashboard is green. The customers are unhappy. Managers are chasing exceptions, and the owner has a strong suspicion that the business is less healthy than the presentation suggests.

Perhaps the traffic lights need a different electricity supply.

More often, the problem is what the measures include, what they exclude and how people use them. This article examines how to build management information that supports decisions and reveals operational performance, rather than simply making it easier to report activity.

1. How a green dashboard can describe a disappointing service

A call centre can answer quickly and still fail to resolve the customer’s problem.

A delivery team can dispatch on time and still send the wrong order. The green status is accurate within a narrow definition; the experience remains poor.

A KPI represents a chosen aspect of performance. It cannot automatically stand in for the whole service.

Begin by establishing the outcome the customer needs and the result the business is trying to achieve. Then examine whether the dashboard covers those outcomes or only the activities leading towards them.

At DeaconOps, I’d trace a sample of real work against the measures. That tests whether a reported success corresponds to an actual completed outcome.

2. Separate activity, output and outcomes

Sending ten emails is activity. Completing a decision is output. Giving the customer a correct, usable answer is an outcome.

A particularly determined operation can increase the first while struggling with the other two.

Activity measures help understand effort and workload. Output measures show what is produced. Outcome measures indicate whether that production achieves its purpose.

You usually need a relationship between them, rather than a declaration that one category is always superior. Low output may reflect insufficient capacity, complex work or poor inputs; customer outcomes can also be influenced by factors outside one team’s control.

Make those relationships explicit. A useful dashboard helps managers ask a better next question instead of treating a single number as the answer.

3. Identify targets that move problems elsewhere

A team clears its queue by transferring difficult cases to another department.

The local result improves. The next department acquires a backlog and a strong opinion about teamwork.

Measures can influence behaviour, particularly when incentives or consequences depend on them. Goodhart’s law is often used to express the risk that a measure becomes less informative when it becomes a target.

The practical issue is whether the behaviour rewarded by the target still supports the intended outcome.

Review closure definitions, transfers, reopened work and exceptions alongside headline productivity. Discuss where staff experience pressure to satisfy the measure in ways that do not help the customer.

This is a design question as well as a conduct question. The target may need balancing, clarification or replacement.

4. Look beyond averages and isolated movements

An average wait of three days sounds reasonable until you discover some customers wait three weeks.

The average has not lied. It has simply declined to introduce you to everyone.

Segment where differences matter: work type, complexity, route or stage. Examine the age and distribution of unfinished work alongside averages for completed items.

Also review performance over time. Deming’s work on variation explains why a movement in a result does not automatically establish a new cause. Reacting to every fluctuation can consume management time and destabilise a process.

Use methods appropriate to the data and decision. The point is to distinguish meaningful signals and persistent limitations from noise, not to decorate every report with unfamiliar statistics.

5. Give each KPI a decision and an owner

A dashboard with forty measures can produce a meeting in which everyone reads out numbers and nobody changes anything.

That is reporting with an audience.

For each measure, identify its purpose, definition, source, owner and the decision it can inform. Agree how often it needs review and what pattern or threshold warrants investigation.

Avoid thresholds chosen solely because a spreadsheet needs a colour. A response should reflect the consequences and the process’s behaviour.

If nobody can explain what they would do differently when a measure changes, question whether it belongs in the main management view. It may still be useful as supporting analysis rather than a permanent headline.

6. Build measures that expose trade-offs

Faster work can be valuable. Faster work that creates more corrections may simply move the cost into next week.

A single speed measure cannot explain that relationship.

Combine a small set of measures covering demand, completion, quality, customer impact and resource use where relevant. The set should make it difficult to improve one result by quietly worsening another.

Definitions need consistency across teams and periods. A change in the way work is classified can create an apparent improvement without changing the service.

I’d keep a short measure dictionary and make important limitations visible. Reliable management information depends as much on shared meaning as it does on the software displaying it.

7. Turn performance meetings into decisions

Reading a dashboard aloud does not become analysis because it happens on a Tuesday.

The useful work is understanding the exception, agreeing a response and checking what happened afterwards.

Structure the discussion around material changes, unresolved problems and decisions needed. Establish what the evidence supports, what remains uncertain and who will do what next.

Record the action, owner, expected effect and review point. Revisit previous decisions so the meeting learns from outcomes rather than continually generating new commitments.

Where a question needs deeper investigation, assign that investigation instead of improvising a confident explanation in the room. Not every variance needs immediate intervention.

8. When management information needs redesign

If managers distrust the numbers, maintain competing reports or make decisions outside the formal meeting, the reporting process deserves examination.

Buying another visualisation tool may leave those problems intact.

DeaconOps’s Improvement Projects can help align measures, definitions and management routines with the decisions the business needs to make. Operational Reviews provide ongoing independent challenge and senior support.

Where weak visibility is part of a wider performance problem, an Operational Diagnostic can establish the priorities.

Start a conversation if the dashboard explains what was counted but leaves you uncertain about what to do.

Questions about operational KPIs

How many KPIs should a business have?

There is no universally correct number. Use a manageable set tied to important outcomes and decisions, with supporting detail available for investigation.

Should we remove all activity measures?

No. They can help explain workload and resource use. The risk is treating activity as sufficient evidence of a successful outcome.

Will a new dashboard improve decisions?

Only if the data, definitions, ownership and management routines support useful action. A new interface does not resolve those requirements by itself.

---

About the author

Dean Cain is the founder of DeaconOps, with more than 15 years’ experience across operational delivery, leadership and regulated environments. Read about the DeaconOps approach.