In this guide
Measure the service you agreed to receive and the outcome you can observe. Do not invent a response-time promise because it sounds normal for the industry. A useful review separates contractual commitments, internal targets and descriptive measurements.
For benefit-plan service relationships, DOL’s fiduciary guide discusses provider selection and ongoing monitoring. Our service-review method is broader operational synthesis; it does not determine fiduciary status or whether a particular provider has met a legal standard.
Give each metric a definition card
Record the question, start event, stop event, population, exclusions, data source and owner. For response time, decide whether the clock starts at a valid case submission or at an incomplete email. Decide whether “response” means acknowledgment, a substantive answer or resolution. These are different measures.
| Measure | Useful definition question | Common distortion |
|---|---|---|
| Acknowledgment time | When was a valid case received and acknowledged? | Calling an automated receipt a resolved case |
| Resolution time | What evidence establishes closure? | Stopping the clock while the affected issue persists |
| Rework | Which completed items needed correction, and why? | Counting every employer-initiated change as provider error |
| Open-case age | How old are unresolved cases at the review date? | Reporting only closed cases and hiding the backlog |
| Handoff completeness | Which required inputs and receipts were present? | Treating “sent” as downstream acceptance |
Use denominators that match the question
Suppose a fictional review finds four corrected payroll items during a month. That number means something different across 20 items and 2,000 items. State the population and unit, then examine the consequence of each case. A low rate can still include one serious issue.
Do not mix employee-level records, payroll runs and support tickets in the same denominator. If one incident creates ten tickets, counting ten independent failures may distort the pattern. Link related cases while preserving the effect on each affected workflow.
Compare case types before blaming the average
A month with more complex location changes or a new integration may take longer than a month of routine questions. Separate ordinary cases from defined exceptions and explain material changes in mix. This is not a reason to excuse poor service; it is a way to ask a more precise question about where the process failed.
Look at the distribution and the unresolved tail, not only the average. If nine fictional cases close in one day and one remains open for 21 days, the average can obscure the long-standing problem. Show the old case, its consequence and the next decision.
Review the handoff on both sides
Classify causes using evidence: missing employer input, unclear responsibility, provider processing issue, downstream rejection or unknown. Avoid forcing every case into a category prematurely. “Unknown” is a valid temporary state when an investigation is still needed.
A provider may receive incomplete data; an employer may receive an ambiguous request. Review the actual exchange and the required process. The goal is a repair that prevents recurrence, not a scorecard designed to allocate embarrassment.
Use an agenda that ends in decisions
- Review material incidents and unresolved consequential cases.
- Check performance against actual commitments, with definitions visible.
- Examine repeated causes and changes in case mix.
- Agree repairs, owners and evidence of completion.
- Confirm upcoming scope, staffing, platform or policy changes that alter the operating map.
Keep a record of proposed versus agreed actions. A suggestion made during a review is not automatically a contractual service-level amendment. Material changes to obligations should go through the authorized agreement process.
Track whether repairs work
If the action is “improve change submissions,” define the new input requirement and how to verify its use. If the action is a provider report change, inspect the next report against the agreed requirement. Closing the action when a meeting is held does not prove the underlying issue improved.
Revisit the cost assumptions when service scope changes, and update the operating map when ownership changes. Keep the review tied to the employer’s real work rather than broad marketing claims or unverifiable peer benchmarks.
Finally, distinguish a renewal decision from a routine scorecard. Renewal may require a fresh look at credentials, security evidence, pricing and exit options. The metrics inform that decision; they do not make it automatically.