Small-business outsourcing research

What share of customer deletion requests close within the stated window?

A protocol for measuring how many customer deletion requests close within the business's stated response window and where the delay occurs.

What share of customer deletion requests close within the stated window?

Research question and scope

This protocol asks what share of customer deletion requests close within the window the business states in its policy or reply. The unit is one request case, not one customer and not one system. The study measures the closure time, the systems searched, and the share of cases with a written response. It does not judge whether every deletion was legally required.

Set the scope to one business, one intake channel set, and one published window. Record the policy version and the date the window was stated. If the window changes during the study, keep the original window for cases already opened and analyze later cases separately. A new intake form can change the mix of requests without changing the process.

Define the population as every deletion request received during the window, including requests that were withdrawn or denied. Record how each case entered the business, because a request made in a support chat may be logged later than one made through a form. State any exclusion and apply it to every case. Test the definition on ten closed cases before collection begins.

Definitions that make the measure reproducible

Define closure as the date the business sent a written response that states what was deleted and what was retained. Define the stated window as the number of calendar days in the policy in force when the request arrived. Record the start time as the date the request was received, not the date it was assigned. The difference between those dates is part of the finding.

Record whether the request needed identity verification, a legal-hold check, or a system search that the business could not complete. Mark a case as open if any of those steps is unfinished. A case that has a reply but no deletion is not closed. Use the same rule for every channel so one team does not look faster because it closes cases earlier.

Use the case log as the source of dates and keep an export timestamp beside it. Record any case that was reopened after closure and the reason. If a case was merged with another request from the same customer, keep the earlier start date. A merged case that adopts the later date can make the closure rate look better than it was.

Evidence standard and data dictionary

For each case, capture the case identifier, the received date, the channel, the verification date, the systems searched, the deletion date, and the response date. Add the retention reason for any record that was kept and the owner of the case. Record whether a legal hold was checked and who confirmed it. Keep the response text linked to the case.

Capture the reason for any delay when it is documented: identity verification, a vendor that needed a manual request, a backup cycle, or a legal review. Do not infer a reason from an empty field. If the reason is unknown, mark it unknown and report the count. A guessed reason hides the process step that actually needs repair.

Record the number of reminders sent and the channel used. A reminder that reaches the system owner on the last day is different from one sent at the start. Note whether the business has a written escalation rule and whether it was followed. A policy that exists but is never used is a different finding from no policy at all.

Sampling and collection

Collect every request in the window rather than a sample, because the closure promise applies to each one. If the business has several brands or locations, collect each separately and record the entity. A combined rate can hide a channel with a persistent problem. Report the number of cases behind every percentage.

Draw dates from the case system and the response log, not from a spreadsheet maintained after the fact. If the case system is incomplete, say so and reduce the claim. Record the export date, the filters, and any case that could not be matched to a response. A missing response record is itself a finding about the process.

Ask the support lead to identify weeks when intake volume was unusually high or a system owner was absent. Those weeks can produce delays that have nothing to do with the process design. Record the event and analyze those cases separately. The reader should be able to tell whether the rate reflects the rule or the calendar.

Analysis and reporting

Report the share of cases closed inside the window, the share closed late, and the share still open at the end of collection. Show the median and range of closure days. Present the results by channel, system, and week so the reader can see whether the problem is concentrated. Use the same denominator for every rate.

Separate time spent waiting for the customer from time spent inside the business. A request that waits for identity evidence is a different finding from one that waits for a system owner. Report both because they lead to different fixes. A business may accept a customer delay but should not accept an internal delay without a documented decision.

Inspect the slowest cases and describe which stage dominated without naming the customer. An intake gap, a vendor request, and a backup cycle each point to a different repair. Compare the delay with the stated window and the reminder schedule. Report cases with incomplete dates rather than dropping them from the denominator.

Limits of the evidence

The case record can show that a response was sent, but it cannot show that every copy was actually deleted. A system that is not in the inventory will not appear in the search list. The protocol also cannot capture a verbal assurance that was never written down. State that limitation before presenting the rate.

A small business may have too few requests to support a stable trend. Describe the case count and the channels instead of ranking staff. If one owner handles every request, the result may reflect that person's availability more than the process design. Do not use the rate as a performance score without the context of volume and turnover.

Do not compare a business that closes cases after a written response with one that closes them when the deletion job is queued. The two operate different definitions even when the rate looks similar. Report the closure rule beside the rate. A target that can be met by sending a reply before the work is done is not a good target.

What to change

Give the result to the owner and the support lead and ask them to choose one observable change. The change might be a single intake form, a named system owner, a reminder at the midpoint, or a legal-hold check at intake. Record the owner, the effective date, and the measure that will show whether the change worked. Repeat the protocol with the same definitions after the change.

If the closure rate rises but the open-case count also rises, the business may be receiving more requests rather than handling them better. If both fall but responses become shorter, the team may be closing cases before the work is complete. Report the movements together before declaring success. Close the review when the owner accepts the finding and the next collection date is set.

Ask the system owner which request was hardest to complete and why. A concrete answer often reveals an unlisted tool, a vendor with no self-service path, or a backup that cannot be searched. Use that answer to improve the inventory rather than to blame the team. The next measurement should test whether the same delay shrank.

Sources

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