Table of contents
Research context
This protocol asks how long a client database change approval takes from submission to decision. The unit is one change request for one production database, identified by the change identifier and the target environment. The study measures the time in review, the time waiting for missing evidence, and the time to execution. It does not judge whether the change was necessary or whether the reviewer was careful enough.
Record whether the request included a rollback plan at submission or only after the reviewer asked. A request that arrives incomplete creates a second queue and inflates the apparent review time. Measure the time to a complete request separately from the time to a decision. That split shows whether the delay is preparation or approval.
Set the scope to one client, one database platform, and one change policy. Record the change types, the approvers, and the review window at the start. If the policy changes during the window, keep the earlier definition for requests already submitted and analyze the later cohort separately. A new reviewer can look like a slower process when it is only a different queue.
Check whether the reviewer had delegated authority for the change type. A reviewer who must consult another person for every schema change adds a hidden stage. Record the delegation and the person consulted. The approval time should reflect the actual decision path, not an assumed one.
Define the population as every change request submitted in the window, including requests that were cancelled, rejected, or still pending at the observation date. The unit is the request, not the migration file, because one request can include several statements. A resubmitted request after missing evidence is a new unit linked to the original. Test the definition on two recent requests before collection.
Measure the time from approval to execution as its own interval. A fast approval that waits days for a deploy window still delays the business. Record the window and the reason for the wait. The client may prefer a faster execution path over a faster approval.
Record the change type, the affected tables, the estimated row count, the risk tier, the reviewer, the submission time, the first review time, the decision time, and the execution time. Add the number of times the request was returned for missing evidence and the evidence that was missing. Keep the change record with the approval thread. Ask the reviewer which part of the request was hardest to assess. The answer may point to a missing row estimate, an unclear rollback, or an unfamiliar table. Record the answer and update the template. A small template change can remove a recurring question.
Report the number of requests, the median time to first review, the median time to decision, and the median time to execution. Show the counts by change type, risk tier, and reviewer. Present the share that waited more than one review window and the share that required an emergency path. Use the same denominator for every rate.
Practical application
A useful additional view is the share of requests that were approved with conditions. A conditional approval can be a good control, but it can also hide an unresolved question. Record the condition, the owner, and the verification date. The review should show whether the condition was closed before execution.
Inspect the slowest requests and describe which stage dominated without naming the reviewer. A wait before first review may point to a missing reviewer or an unclear queue, while a wait after review may point to missing rollback evidence. Compare the result with the change type and the row count. Report cancelled and pending requests separately.
Do not compare approval times across clients with different change policies. A client with a weekly committee and one with a daily delegate are not measuring the same process. Keep the comparison inside one client and one policy, or state the difference clearly. The record cannot show whether the reviewer made a good decision or whether the change was safe. It can show only the submission, review, and decision times. A fast approval is not proof of good control, and a slow one is not proof of neglect. The protocol also cannot capture an approval given in a meeting and recorded later.
Check whether the request was submitted during a holiday or a release freeze. A long wait in that period is a scheduling effect, not a review failure. Record the calendar context and report it beside the median. The reader should not attribute a planned pause to the reviewer.
Give the result to the client owner and the offshore lead and ask them to choose one observable change. The change might be a pre-review checklist, a named backup reviewer, a weekly approval slot, or a smaller change template. Record the owner, the effective date, and the measure that will show whether the change worked. Repeat the protocol with the same definitions.
A useful additional view is the number of requests that were withdrawn before a decision. A withdrawn request may indicate that the requester found a simpler path or that the process was too heavy. Record the reason where it is known. The pattern can reveal unnecessary changes as well as unnecessary delays.
Review the emergency path separately from the normal window. An emergency that is used often may mean the normal window is too slow, not that the team is careless. Record the emergency reason and the post-change review. The owner should fix the normal path before tightening the exception.
Operating controls
Close the review with one change and a date to re-measure. The change might be a pre-review checklist, a delegate for low-risk changes, a weekly slot, or a clearer rollback template. Record the owner and the expected effect. Repeat the protocol with the same definitions and compare the same fields.
A useful additional view is the share of change requests that were submitted with a tested rollback. A request without a tested rollback often returns for more evidence, which inflates the approval time. Record the rollback evidence and the reviewer's question. The pattern can point to a template that is missing a required field.
Check whether the reviewer had access to a staging copy that matched production. A reviewer who cannot see the result must rely on the description, which adds a round of questions. Record the environment and the evidence. A shared staging environment can shorten the review without weakening the control.
Ask the offshore lead which change was easiest to approve and why. The answer often reveals a template or an example that worked. Reuse that example for the next request instead of writing a new format. Consistency is a cheap way to reduce review time.
Do not treat a rejected request as a failure of the process. A rejection can prevent a risky change, which is a success for the control. Record the rejection reason and whether the requester submitted a safer alternative. The review should value the catch, not only the speed.
A useful additional view is the share of requests that were approved by a backup reviewer. A backup can keep the window moving, but a high share may mean the primary is overloaded. Record the backup and the reason. The owner can then decide whether to add a reviewer or redistribute the queue.
Record the reviewer's queue depth on the day of the decision. A reviewer with ten pending changes will take longer than one with two, even when the request is identical. The queue depth is context, not an excuse, but it helps the owner decide whether to add a reviewer or reduce the volume. Report it beside the median so the number is not read as a personal performance measure. A queue-depth note is context, not an excuse, and it belongs in the same sentence as the review window.
Sources
Frequently asked questions
What should happen first?
Begin with a bounded ticket, approved access, explicit acceptance evidence, and a named reviewer.
Who approves production changes?
The client's technical owner keeps final merge and release authority.
