A KPI is supposed to be a proxy for business health. When it stops being a reliable proxy, it becomes something worse than useless. It becomes misleading.
The danger is not that KPIs are inaccurate. It is that they can be accurate, improving exactly as reported, while the underlying business deteriorates in dimensions the KPI does not capture.
EBITDA improves because maintenance capex was deferred, not because operations improved. Adjusted margins expand because more items are classified as “non-recurring.” ARR increases because contracts are lengthened with discounts that reduce lifetime value. Utilisation improves because the denominator (capacity) was reduced, not because the numerator (output) increased. Customer acquisition cost declines because the company stopped acquiring the customers that cost the most to win, which were also the most valuable.
In each case, the metric moves in the direction management wants. The business moves in the opposite direction. And the board, relying on the metric, approves the trajectory.
The Mechanism: Goodhart’s Law Applied to Corporate Management
Goodhart’s Law states: when a measure becomes a target, it ceases to be a good measure. In corporate management, this means that any KPI attached to executive compensation, board reporting or investor expectations will eventually be optimised for the metric rather than for the business outcome the metric was supposed to represent.
This is not fraud. It is rational behaviour within an incentive system. A CEO whose bonus depends on EBITDA growth will, when facing a trade-off between improving EBITDA and investing in long-term capability, choose the option that improves EBITDA. The investment that would strengthen the business but reduce near-term EBITDA is deferred. The cost cut that improves EBITDA but weakens the business is implemented.
The metric improves. The business deteriorates. And the deterioration is invisible until the KPI is replaced or the underlying damage becomes too severe to conceal.
Six KPIs That Can Improve While the Business Declines
EBITDA improves by deferring maintenance, reducing R&D, cutting training, stretching payables, reducing inventory below optimal levels, reclassifying operating costs as exceptional items and capitalising expenses that should be charged to the P&L. Each action increases reported EBITDA. Each weakens the business.
Revenue growth accelerates by offering pricing concessions that destroy margin, extending credit to customers who may not pay, channel-stuffing distributors near quarter-end, recognising revenue earlier in the delivery cycle and acquiring revenue through M&A without generating organic demand.
ROIC improves by reducing invested capital: selling assets, entering sale-and-leaseback arrangements, writing down assets aggressively. Each action reduces the denominator and increases the ratio without improving the company’s actual ability to generate returns.
Customer metrics improve by selection bias. Net Promoter Score improves by surveying only satisfied customers. Retention rate improves by redefining “active customer.” Average revenue per user improves by losing the smallest users. Each metric improves. The customer base may be weakening.
Utilisation improves by reducing capacity. Closing a factory, retiring equipment, reducing headcount: each increases utilisation by shrinking what is available, not by increasing what is used. The company looks more efficient. It may simply be smaller.
Cash conversion improves by extending payables (damaging supplier relationships), accelerating receivables through discounts (reducing total revenue), reducing inventory (creating stock-out risk) and delaying capital expenditure (depleting future productive capacity).
The Diagnostic: The Three-Part Test
In Northrop Management governance advisory work, we apply a three-part test to every KPI the board relies on.
Test 1: Can management improve this metric while making the business worse? If yes, the KPI is vulnerable to gaming. It may still be useful, but it should not be the primary governance metric without a complementary measure that captures what it misses.
Test 2: What is the KPI not measuring? Every metric captures one dimension and ignores others. EBITDA captures operating earnings and ignores capital intensity, working capital and investment quality. Revenue captures demand and ignores margin and cash conversion. The board should explicitly identify what each KPI omits and ensure that a complementary metric covers the gap.
Test 3: Is the KPI attached to compensation? Any KPI that determines executive pay will be optimised. This is not cynicism. It is the predictable consequence of attaching financial rewards to a metric. The board should assume that compensated KPIs will be managed to the limit of what is permissible and design the metric set accordingly.
Ashish Chaudhary, frames the governance risk directly: “A KPI is only useful if management cannot optimise the metric while destroying the underlying business. The moment a metric can be improved by making the company weaker, it has stopped being a diagnostic and started being a danger.”
Questions for the Boardroom
- For each KPI we report to investors, can management improve it through actions that weaken the business?
- What is each KPI not measuring, and do we have a complementary metric that covers the gap?
- Has any KPI improved consistently while cash generation, customer quality or operational capability has deteriorated?
- Are our executive compensation KPIs designed so that meeting the target requires genuinely improving the business, not merely managing the metric?
- When was the last time we retired a KPI that had become unreliable or introduced a new one that better reflected the current business?
Closing Implication
KPIs are instruments, not outcomes. They measure what they are designed to measure, and they ignore everything else. A board that governs by KPIs alone is governing by instruments that management has the ability, and frequently the incentive, to optimise at the expense of the business they are supposed to represent.
The discipline is not in having KPIs. It is in understanding their limitations, testing their reliability and replacing them when they stop reflecting the truth about the business.
