Research question and scope

This protocol asks how often a client changes the priority of work in a shared queue and how much previously planned work is displaced by that change. The unit is one priority-change event that affects at least one item in the queue. The study measures the frequency, the source, the notice given, and the work that moved as a result. It does not judge whether the new priority was correct, because that judgment belongs to the client owner.

Set the scope to one queue, one client, and one planning cycle, and record the intake process and the prioritization rule in force at the start. If the client changes its planning cadence during the window, keep the earlier definition for the full sample or analyze the change separately. A new weekly meeting can look like more churn when it is only a new rhythm. Record the meeting cadence beside the event data.

Population and unit of analysis

Define the population as all priority-change requests that reached the queue during the window, including those that were rejected or deferred. The unit is the change event, identified by the request date, the affected item, and the person who requested it. A single message that changes five items is one event with five affected items. A follow-up that repeats the same request without new information is not a new unit.

Record the requester's role, the channel used, the notice given, and whether the change affected a hard deadline. These strata matter because a change from the client owner with a week of notice is different from a change from an end user the day before delivery. Include changes that were reversed within the same cycle, because reversal is a distinct cost. Test the unit definition on two ordinary cases before collection begins.

Definitions that make the measure reproducible

Define a priority change as an instruction that alters the order, the deadline, or the scope of an item already accepted into the queue. Define displacement as the number of planned hours or items that moved to a later position because of the change. Define notice as the time between the change request and the original deadline of the affected item. Define reversal as a later instruction that restores the earlier priority.

Do not count a clarification that leaves the priority unchanged. Do not count a new request that was never accepted into the queue, although it should be recorded separately. A change that only adds information without moving the order is not a priority change under this definition. Each definition should include one positive example and one boundary example so a second coder can apply it consistently.

A priority change that arrives with no reason should be coded as unexplained rather than assigned to a guessed category. The reason field is often the weakest part of the record, and a guessed category can make the analysis look more precise than the data. Ask the requester for one sentence of context at the time of the change. That sentence is usually enough to separate a customer escalation from a preference.

Evidence standard and data dictionary

For each event, capture the request date, the requester, the channel, the affected item or items, the original deadline, the new deadline, and the reason given. Add the planned effort, the actual effort, and the item that was displaced. Record whether the change was approved by the client owner and whether the queue owner confirmed the new order. Code each field as present, missing, or not applicable.

Also record whether the change arrived through the approved intake channel or through a side conversation. Side-channel changes are a useful signal because they bypass the queue's normal tradeoff discussion. If the reason is vague, record the exact words rather than an interpretation. A second coder should be able to classify the event from the packet without asking the queue owner what happened.

Sampling and collection

Draw the sample from the queue system and the client communication log rather than from a list of memorable escalations. Record the export date, the filters, the row count, and the selection rule. Include changes that were accepted, rejected, and reversed so the analysis can show the full cost. If the population is ordered by request date, a fixed interval spreads the sample across the window.

Collect the evidence from the systems that existed at the time of the change. Do not rebuild a missing approval from memory, because reconstruction measures the investigator rather than the process. If a message thread has expired, record the expiry and preserve the metadata. Note the collection date separately from the request date so the reader can see how much of the record has aged.

Analysis and reporting

Report the number of priority-change events per week, the share that arrived through the approved channel, and the median notice before the original deadline. Show the displaced effort in hours and in items, and present the counts by requester role and by queue type. Include the number of events with a missing reason so the reader can see the size of the gap. Use the same denominator for every rate.

Inspect the events that displaced the most work and describe the pattern without naming individuals. A cluster of changes from one requester may point to a planning gap, while a cluster near the deadline may point to an intake problem. Compare the cost of the change with the value the client described, but do not pretend the protocol can measure value. Report the reversal rate separately because a reversed change often costs twice.

Show a weekly trend rather than only a quarterly average, because churn often clusters around planning meetings or month-end. A spike that follows a recurring meeting may be a predictable process cost rather than a random disruption. Compare the trend with the client's calendar and with the queue's own release schedule. That comparison can turn a vague complaint about constant changes into a specific timing problem.

Limitations and sources of error

The record cannot show whether the original priority was wrong or whether the new one was better. It can show only how often the order changed and what moved as a result. A high churn rate is not proof of a bad client, and a low rate is not proof of good planning. The protocol also cannot capture the morale cost of repeated changes unless the team records it separately.

Queue exports can omit verbal instructions, side conversations, and changes that were never logged. A change that was absorbed without moving any item may not appear in the sample. Record those cases and their treatment instead of dropping them silently. If the queue owner and the requester are the same person, note the independence limitation and analyze that subgroup separately.

The protocol also cannot measure the cost of context switching, which may be larger than the displaced hours suggest. A specialist who returns to an interrupted item often needs time to rebuild the state, and that time is rarely recorded. Add a field for the number of days an item was paused, even if the effort is not measured. The pause duration is a useful proxy for the disruption that the queue actually experienced.

Review cadence and next action

Give the result to the client owner and the queue owner and ask them to choose one observable change. The change might be a single intake channel, a weekly priority freeze, or a rule that late changes require an explicit tradeoff. Record the owner, the effective date, and the measure that will show whether the change worked. Avoid changing three things at once if the team needs to learn which one mattered.

Repeat the protocol with the same definitions after the change and compare the same fields. If churn fell but displacement did not, inspect whether the remaining changes are larger or later. If side-channel changes fell but the total did not, the approved channel may simply be capturing what already happened. Close the review when the owner accepts the finding and the next check is scheduled.

If the client agrees to a priority freeze, define the window and the exception path before the first cycle. A freeze without an exception path will be bypassed by the first urgent request and will lose credibility. Record the exceptions and review them at the end of the freeze. The point is to make the tradeoff visible, not to prevent the client from responding to a real emergency.