How should an insurance agency evaluate outsourced work?
Start with a small set of measures for each workflow: work accepted without required rework, completion within the agreed window, aging items, timely escalation, and record completeness. Review the underlying task sample as well as the percentages. Separate provider-controlled time from waiting on the client, carrier, or agency approver, while also showing the total time the client waited.
This is a management method, not a universal benchmark. A certificate request, a renewal preparation file, and a carrier follow-up have different completion points. One combined "tasks completed" or "average turnaround" figure can hide those differences. An agency should define the measures before asking a provider to demonstrate results.
QA, KPI, and SLA do different jobs
QA is the planned review of how work is prepared, checked, corrected, and documented. A reviewer might compare a drafted certificate with the request and agency instructions, or check that a policy change file includes the returned carrier document and a clear next action. ASQ distinguishes quality assurance from quality control, while noting that organizations sometimes use the terms interchangeably. The agency's actual checklist and reviewer responsibilities matter more than the label.
A KPI summarizes an observed result over a defined period. "First-pass acceptance" is a KPI only after both sides agree what counts as eligible work, what counts as rework, and which source record supplies the numerator and denominator.
An SLA records an agreed service expectation. It might define the covered workflow, operating hours, response or completion window, exclusions, evidence, review cadence, escalation, and what happens after a miss. A dashboard metric is not automatically a contractual SLA. An SLA can be useful without a punitive penalty, but it must be specific enough for the agency and provider to interpret the same event in the same way.
A five-measure scorecard an agency can actually audit
The following is an illustrative scorecard design. Set workflow-specific definitions and targets with your provider; the table does not describe United Alliances' actual performance or a standard across agencies.
| Measure | Calculation or record | What the agency should inspect |
|---|---|---|
| First-pass acceptance | Eligible completed items accepted without provider-caused rework divided by eligible completed items | Reopened tasks, defect reason, and who judged the correction necessary |
| Within-window completion | Eligible items finished by the agreed workflow deadline divided by all eligible items due in the period | Intake time, actionable start, pause reason, completion evidence, and misses |
| Open-item aging | Count and age of open items by workflow and status, including waits | Oldest items, next owner, follow-up date, and unresolved exceptions |
| Escalation timeliness | Exceptions sent to the assigned agency owner within the agreed interval divided by exceptions requiring escalation | Whether the message included facts, urgency, owner, and next action |
| Record completeness | Sampled files with all required AMS notes, attachments, status, and handoff information divided by sampled files | The checklist version, sample selection, and missing elements |
Keep the five figures separate by work type and, where useful, by team or queue. A routine download check may have a different deadline and review rule from a nonstandard certificate or a commercial renewal. Report the number of eligible tasks alongside every percentage. A result from five sampled files should not be presented with the same confidence as a result from a much larger, consistently selected sample.
Illustrative calculation: If 60 eligible completed requests were reviewed and nine needed provider-caused correction, first-pass acceptance is 51 ÷ 60, or 85%. That is not a recommended target. The next question is what those nine errors were and whether any reached a client or carrier before a reviewer caught them.
Define the turnaround clock before comparing providers
"Fast turnaround" is ambiguous until the agency identifies the event being timed. For a COI request, track when it arrives, when it becomes actionable with the required information, when the assigned team acknowledges it, when its preparation is ready for agency review, and when the final communication is sent. Those points may be hours or days apart for reasons the support team cannot resolve alone.
Choose whether a deadline uses business hours or calendar time and specify the time zone, weekends, and holidays. Decide which waits can pause a provider-controlled clock, such as missing client details or a documented carrier response. Record the pause start, reason, owner, and restart time. A task should not disappear from the client-elapsed view simply because it is waiting on someone else.
If a request is reopened because the provider omitted a required document, preserve its link to the original task and count the rework. If the client makes a genuinely new request after completion, record it separately. Otherwise, closing and reopening items can make a weak process look fast.
Do not compare advertised turnaround times between companies unless they describe the same workflow, clock, exclusions, and review point. A 24-hour promise for a prepared file cannot be fairly compared with a 24-hour promise for carrier completion. Ask each provider for a sample report and the written metric definition before treating a number as evidence.
Make QA checks proportionate to the work
An agency needs both checks before release and a later sample of completed records. The pre-release review catches errors before the output leaves the team. A retrospective sample finds recurring defects in otherwise routine work and tests whether the checklist is still useful. Sampling only easy files, or only files that already generated complaints, gives a distorted picture.
For a commercial renewal preparation queue, a QA reviewer might verify that the requested loss runs were ordered, source details match the AMS, missing applications are flagged, and the file has reached the account manager with the open questions visible. For a certificate queue, the reviewer may compare names and requested wording with approved source documents and escalate language or coverage questions. The agency should identify which steps require review by its authorized personnel. The NAIC explains that state regulators license producers to sell, solicit, and negotiate insurance; a task scorecard does not transfer those responsibilities to an administrative assistant.
Use a defect log that records the task ID, defect, severity assigned under agency criteria, discovery stage, correction owner, and root cause. A missing internal file label and a client-facing mistake should not carry the same meaning in a single "accuracy" figure. Review the trend and the corrective action, not only the score.
What should a monthly performance review contain?
A monthly review should reconcile the scorecard to the task register, examine a representative QA sample, discuss exceptions and repeated misses, and assign corrective actions with owners and dates. Weekly queue reviews can handle urgent aging items; the monthly meeting should decide whether the process, training, staffing, or agency instructions need to change.
Ask the provider to bring a short list of the oldest open tasks, each with its next action and dependency. Ask the agency reviewer to bring examples of accepted work and work returned for correction. When a measure improves, check whether volume, task mix, or eligibility rules changed. When it worsens, inspect individual cases before concluding that the team lacks skill or capacity.
The review should end with a decision record: revised instruction, added approval point, owner for training, adjusted capacity, or an agreed exception to the measurement rule. A report that merely repeats green and red percentages cannot show whether anyone acted on what those numbers mean.
Questions to ask a provider before you rely on its reporting
Request a blank or anonymized sample of its task log, QA checklist, escalation register, and periodic scorecard. Then ask who enters each status, whether the agency can see the underlying AMS or queue evidence, who audits the sample, and how disagreement about an error is resolved. Confirm which reporting frequency and review support are included in the particular service model you are buying.
Also define access and change control. A reliable audit trail needs named accounts and a way to identify who changed a record or closed an activity. NIST's small-business cybersecurity guidance recommends multi-factor authentication where offered and continuing risk review. Align the reporting workflow with the agency's system permissions and retention rules.
For a broader selection process, see how to choose an insurance back-office outsourcing partner. This article addresses the performance evidence that should follow that choice. United Alliances' earlier piece on slow insurance operations describes why delays are visible to clients; the scorecard here defines how to measure and investigate them.
How United Alliances describes its quality and reporting model
United Alliances' insurance back-office service page describes intake, AMS or portal updates, a peer or supervisor QA check, completion logging, and escalation of nonstandard requests. The page also describes different reporting by service option: daily task logs and weekly summaries for a full-time dedicated VA, and real-time task tracking and weekly SLA scorecards for a dedicated team. Its insurance virtual assistant page describes task tracking and updates across agency workflows.
Those published descriptions are a useful basis for a conversation, not a substitute for an agreement or an actual sample. Ask United Alliances which model fits your queue, which QA steps apply to your tasks, who reviews exceptions, and which of the five measures above can be produced from your agency's records. Agree on the definitions and reporting frequency for the proposed engagement rather than assuming every option has the same scorecard.
