Question and scope

This protocol measures how often an escalation reaches a second person without the assigned owner being told. An escalation can look properly routed in the system while the original owner learns about it only after the decision. The measure counts skips and describes what the owner lost: context, authority, or the chance to respond. It does not judge whether skipping the owner was the right call in a particular incident.

Scope is one escalation path between two teams over a stated period, such as four consecutive weeks. Include escalations that were later resolved and escalations that were reopened after a skip. Exclude requests that were never assigned to an owner in the first place. Write the boundary in the report so another reviewer can repeat the count on exactly the same set.

Population and unit of analysis

The unit is one escalation record tied to one owner and one moment of transfer. An escalation that moves through three people counts as three transfers but only one record. Record the owner at the moment the transfer begins, not the owner listed after the work was reassigned. A stable unit prevents a busy incident with many handoffs from looking like many separate escalations.

Build the list from the escalation log rather than from a summary prepared for the review. Include records that were edited after the fact and keep the version that was visible during the transfer. Exclude duplicate records created by two systems for the same event, but report how many duplicates were removed. A clean denominator is the difference between a useful rate and a number nobody trusts.

Definitions and measures

A skip occurs when an escalation reaches another person before the assigned owner is notified or asked to act. Notification means a message that the owner can see, not a general channel post that nobody is required to read. A skip can be intentional, accidental, or caused by an approval rule that bypasses the owner. Count each cause separately because the repair is different for each one.

Report the share of escalations with at least one skip and the average number of skips per escalation. Break the result down by severity, team, and time of day, since overnight routing often behaves differently. Show raw counts beside every percentage so a small sample is not mistaken for a stable rate. A rate from six escalations should never be presented with the same confidence as a rate from sixty.

Evidence and data dictionary

Each record needs the escalation identifier, the assigned owner, the time of assignment, and the time the next person acted. Evidence of a skip is a message or action that shows the owner was not consulted before the transfer. A system field that says reassigned is not enough because it may describe a routine correction. Keep the log export and the coding sheet together so the result can be reconstructed.

Write a one-page dictionary that defines assigned owner, notification, skip, and legitimate bypass. Include the rule for a manager who takes over during an outage and the rule for an owner who is on leave. Show one example of each category so two reviewers can apply the same words. Mark unclear records as unresolved instead of forcing them into a category that does not fit.

Test design and sampling

Use a census when the escalation volume is small and a documented sample when it is large. The sample should include every severity level and at least one overnight period and one incident. Freeze the sample before coding so the reviewer cannot choose records that support a preferred conclusion. Have a second person review a subset and record disagreements before the group resolves them.

A practical window is four weeks or the most recent complete operating cycle. If the routing rules change during the window, split the analysis at the change and report both parts separately. A comparison across a rule change mixes two different processes and produces a misleading average. The measure is only meaningful when the definition of a skip stayed stable across the whole period.

Decision rules before the test

Decide before coding whether a manager who acts during an outage counts as a skip or a legitimate bypass. State whether a message sent to a shared inbox counts as notification to the owner. Decide how to treat an owner who was unavailable and had no recorded backup. These choices change the rate, so they must be written before the first record is read.

Agree on the action that follows a high skip rate. One option is to require an owner acknowledgement before an escalation can leave the queue. Another is to publish a short routing map that names the backup for each severity level. Avoid a general reminder to communicate better, because the gap usually comes from a rule rather than from individual intent.

Analysis and reporting

Report the share of escalations with a skip, the share with more than one skip, and the count of unresolved records. Show the oldest escalation in the sample that skipped its owner and the team where it sat. Pair the count with a short review of what the owner lost, because a skip is not equally serious in every context. The report should help a manager decide whether to change the rule or the roster.

Break the result down by severity and by hour of day, and show the number of records behind each figure. A comparison between a weekday afternoon and a weekend night is useful only when both periods have enough records. If the sample is small, report counts and describe the uncertainty in plain language. A useful report ends with the change the team will test next.

Reporting template

Use a one-page template with four blocks: scope, share with a skip, share with an unresolved owner, and the action agreed at review. Keep raw counts beside every percentage so the reader can judge the evidence behind it. Add a short note for each exclusion and each routing change during the period. The template should be short enough to complete in the review meeting.

Store the completed template with the export and the coding sheet in one shared folder. A later reviewer should be able to reproduce the main number without asking anyone for extra context or help. If several versions exist, mark the final one clearly and date each change. Evidence that cannot be found is treated as missing when the result is questioned.

Limits and sources of error

The cited public sources describe records, internal control, and privacy in general terms, not the routing behavior of one team. A recorded skip does not prove that the owner was harmed, and a missing record does not prove that no skip occurred. The measure also depends on how the team defines notification. A loose definition can count a channel post as a real handoff.

Edited logs and reconstructed records are common sources of error in this kind of test. Use the version that was visible during the transfer and mark any record that changed later. Exclude a record when the owner field cannot be verified, but report the exclusion count and the reason. A clear rule is better than a precise-looking number built on uncertain evidence.

Review cadence

Review the result at a fixed date and end with one change, one owner, and one check date. A repeated skip should change the routing rule or the backup roster before it changes coaching. Keep the export, the coding sheet, and the notes together for the next reviewer. Compare like periods and mark any routing change on the same timeline as the result.

Repeat the review after two full cycles, then move to a monthly rhythm if the measure is stable. Retire the measure if it stops informing a decision, because a metric nobody uses will be maintained badly. The purpose is an escalation that reaches the right person with context, not a perfect audit trail. Stop when the team can show that outcome consistently.

Sources