Key stats
- 4 advisory segments: preparation, assignment, decision, and scheduling
- 3 timestamps to trust: ready, decision recorded, and window scheduled
- 2 views: median lead time and the share above the tier threshold
- 1 rule: report the median and the tail, never the average alone
Key takeaways
- Start the clock when the change note and evidence are complete, not when the idea first appears.
- Separate preparation, assignment, decision, and scheduling so the largest wait is visible.
- Match the approval path to a small number of risk tiers instead of adding meetings to every change.
- Include the reviewer's overlap window and queue depth when explaining a delay.
- Report the median and the share above a threshold, because one parked change can distort an average.
Question and scope
This report examines change advisory lead time for maintenance windows staffed by an offshore pod. The unit is one proposed change from the moment it is ready for review to the moment a named approver authorizes the window. The scope covers the advisory steps that sit between a finished change plan and an approved maintenance slot. It excludes the build work and the execution of the change itself.
The question is whether the advisory path adds avoidable delay or appropriate control. A short lead time can mean a well-prepared team, or it can mean nobody reviewed the risk. The report decomposes the path so a buyer can see where a change waits and which wait is worth keeping. It treats lead time as a coordination measure, not a speed target.
Start and end of the clock
Choose a start the delivery team controls, such as the moment the change note and rollback evidence are complete. A start at the customer request date mixes discovery and approval, which makes the result hard to act on. The start should mean the pod has done its part and is waiting on a decision. Write the start rule in the change template so a submitter cannot choose a later date by accident.
Choose an end that reflects real authority, such as the approval recorded in the change system. A verbal yes in a meeting is not the end unless it is written down. If the approval only becomes usable when the window is scheduled, record both times and report them together. The end rule should also say who records the approval when the approver is unavailable.
The four segments
Split the advisory path into preparation, review assignment, reviewer decision, and scheduling. Preparation is the time to complete the note and evidence. Review assignment is the wait for a qualified reviewer. Reviewer decision is the time that reviewer spends before approving or returning the change. Scheduling is the wait for a window that fits the risk, the approver, and the calendar.
Report each segment as a share of the total. A ten-day path with one day of review and seven days of scheduling points to a different fix than a path with six days of reviewer silence. The share shows where the coordination cost sits, while the absolute number shows how much room there is to change it. The share is more useful than the total when the team compares changes of different risk.
Check that the segments add up to the total. A gap usually means a step nobody named, such as a security sign-off that happens outside the change tool. Reconcile the gap before using the number for a decision. An incomplete decomposition can send the team toward the wrong bottleneck. A missing segment is a finding in itself because it means the process has an unmeasured step.
Risk tiers and proportionality
A routine configuration change and a schema migration should not follow the same approval path. Use a small number of risk tiers and define what each tier requires. The tier should raise the evidence and the approver level, not simply add waiting. A tier that only adds a meeting teaches the team to avoid it. The tier definitions should name the evidence required, not only the title of the approver.
Review the tier assignment when a change is returned or when an incident follows an approval. A change that needed more review than its tier allowed is evidence that the criteria are too loose. A change that waited for an approver who added no value is evidence that the tier is too broad. Review the tier definitions at the same time as the approval data so the labels stay meaningful.
Coordination across time zones
Every handoff between the pod and the buyer adds a waiting period that is bounded by the overlap window. A reviewer who is available for two hours per day will approve fewer changes than one who shares a longer overlap, even when both work at the same speed. Record the overlap and the reviewer's queue depth beside the lead time. The overlap should be recorded in the reviewer roster so the pod can plan around it.
Batch approvals can reduce the number of handoffs, but they also increase the time each change waits for the next batch. A daily review slot may be better than an ad hoc request that depends on catching someone online. Test the schedule with the actual reviewers before changing the process. The test should include a change submitted after the cut-off to see how the schedule handles it.
Measurement pitfalls
An average hides the changes that wait for weeks because a single approver is on leave. Report the median and the share above a threshold, such as three business days. The tail is where the coordination failure usually lives. Segment by risk tier so a rare migration does not distort the routine path. The tail view should name the changes that waited longest so the team can inspect the cause.
A change can look fast if the clock starts late. Moving the start to the review request excludes the preparation wait, which may be the largest part. Keep the definition stable and publish it beside the chart. A trend built on a changed definition is not comparable with the old one. Publish the definition beside the trend so a reader can compare periods without guessing.
A practical protocol
Instrument four timestamps: ready for review, reviewer assigned, decision recorded, and window scheduled. Add the risk tier, the reviewer, and the overlap window. Collect one quarter before comparing teams or periods. Check a sample of changes against the tool history to confirm the timestamps are real. The sample should include changes that were returned, because a return adds a segment the timestamps may not capture.
Pick the largest segment and change one thing. If assignment is the bottleneck, publish a reviewer roster for each tier. If scheduling is the bottleneck, reserve a weekly window for routine changes. Measure the same segments for a month and record the result with the change date. Record the change date beside the result so the next review can separate the fix from other events.
Review the result with the approvers as well as the pod. The approvers can explain a delay that the data only hints at, such as a missing security review or a change that was not ready. Their account should lead to a process fix rather than a request to respond faster. The conversation should produce one process change with an owner and a review date.
A worked example
A pod submits twelve maintenance changes in a quarter. The median advisory lead time is eight business days. The decomposition shows one day of review and six days of scheduling because the only approved window is a Saturday night that the buyer's database administrator must staff. The Saturday constraint is a capacity fact, not a failure of the pod, and the review should say so.
The team adds a Tuesday window for low-risk changes and keeps the Saturday slot for migrations. The median falls to three days, while the review segment stays the same. The change adds capacity for routine work without weakening the approval step that matters for risky changes. The result should be compared over the same quarter length before the team treats it as stable.
Data to collect first
Record the change identifier, the risk tier, the ready time, the assigned time, the decision time, the scheduled window, and the approver. Add whether the change was returned and the reason. Keep the raw timestamps so a reviewer can recheck the segments rather than trusting a summary. The record should also note whether the change was returned to the submitter and how long that took.
Collect at least a quarter, because maintenance windows are often monthly and a single cycle is too small to show a pattern. Mark periods when the approval policy changed. If the tool does not record assignment, use the first reviewer comment as a proxy and state the limitation. A quarter with a policy change should be split rather than averaged into one number.
Counterevidence and limits
A long advisory lead time can be appropriate for a high-risk change that needs a security review and a rollback rehearsal. The measure should not push a team to skip evidence. Compare the path with the risk tier and the incident history before calling the delay waste. The review should separate a deliberate delay from an unowned queue before proposing a fix.
Lead time also depends on the number of qualified approvers, which may be outside the pod's control. A buyer with one database administrator will have a different path from one with a team. Report the constraint honestly and treat the result as a prompt to inspect the process, not as a score for the pod. The result should be reported with that constraint so the reader does not read it as a pod score.
Evidence table
| Signal | What to inspect | Owner |
|---|---|---|
| Outcome | Acceptance evidence for the bounded task | Task reviewer |
| Control | Access, test, and approval boundary | Internal owner |
| Handoff | Open risks and next decision | Next owner |
Good distributed work is observable at the handoff: the result, evidence, limitations, and next owner are explicit.
Frequently asked questions
Should advisory lead time include weekends?
Use business time for the decision segments and calendar time for the scheduled window. State which clock each segment uses so the total is interpretable.
Is a shorter advisory path always better?
No. A high-risk change may need a security review and a rehearsal. Judge the path against the risk tier and the incident history.
What if the buyer has only one approver?
Record the constraint and a backup plan for leave. A single approver is a capacity limit, not a reason to remove the approval.
