Research question and scope

This protocol measures how much of a delegated inbox is routine work and how much is exception work. The unit is one handled item, identified by the message, the record it touches, and the person who handled it. Exception work is any item that cannot be completed with the written routine and needs a decision, a search, or a new rule. The study measures work mix, not individual productivity.

Set the scope to one client, one inbox, and one shift pattern for a fixed window. Record the routine guide version, the queue rules, and the tools used when collection starts. If the guide changes during the window, keep the earlier definition for the full sample. A new guide can move items from exception to routine without changing the underlying demand.

Population and unit of analysis

Define the population as every item handled in the inbox during the window. The unit is the item, not the thread, because one thread can contain a request and a later correction. An item that is reopened remains one unit with a second handling event. Record both events so the analysis does not hide the repeat.

Exclude spam and clearly personal messages, and record the exclusion count. Include items that were forwarded to another team, because the inbox still spent time classifying them. Include items that were answered with a saved reply, because routine work is part of the mix. Test the unit definition on ten ordinary items before collection begins.

Classifying routine and exception work

Define routine work as an item that matches a written rule and can be completed without a new decision. Define exception work as an item that needs a policy decision, a missing record, a vendor contact, or a customer promise outside the guide. Add a third category for unclear items that cannot be classified without more information. Publish one positive and one boundary example for each category.

Do not classify by how long the item took. A routine item can be slow because the system is slow, and an exception can be fast because the owner has seen it before. Classify by what the work required, then measure time separately. That separation is what makes the work mix comparable across shifts.

Evidence standard and data dictionary

For each item, capture the message, sender, record identifier, routine guide section, classification, handling time, and outcome. Add the person who handled it and whether they asked for help. Record the reason for the exception in a short code. Store the packet with the inbox record so the evidence survives a later review.

Also capture whether the item returned within seven days. A routine answer that creates a second question may be a hidden exception. Record the customer's next message and the reason it arrived. If the reason code is other, require a note. A reviewer should be able to classify the item from the packet without asking the handler.

Sampling and collection

Draw the sample from the inbox export rather than from a list of difficult cases. Record the export date, filters, row count, and selection rule. Include every weekday and every handler so the result is not dominated by one busy Monday. If the inbox is ordered by time, a fixed interval spreads the sample across the window.

Collect the message as it existed when it arrived, not as it looks after the reply. If the customer added detail later, record the edit time and analyze the first version separately. Do not reconstruct a missing field from the final answer. Note the collection date separately so the reader can see how much of the record has aged.

Measuring time by work type

Report the median handling time for routine, exception, and unclear items separately. Use the same clock and include the time spent searching, asking, and waiting. A routine item that takes twenty minutes may be a guide problem, not a slow person. Show the slowest decile as well as the median so the tail is visible.

Break the time down by handler and by work type within the same week. A handler with a high exception share will look slower even if their routine work is fast. Do not compare raw averages across handlers without adjusting for the mix. The mix is the finding, not a nuisance to control away.

Analysis and reporting

Report the number of items, the share in each category, and the median time by category. Show the exception reasons in a small table and rank them by count and by time. Use the same denominator for every rate and state the exclusions. A percentage without its base is not actionable for the queue owner.

Compare the exception share across weeks and shifts. A rising share may mean the guide is missing a case, the product changed, or the team is receiving a new request type. A falling share may mean the guide improved or the team is classifying difficult work as routine. Check a sample of classifications before trusting the trend.

Limitations and sources of error

The record cannot show work that was handled outside the inbox, such as a phone call or a direct message. It also cannot show the time spent worrying about an item after it was closed. A handler may classify an item as routine to avoid asking for help. The study measures recorded classification and recorded time.

Messages can be deleted or moved, and a thread may span several systems with different clocks. A verbal decision may never appear in the record even though the work moved. Record those cases and their treatment rather than dropping them. If the person who classified the item also set the time target, note the independence limitation.

Review cadence and next action

Give the result to the queue owner and the training lead and ask for one observable change. The change might be a new guide section, a saved reply, or a rule that a certain exception no longer needs a manager. Record the owner, the effective date, and the measure that will show whether the change worked. Avoid changing the guide and the classification rule at the same time.

Repeat the protocol with the same definitions after the change and compare the same categories. If the exception share fell but handling time rose, the new guide may be adding work without removing decisions. If only one handler improved, the change may not be the cause. Close the review when the owner accepts the finding and the next sample is scheduled.

Queue design implications

A high exception share is not automatically a staffing problem. It may mean the inbox is collecting work that belongs in a different queue. Split the inbox by request type and route the exceptions to a lane with the right decision owner. A specialist who handles routine work and exceptions at the same time will switch context and slow both.

Use the exception reasons to decide what to automate or document first. A missing record that appears in many items is a data problem. A policy question that repeats is a guide problem. A vendor delay is a follow-up problem. Each cause points to a different owner, and the work mix shows which one deserves the next improvement.

Reporting to the client

Report the work mix in a form the client can act on: which request types create exceptions, which reasons dominate, and what the team does when the guide runs out. Avoid presenting the numbers as a grade for individual specialists. Use examples with identifying details removed. The client should leave with one change it can approve.

Show the denominator and the exclusions so the client can judge the sample. If the sample is too small to support a conclusion, say so. A cautious report is more useful than a confident one built on a handful of items. Offer to repeat the measurement after the change and name the date.

Replication and version control

Keep the definitions, the guide version, and the export query with the report. A second analyst should be able to repeat the measurement on a later window and compare the same categories. If the definition changes, publish a new baseline rather than mixing periods. Replication is what turns a one-time review into an operating measure.

Schedule the next measurement before the first report is filed. The date should be far enough away for the change to take effect but close enough to keep attention. Assign the owner who will run it. A protocol that is never repeated becomes a story about the past. Put the date in the same calendar as the queue review so it is not forgotten.