The Dashboard That Everyone Trusts and Nobody Checks
Kestrel Home Services operates 140 branches. Its monthly performance pack reports branch profitability, customer satisfaction and staff turnover. The COO has noticed that the top 20 branches by reported profit are also the top 20 by customer satisfaction - an unusually clean relationship. Investigation shows: branch profit is calculated after allocating head-office costs by revenue; satisfaction is measured by a survey sent only to customers who logged a complaint that was resolved; turnover counts leavers but not transfers between branches. Regional managers are bonused on branch profit and satisfaction. Two branches with high reported profit have the highest complaint volumes in the group.
1. Critique the performance measurement system, explaining how each metric is distorted and what behaviour it drives.
Read the worked answer
The clean correlation is an artefact, not a finding.
Branch profit
- Head-office costs allocated by revenue means larger branches absorb more central cost; profit therefore partly reflects size and allocation policy, not branch performance. Managers can improve reported profit by shrinking revenue (perversely) or by disputing allocations, which consumes management time.
- Bonusing on it drives cost-shifting and lobbying over allocation methodology rather than operating improvement.
Customer satisfaction
- Surveying only complainants whose complaint was resolved is a severe selection bias: it measures recovery from failure among the already-dissatisfied, and excludes silent dissatisfied customers and the majority who never complain. Two high-profit branches with the highest complaint volumes can look good because they resolve complaints well - the metric rewards firefighting, not prevention.
- Bonusing on it encourages complaint suppression or selective resolution of easy cases.
Turnover
- Counting leavers but not internal transfers understates churn where staff move between branches; a branch can appear stable while losing its best people to transfers. It also ignores absence and engagement, which precede exit.
System-level problem
- All three metrics are internally generated, lagging and gameable, and all are tied to the same bonus. That combination produces convergent, flattering numbers - exactly the pattern observed. The COO should treat the correlation as a warning, not a comfort.
2. Recommend a redesigned measurement approach, addressing metric choice, data source and governance.
Read the worked answer
Metric choice
- Profit: report branch contribution before central allocation, with allocation shown separately. Bonus on contribution and on controllable cost, not on allocated profit.
- Customer: measure satisfaction across a random sample of all customers, plus complaint rate per 1,000 customers and first-time-fix rate. Prevention metrics (repeat contact rate) matter more than recovery scores. Include a silent-customer signal such as retention or repeat purchase.
- People: total attrition including internal transfers, plus absence and engagement, and voluntary-leaver quality (regretted attrition).
- Balance: add at least one forward-looking or external measure per branch (e.g. market share, new customer growth) so the pack is not purely backward-looking.
Data source
- Independent survey sampling rather than complaint-triggered distribution; system-generated complaint counts rather than manager-reported.
- Reconcile reported branch profit to the general ledger centrally, and publish the allocation method so it cannot be lobbied.
Governance
- Separate the reporting of metrics from the rewarding of them: the finance function owns the numbers, HR owns turnover data, and neither reports to the bonus-holder for the metric.
- Introduce a data-quality review each month with a named owner and an exception log; require branches to explain outliers in both directions (unusually high satisfaction as well as low).
- Test the bonus design against gaming: for each metric ask 'how would a rational manager improve this number without improving the business?' and close that route.
- Pilot the revised pack for one quarter before linking it to pay, and compare rankings under old and new measures to surface distortion.