When KPIs Create the Wrong Behavior

A KPI can improve while the business outcome deteriorates.

The issue is often the gap between the result an organization wants and the proxy it rewards. A measure can be accurate on its own terms and still encourage choices that weaken service, quality or control.

Start with the outcome, then examine the measure

Consider an illustrative order-processing team assessed mainly on how quickly it closes orders. If 'closed' means an administrative status change rather than verified fulfillment, reported speed may improve while incomplete deliveries or reopened orders remain elsewhere in the process. This is a possible failure mode, not an observed company incident or evidence of misconduct.

A stronger definition asks what must be true for the customer and the business before an order counts as complete. It then identifies the records that demonstrate that outcome.

Four different problems need four different responses

A poorly chosen measure tracks the wrong outcome. An incentive arrangement gives too much weight to that measure. A data-quality problem makes the reported result unreliable. Deliberate gaming involves intentional behavior that must be established through evidence. These issues can interact, but they should not be treated as interchangeable.

For example, a rise in reopened orders may indicate premature closure, but it may also reflect demand mix, system changes or capacity constraints. A pattern creates a question to investigate; it does not settle the explanation.

Ask six questions before rewarding the result

1. What business outcome should this KPI support?
2. What shortcut could satisfy the measure while undermining that outcome?
3. Where would the resulting cost or risk appear: later periods, another team, the customer or the balance sheet?
4. Which complementary measures would expose the trade-off?
5. What source evidence confirms the result and the completeness of the underlying population?
6. Who owns the design, and who can challenge the reported performance?

These questions are an Inside Assurance practitioner framework, not a prescribed regulatory scorecard.

Balance the measure, then test the evidence

In the order example, speed can be considered alongside verified fulfillment, reopen events and returns or errors. The appropriate combination depends on the process. Adding another dashboard field is insufficient if definitions remain ambiguous or source records are incomplete.

A practical assurance test traces selected dashboard entries back to the order, completion rule, delivery evidence and later adjustments. It reconciles the population, examines exceptions and evaluates management's explanations against the available records. Selection and testing should reflect the risk; the absence of an exception in a small sample does not prove the entire incentive system is sound.

Treat incentives as part of governance

Financial-sector guidance provides a useful foundation: the FSF's 2009 compensation principles connect oversight and reward design to risk, while the US banking agencies' 2010 guidance addresses incentive compensation that can promote excessive risk-taking. Those documents have financial-institution scopes. The operational application here is our broader-business interpretation, not a claim that their specific rules apply to every company.

Management owns KPI definitions, incentives and corrective action. Appropriate governance and risk functions challenge the design and outcomes. Internal audit independently evaluates whether performance reporting and related controls support the intended objectives; it should not become the owner of the incentive scheme.

The useful question is not only whether the target was achieved. It is whether the way it was achieved strengthened the business.