The 96% SLA That Nobody Believes
PayrollCo provides managed payroll for 40 clients. Its SLA promises that 95% of priority-1 incidents will be resolved within four hours, measured monthly. For six months the provider has reported 96.2% average compliance. Yet three clients missed payroll deadlines last quarter, and the average client satisfaction score fell from 4.4 to 3.6 out of 5. An internal review found that agents reclassify hard incidents as priority 2 to protect the P1 clock, and close tickets once a workaround is in place even when the root cause remains. The account manager’s fix is to raise the target to 98%.
1. Explain why reported SLA compliance can rise while service outcomes worsen. What does this reveal about the metric design?
Read the worked answer
The SLA measures a supplier-side activity - time to close a P1 ticket - rather than the outcome the client buys: accurate, on-time payroll. Three mechanisms explain the divergence:
- Definition gaming: if “resolution” means a workaround is applied, the clock stops before the service is truly restored. Root causes persist and recur.
- Classification gaming: reclassifying a P1 as P2 removes the ticket from the measured population. Compliance rises while the client’s experience worsens.
- Boundary problem: the timer starts and stops inside the provider’s process, not at the client’s payroll run.
The metric is technically valid but managerially misleading. It rewards behaviour that protects the number rather than the service, which is a classic perverse incentive. The problem is not that the target is too low; it is that the measure is disconnected from value.
2. Redesign the SLA and supporting measures for this payroll service. Specify at least three changes and justify each.
Read the worked answer
Redesign around outcomes and guardrails:
- Add an outcome SLA: percentage of payroll runs delivered complete and accurate by the agreed cut-off, plus the number of post-run pay corrections. This measures what clients actually value.
- Measure end-to-end resolution: start the clock at client report and stop only at confirmed restoration. Treat “workaround in place” as a separate status, not as resolved, and define the evidence required to close.
- Segment by criticality: apply the four-hour target only to incidents that threaten a payroll cut-off; set different targets for lower-impact issues.
- Add a quality or experience measure: post-incident client confirmation or a short transactional survey, used as a balancing indicator rather than a stand-alone target.
- Add underpinning measures: OLAs between resolver groups and the service desk, plus monitoring of recurring root causes.
Justification: multiple measures reduce gaming; outcome measures align to value; experience measures capture what the current SLA misses.
3. The account manager proposes raising the P1 target from 95% to 98%. Evaluate this proposal.
Read the worked answer
Raising the target tightens the wrong control. It increases pressure on the same flawed definition, so classification and early-closure gaming are likely to intensify. It may also push agents to close tickets faster without fixing root cause. It does not address the missed payroll runs or the fall in satisfaction.
A better sequence is:
- Fix the definitions and add outcome measures first.
- Then set targets on the redesigned measures.
- If the client wants a headline number, use a small balanced scorecard or a composite, not a single percentage.
- Consider a service credit tied to payroll accuracy or on-time delivery, not only P1 speed.
The proposal is understandable commercially, but it would make the measurement problem worse.