Research question and scope

This protocol measures the share of open items in a delegated queue that cannot move without a written decision from someone outside the queue. The unit is one open item at a fixed observation time, identified by its record and its current state. The study measures waiting caused by a decision gap, not the total age of the item. An item that is old because the customer has not replied is outside the decision measure.

Set the scope to one client, one queue, and one observation window. Record the queue definition, the decision owners, and the system used to store the records. If the queue changes during the window, keep the original definition for the full sample or treat the change as a separate period. A new approval rule can move the share without improving the work.

Keep the observation separate from a service review. The purpose is to describe the queue at one moment, not to grade the team or the approver. A neutral description makes it easier for the owner to accept the finding and choose a change. State the observation time in the report so the snapshot cannot be mistaken for a trend.

Name the system of record before collection so every reviewer looks at the same place. If two systems can hold the state, choose one as the source and note the other as context. A review that mixes sources will produce a share that nobody can reproduce. The system name belongs in the report header.

Population and unit of analysis

Define the population as every item that was open at the observation time, including items that arrived before the window. The unit is the item, not the message, because one item can generate many messages and one message can cover several items. A reopened item remains one unit and receives a second observation. Record both observations if the item is reopened during the window.

Exclude items that are closed but awaiting an administrative step, such as a receipt filed after the customer was served. Record the exclusion count and the reason. Include items that are waiting for a decision even if the decision has already been requested. The measure should capture the wait as it exists, not only the wait that was formally logged.

Record the priority of each item so a high-priority wait is not hidden inside a low-priority average. A queue can have a modest overall share while its urgent items wait the longest. Report the priority mix and the decision share for each level. The owner needs to see whether the delay touches the work that matters most.

Record whether the item has a customer-facing deadline. An item with a deadline and a decision wait is more urgent than one without. Report the share separately for items with and without a deadline. The owner can then see whether the decision gap touches promises the customer is tracking.

Definitions that make the measure reproducible

Define a written decision as a recorded approval, denial, or instruction that names the action and the condition. A verbal yes in a hallway does not count unless someone records it in the system. Define the decision owner as the single person with authority to decide, not the group that receives the request. A clear definition prevents the share from depending on who is counting.

Separate a decision wait from an evidence wait by asking what is missing. If the approver has everything needed and has not decided, that is a decision wait. If the packet is incomplete, that is an evidence wait even if a decision was requested. Apply the same rule to every item and publish one boundary example. A second reviewer should reach the same classification without asking the first.

Include a boundary case in the definitions, such as an item with a verbal approval that was never written down. Decide before collection whether that case counts as a decision wait or as an evidence gap. Apply the same choice to every record and write the choice into the report. A clear boundary prevents two analysts from producing different shares.

Write the definition of a decision request as a recorded ask, not a thought. If the specialist mentioned the issue in a meeting but never asked the owner, the item is not yet waiting on a written decision. That distinction keeps the measure honest. It also shows how much work is stuck before anyone asks.

Sampling and data dictionary

Draw the sample from the queue export at a fixed time rather than from a list of complaints. Record the export time, the filters, the row count, and the selection rule. Include items from every owner, every priority, and every work type so the result is not dominated by one part of the queue. If the population is small, observe every item instead of sampling.

For each item, capture the record identifier, the current state, the age, the decision type, the owner, and the date the decision was requested. Add the last message and the last system event so the reviewer can verify the state. Store the export with the report so the observation can be repeated. A number without the underlying records cannot be checked.

Record the time the decision was requested, not only the time the item entered the queue. The difference between those times is the part of the wait that the team controls. If requests are batched to the end of the day, the queue will look slow even when the approver responds quickly. The timestamp exposes that pattern.

Include a field for the decision owner's last response time so the wait can be split into request and response. A long wait may be caused by a late request rather than a slow approver. The split tells the owner where to intervene. Without it, the team may add reminders that do not address the real delay.

Measuring the waiting share

Report the share of open items that are waiting on a written decision at the observation time. Also report the median wait for those items and the slowest decile. A single share can look modest while a small group of items waits for weeks, and those items often carry the most customer risk. Show both the typical case and the tail.

Break the share down by decision type and by owner. A high share in one type may mean the policy is unclear, while a high share in one owner may mean the approval load is concentrated. Do not compare decision types with different risk without saying so. The reader should be able to see where a fix would have the largest effect.

Show the distribution of waits as a small table with the median, the slowest decile, and the count. A dashboard that shows only an average will hide the cases that damage trust. Include the number of items that have waited longer than the lane threshold. The reader should be able to see the tail without opening the raw data.

Show the number of items that changed decision owner during the wait. A reassigned decision often restarts the clock and hides the original delay. Report the count and the median wait after reassignment. Frequent reassignment is a sign that the authority is unclear.

Analysis, limitations, and next action

The record cannot show how often a specialist worked around a missing decision without telling anyone. It also cannot show the cost of a delay that the customer never noticed because the team managed expectations. Treat the share as a lower bound on decision waiting, not as the full customer experience. Note any workaround reported by the team.

Give the result to the queue owner and the decision owners together and ask for one change. The change might be a named backup, a shorter approval form, or a rule that small decisions do not need a second approver. Record the effective date and the measure that will show whether the change worked. Repeat the same observation after the change and compare the same categories.

Repeat the observation at the same time of day and the same day of week if possible. A Monday morning snapshot and a Friday afternoon snapshot may not be comparable. Note any seasonal or staffing event that could affect the result. The second observation is what turns a one-time count into an operating measure.

Agree on the next observation date before the meeting ends. A measure that is not repeated becomes a one-time story. Put the date in the same calendar as the operations review. The second observation should use the same definitions and the same system of record.