Research question and scope

This protocol measures how often a delegated queue pauses without a recorded outage notice. The unit is one pause event for one queue. An unrecorded pause means work stopped or slowed for a system reason and the queue record shows no notice, start time, or scope. The measure describes the visibility of downtime, not the reliability of the platform.

Set the scope to one queue, one client, and a fixed period. Record the systems the queue depends on and the definition of a pause in force. If the definition changes during the period, split the sample at the effective date. A different threshold creates a different denominator.

Keep the measure separate from a missed service level. A recorded pause can still cause a late reply, and an unrecorded pause may have no customer impact. The purpose is to test whether the desk knows when it stopped. Report service performance separately.

State the decision the result will inform before collection begins. The client may use it to justify a status page, a queue banner, or a faster escalation to the system owner. A measure without a decision tends to become an unused report. Write the intended action beside the scope.

Name the report audience and the review date before collection begins. A report for the platform owner may need different detail from a report for the client service lead. Record the audience and the meeting where the result will be discussed. A measure with no scheduled review is easy to postpone indefinitely.

Population and unit of analysis

Define the population as every pause that affected the queue during the period. Include pauses called by the team, pauses announced by the platform, and pauses inferred from a gap in completions. The unit is the pause event, not the ticket or the hour. One outage that stops three task types is one event when it shares a start time and cause.

Include partial pauses where some tasks continued. Mark the stopped and continuing tasks on the record. Exclude a scheduled maintenance window that was announced in advance and respected by the queue. Report the excluded count so the reader can see the desk's share of unexpected downtime.

Exclude a pause caused only by a single specialist stepping away. That is a staffing event, not a system event. Record it in a separate staffing log if the client tracks coverage. State the exclusion rule before sampling so a second reviewer can apply it the same way.

Decide how to treat a pause that began before the period and continued into it. Count it at the start date if the start is visible, and mark it as carried in. A pause that began without a record is itself a finding. Record the rule in the report so the boundary is clear.

List every queue that shares the same platform before drawing the population. A platform outage can stop three queues at once, and each queue may record it differently. Record the shared dependency and whether the queues are counted together or apart. The choice changes the denominator and must be made once.

Definitions that make the measure reproducible

Define a recorded pause as a notice with a start time, a scope, a cause, an owner, and a resume decision. Define a partial record as one missing element. Define an unrecorded pause as no notice at all, even when the team later described it in chat. Publish one example of each before sampling.

Define the pause threshold in measurable terms. A queue may count as paused when no task can be completed for fifteen minutes, or when a named system returns an error on save. Avoid a vague rule such as unusually slow. A threshold that two reviewers interpret differently will not reproduce.

Define the resume check as a test transaction completed by a named person after the system owner reports recovery. A resume without a test is a hope, not a control. Record the test reference and the time. The gap between recovery notice and resume test is part of the result.

Decide how to classify a pause with an unknown cause. Keep it as unknown and report it separately rather than assigning a likely reason. An invented cause hides the need for better monitoring. The unknown group is often the most useful part of the report.

Agree how a pause ends before you count duration. A pause may end when the system owner reports recovery, when the first test passes, or when the queue reopens. Record the end definition beside the start definition. Using different ends for different systems makes the duration comparison meaningless.

Sampling and data collection

Draw the sample from the queue export, the platform status feed, and the team chat for the full period. Record the export date, the filters, and the row count. Do not build the sample from customer complaints alone. A complaint-based sample cannot show the pauses that customers never noticed.

Capture the queue, the start and end time, the cause, the scope, the notice channel, the owner, the resume test, and the customer-facing message. Store the exports with the report. The notice time and the resume test are the core evidence. A record missing either field counts as partial even when the chat thread is complete.

Check a small overlap sample with a second reviewer. If the reviewers disagree about whether a chat message counts as a notice, the definition is too vague. Rewrite the rule and retest before the full sample. Agreement matters more than sample size.

Record the time spent reconciling the queue export with the status feed. A report that requires reading every chat thread will not be repeated. Note the systems that do not keep a status history. The collection cost belongs beside the result.

Keep a small pilot sample of ten pause events before the full collection. Use the pilot to test the definitions and the export fields. Record what was missing and how long each record took. A pilot that takes an hour per event is a warning about the full sample cost.

Measuring unrecorded pauses

Report the share of pauses with a complete record, a partial record, and no record. Show the counts behind each share. The three groups describe different levels of visibility and should not be collapsed. A single unrecorded rate hides the pause that had a start time but no resume test.

Break the result down by system, shift, and task type. A platform with no status feed may show more unrecorded pauses. A night shift may rely on chat because the system owner is offline. Compare only groups with similar volume and dependency.

Report the median time from the first failed task to the first notice. A long gap means the team is discovering outages through customer reports. Include the median time from notice to resume test. Both numbers describe the response, not only the record.

Report the share of pauses that produced a customer-facing message. A pause can be fully recorded internally and still leave customers without an explanation. Include the message channel and the approval owner. The external silence is a separate risk.

Compare pauses that were announced before the first customer complaint with those announced after. The comparison shows whether the desk is monitoring or reacting. Record the source of the first signal. A team that learns from customers is not monitoring its own queue.

Limitations and next action

The measure cannot show whether a pause was necessary or whether work could have continued safely. It also cannot show a platform outage that the team worked around without stopping. Treat the result as a measure of recorded visibility. Note the systems and shifts that were not included.

Give the result to the service owner and the platform owner with one change to test. The change might be a queue banner, a required pause form, or a status-feed subscription. Record the effective date and repeat the measure with the same definitions. A controlled change is easier to judge.

Review the boundary cases before widening the threshold. A few short pauses with a clear chat notice may explain a large share of partial records. Fix the notice template first if that is the pattern. The goal is a queue that knows when it stopped, not a lower number by definition.

Repeat the measure after the change with the same scope and definitions. Report the before and after counts together, including the sample size. Record any new system or queue that entered the period. A comparison with a changed denominator is not a comparison.