Question and scope
This protocol measures the share of cross-team requests that change scope after work has started. A scope change is any addition, removal, or material redefinition of the expected deliverable after the receiving team begins work. The result is a percentage, a median count of changes per request, and a breakdown by request type and sending team. It describes changes that were visible in the request record, not every informal shift in what a requester hoped to receive.
Scope is one coordination team and the requests it sent or received during a fixed window. Include requests that crossed a team boundary in either direction, since a team that changes scope often may also absorb changes from its own requesters. Exclude routine clarifications that leave the deliverable unchanged, such as a corrected file name. Record the trackers searched, the window dates, and the reviewers so the count can be rebuilt.
Why post-kickoff change is a coordination cost
Work that starts before the deliverable is settled consumes capacity twice. The team plans an estimate, reserves a slot, and begins, then has to re-plan when the request grows. Even a small addition can push a request past a weekly cutoff and into the next queue cycle. The visible cost is the extra hours, while the hidden cost is the disruption to every request scheduled behind it.
Repeated scope change also weakens the estimate itself. If requesters learn that an initial ask can be expanded later at no cost, they have little reason to think carefully at intake. The team then pads estimates or adds review steps to protect itself, which slows every request. Measuring the rate gives both sides a shared fact to discuss instead of trading complaints about vagueness.
Defining the coding unit
The unit is one cross-team request with one named requester, one receiving team, and one expected deliverable. Split a bundle of unrelated asks into separate requests, because each can change on its own schedule. Keep one request as one unit even when several people contribute to the output. Write the split rule into the codebook before sampling so two reviewers classify similar records the same way.
Use the request as it existed when work began, not the final version in the tracker. The final version already contains the changes, so starting there makes the change invisible. If the tracker overwrites the original description, pull the baseline from the kickoff message or the first assignment note. Note which source you used for every request, because a baseline from memory is weaker evidence than a baseline from a timestamped message.
Pinning the baseline
Record the baseline fields that matter for judging change: deliverable, format, quantity, deadline, and acceptance criteria. A change to any of these counts as a candidate, while a change to a spelling error or an internal label does not. Capture the baseline before reading later messages, so the later discussion cannot quietly redefine what counts as new. Two reviewers should agree on the baseline for a sample of ten requests before the full review begins.
Timestamp every baseline element you can. A request that says "a monthly report" without a format becomes ambiguous once someone adds a dashboard, because nobody can prove whether the dashboard was implied. A request that never states its acceptance criteria is under-specified, and the definition-of-done measure shows how common that gap is. Where the baseline is genuinely thin, code the request as under-specified and report it separately, since that category often points to an intake problem rather than a discipline problem.
Classifying the change
Sort each candidate change into one of four types: addition, removal, redefinition, or deadline shift. Addition adds a deliverable or recipient, removal takes one away, redefinition changes what the existing deliverable must contain, and a deadline shift moves the date without changing the work. Report the types separately, because a team that absorbs many deadline shifts faces a different problem from one that keeps adding outputs. A single request can carry more than one type, so allow multiple codes and count the request once in the headline rate.
Distinguish a change requested by the sender from one proposed by the receiver. A receiving team that suggests a better format is improving the request, while a sender who keeps adding recipients is expanding it. Code the source of each change from the message author and the surrounding thread. That split keeps the measure from blaming the receiving team for changes it did not request, and it shows where the pressure actually originates.
Sampling and evidence
Draw the sample from the tracker export rather than from memory or a recent chat search. A random sample of at least thirty requests gives a rate with a wide but usable margin, and a census of a small queue is better than a sample if the volume allows it. Stratify by sending team when one team dominates the queue, or the headline rate will mostly describe that team. Record the random seed or the selection rule so the sample can be reproduced.
For each request, read the baseline message, the kickoff note, the completion record, and every message that mentions a new requirement. Save the message identifiers and timestamps in the coding sheet. Do not rely on a summary written after the fact, because summaries tend to fold changes into the final scope. If the record is incomplete, mark the request as unverifiable instead of guessing, and report the unverifiable share alongside the rate.
Handling requests that span several systems
Pull the record from every system a request touched, because a baseline in one tool and a completion in another can hide a change. If the request began in email and finished in a tracker, treat the first tracked assignment as the kickoff only when the email already defined the deliverable. Otherwise, count the email as part of the baseline and note the split. A coding sheet should list every system consulted, not just the tracker.
When two systems disagree, prefer the message with the earliest timestamp for the baseline and the latest timestamp for the completion. Do not average the two or pick the more convenient version. Record the conflict in a note so a reviewer can see why the request was coded as it was. Conflicts are worth counting separately, since a high conflict rate means the team needs one authoritative record.
Calculating the rate
The headline rate is the number of requests with at least one post-kickoff change divided by the number of verifiable requests. Report the numerator, the denominator, and the unverifiable count next to the percentage. A rate without its denominator hides whether the finding rests on twelve requests or two hundred. Round to whole percentages and avoid false precision, since the coding decisions carry more uncertainty than a decimal place suggests.
Add a secondary measure for intensity: the median number of changes per changed request. A team with a low change rate and a few very large changes has a different problem from a team with many small changes. Show both measures by request type and by sending team. The pair prevents one large project from distorting the story and gives the coordinator a concrete place to start a conversation.
Reporting limits
This measure counts changes that left a trace in the request record. Changes agreed in a hallway, a phone call, or a direct message that was never logged will be missed, so the rate is a floor rather than a complete count. State that limitation in the report instead of presenting the number as the full picture. If the team suspects a large untracked channel, pair this protocol with an observation window that captures it.
The measure also says nothing about whether a change was justified. A client who genuinely needs a new compliance field is not behaving badly, and a scope change that prevents a costly error is good work. Avoid publishing the rate as a performance score for a person or a team. Use it to locate where intake is thin, where approval is unclear, and where a change should have triggered a new estimate.
Using the result without blame
Share the findings as a description of the system, with examples that are stripped of names. Start with the request type that changes most often and ask the people involved what they know about it. Often the cause is a missing template field, an approval that arrives late, or a requester who never sees the original brief. Fix the smallest cause first and re-measure after a quarter.
Pair the rate with a simple rule for handling changes, such as a new estimate whenever a change adds a deliverable or moves a deadline by more than two days. The request acceptance checklist gives the intake fields that make such a rule workable. Write the rule into the form so both sides see it at the same moment, then track whether changes arrive earlier and whether rework falls. The point of measuring scope change is a calmer queue, not a cleaner-looking chart.
Sources
- U.S. Government Accountability Office, Cost Estimating and Assessment Guide (GAO-20-195G), reviewed November 21, 2026.
- National Institute of Standards and Technology, Guide for Security-Focused Configuration Management of Information Systems (SP 800-128), reviewed November 21, 2026.
- U.S. Government Accountability Office, Standards for Internal Control in the Federal Government (GAO-14-704G), reviewed November 21, 2026.
- U.S. Small Business Administration, Manage your business, reviewed November 21, 2026.
