Table of contents
Research context
This protocol measures the share of privileged production sessions that receive a recorded review. The unit is one session on a named privileged path, identified by the account, the start time, and the ticket. A recorded review means a second person inspected the session evidence and left a dated note. The study describes review coverage, not the quality of the work in the session.
Define the privileged paths before collection. Include database consoles, cloud shells, admin panels, and deployment tools that can change production state. Exclude read-only dashboards that cannot mutate a record. Publish the path list with the version and the date so a later reviewer can reproduce the scope.
Define a session from the platform log. Use the login time, the logout time, and the account, and split a long session when the ticket changes. A session that spans two tickets should be counted twice with the matching portion of the log. Test the split rule on one long session before the full sample.
Define a recorded review as a dated note by a person other than the session user. The note should name the ticket, the reviewed window, the findings, and the reviewer. A comment that says looks fine without a window is not a review for this measure. Test the definition on three historical sessions.
Set the population as every privileged session in the window, including sessions with no ticket and sessions that changed nothing. A session that only read data still belongs if the path can read customer records. Record the path, the account, the ticket, the duration, and the review status. Show the sessions that could not be matched to a ticket.
Practical application
Record the evidence source for each session. A platform recording, a command log, and a ticket note are different levels of evidence. Classify the session by the strongest source available and note the missing sources. The study should report how many sessions had no usable log.
Report the number of sessions, the number reviewed, and the share reviewed by path. Show the counts by account and by week. Present the sessions with no ticket separately from the sessions with a ticket and no review. Use the same denominator for every rate.
Separate a missing reviewer from a missing log. A session with a clear log but no review is a process gap, while a session with no log is an observability gap. Report the two categories separately so the client can choose the right fix. Record the reason code beside each unreviewed session.
Inspect the unreviewed sessions around weekends and holidays. A cluster may mean the reviewer rota has a gap or the review step depends on one person. Compare the pattern with the session volume. A coverage number that hides a weekly hole is not useful.
Do not treat a high review share as proof that the sessions were safe. A reviewer may approve every session without checking the window. Sample a few reviewed sessions and compare the note with the log. Report the share of reviews that name a specific finding or a specific check.
Operating controls
The record cannot show whether the reviewer understood the system or whether the session caused harm. It can show only whether a second person left a dated note for a defined window. A missing review is not proof of misuse, and a present review is not proof of control. The protocol also cannot capture a session on a path that exports no log.
A useful additional view is the median time between the session and the review. A review that happens a month later is weaker evidence than one that happens the next day. Record the session end time and the review time. That field turns a coverage count into a review of timeliness.
Give the result to the client security owner and the offshore delivery lead with one observable change. The change might be a reviewer rota, a session-to-ticket link, a shorter review window, or a log export from a missing path. 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.
Close the review with one change and a date to re-measure. The change should address the largest unreviewed category, not the easiest one to report. Record the expected effect and the evidence that will prove it. A review share that is never re-measured is a one-time snapshot.
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.
