Small-business outsourcing research
How long do customer refunds wait for owner approval?
A protocol for measuring the delay between a refund request and the owner decision, with stage times and the customer's waiting experience.

Research question and scope
This protocol asks how long a customer refund waits between the request and the owner's decision. The unit is one refund request for one order, identified by the order number and the requested amount. The study measures the time to evidence complete, the time to owner decision, and the time to the customer seeing the result. It does not judge whether the refund was the right business decision.
Set the scope to one small business, one sales channel, and one refund policy. Record the approval thresholds, the payment methods, and the channels through which a request can arrive. If the policy changes during the window, keep the earlier definition for requests already in progress and analyze the later cohort separately. A new threshold can look like a faster process when it is only a different rule.
Define the population as every refund request recorded during the window, including requests that were withdrawn, rejected, or partially approved. The unit is the request, not the customer, because one customer can request several refunds. A request that is reopened after a denial is a new unit linked to the original. Test the definition on two recent requests before collection begins.
Definitions that make the measure reproducible
Define the request time as the moment the customer's message or form was recorded, not the moment an agent first read it. Define evidence complete as the point when the order reference, the reason, and the required proof are present. Define the owner decision as the recorded approval, rejection, or partial amount. Define customer result as the refund appearing on the customer's statement or the customer being told the outcome.
Do not count an internal note that says refund as a decision. Do not count an approved amount that has not been sent to the payment processor as a customer result. A request that is approved and later reversed keeps both dates. Each definition should include a positive example and a boundary example so a second reviewer classifies the same record the same way.
Record the request channel, the amount band, the payment method, and whether the customer is new or repeat. These strata matter because a card refund and a cash refund have different settlement times. Include requests that arrived through a phone call, because those are often logged later and appear slower. Use one category list for every request.
Evidence standard and data dictionary
For each request, capture the order number, the amount, the reason, the requester, the approver, and the channel. Add the request time, the evidence complete time, the decision time, and the customer result time. Record the number of follow-ups from the customer and the number of times the request was returned for missing information. Keep the message thread with the record.
Also capture whether the refund was full or partial, whether a return was required, and whether the customer received a store credit instead of money. These fields change the customer's experience even when the decision time is the same. A second reviewer should be able to rebuild the timeline from the record without reading every message. That test is the evidence standard for the sample.
If the business uses a payment processor with a settlement delay, record the processor's stated timeline separately from the business's decision time. A customer may wait days after approval for the money to appear. The protocol should not attribute the processor's delay to the owner's decision. Report the two intervals separately in every table.
Sampling and collection
Draw the sample from the refund log rather than from a list of complaints. Record the export date, the filters, the row count, and the selection rule. Include requests from every channel and every day of the week so the result is not dominated by one busy period. If the population is ordered by time, a fixed interval spreads the sample across the window.
Collect the request and the decision as they existed at the time, not after a later cleanup. If a note was added days later, record the addition date and analyze the original delay separately. Do not infer a decision from a payment that was later reversed. Note the collection date separately from the request date.
A useful additional measure is the share of requests that were complete on the first contact. A complete request names the order, the reason, and the evidence. Record the missing fields and the time spent collecting them. That view separates a customer-preparation problem from an approval bottleneck.
Analysis and reporting
Report the number of requests, the median time to evidence complete, the median time to decision, and the median time to customer result. Show the counts by channel, amount band, and payment method. Present the share of requests that waited more than one business day at any stage. Use the same denominator for every rate and show the number of records with incomplete times.
Inspect the slowest requests and describe which stage dominated without naming the customer or the employee. A wait before evidence complete may point to a missing template, while a wait after that may point to an owner who is unavailable. Compare the result with the amount band, because a large refund may need more review. Report rejected requests separately so their timelines do not distort the approved cases.
A useful additional view is the time from the owner decision to the customer message. A fast decision that is not communicated does not reduce the customer's wait. Show the communication delay beside the decision delay. That view helps the owner fix the notification step rather than adding more approvers.
Limitations and sources of error
The record cannot show whether the owner made a fair decision or whether the customer was satisfied with the outcome. It can show only the request, the evidence, the decision, and the result times. A fast decision is not proof of good service, and a slow one is not proof of neglect. The protocol also cannot capture a refund handled in a phone call that was never logged.
Payment systems can post a refund with a delay that the business does not control. A customer may see the money before the record is updated, or after. Record those gaps and their treatment instead of dropping them silently. If the owner and the approver are the same person, note the independence limitation and analyze that group separately.
A useful check is whether a new assistant can follow the refund guide and predict which requests need owner approval. Ask the last two hires which case they had to escalate and why. That answer tests the guide, not the person. Update the guide with the missing case and the owner.
Review cadence and next action
Give the result to the owner and the support lead and ask them to choose one observable change. The change might be a complete-request template, a named backup approver, a smaller evidence requirement for low amounts, or a daily approval slot. Record the owner, the effective date, and the measure that will show whether the change worked. A change without a measure becomes an opinion at the next review. Avoid changing the threshold and the tool at the same time.
Repeat the protocol with the same definitions after the change and compare the same fields. If decision time fell but customer result time did not, the bottleneck may have moved to communication or settlement. If evidence time fell but decision time rose, the queue may have moved to the owner. Close the review when the owner accepts the finding and the next cohort is scheduled.
A useful additional view is the share of refunds that were approved within the first response. A customer who receives a decision in the same conversation has a shorter experience even when the payment settles later. Record the first response time and the decision time together. That comparison helps the owner see whether the team is slow to decide or slow to reply.
Ask the last five customers who requested a refund whether they knew what would happen next. A clear expectation reduces follow-up messages even when the wait is unavoidable. Record the answer and the point in the process where the expectation was set. Use the finding to improve the first reply, not to add another approval step.
Sources
- Federal Trade Commission, consumer protection guidance, Federal Trade Commission; checked 2026-11-08
- U.S. Small Business Administration, manage your business, U.S. Small Business Administration; checked 2026-11-08
- Better Business Bureau, customer service guidance, Better Business Bureau; checked 2026-11-08
- U.S. Government Accountability Office, internal control standards, U.S. Government Accountability Office; checked 2026-11-08
Frequently asked questions
Does a favorable sample prove the routine caused the result?
No. The design is observational and supports a bounded operating decision, not a universal causal claim.
What should an owner review first?
Start with high-consequence defects, missing denominators, exceptions, and cases that required the owner to reconstruct the evidence.
Related research
What share of supplier invoices arrive without a purchase order?
A measurement protocol for invoice matching gaps, with channel, supplier type, and the time spent resolving a missing order.
How often does a small-business assistant escalate a support ticket?
A protocol for measuring escalation frequency, the reason, the time to resolution, and whether escalation helped the customer.
How long does month-end close take with a virtual assistant?
A protocol for measuring close duration by task, dependency, and rework instead of a single number at the end of the month.