Question and scope

This protocol measures the elapsed working time between the moment a cross-team commitment is missed and the moment someone with authority to act is first told about it. The result is a median delay, a distribution, and a count of misses that were never escalated at all. It is not a measure of how quickly a problem was solved. It measures how long the organization allowed a known problem to remain private.

The distinction matters because teams often report escalation speed from the moment a manager hears about the issue. That clock starts too late. The useful delay begins when the promised date passes without delivery, regardless of whether anyone has labeled it a miss. A commitment can be three days late while the owner waits for more information, and the organization has already lost those three days.

Scope is one coordination team and commitments made to or by it during a fixed window, as recorded in the commitment log. Include commitments that were eventually delivered late. Exclude deliverables that were formally descoped before the original date. Record the systems searched and the review window so another analyst can reproduce the population. Include the commitment log export date in the file name.

Defining the missed commitment

A commitment is missed when its promised date passes without delivery or an accepted change to the date. The promised date must be a specific calendar date, not a range or a relative phrase such as "next sprint." If the original promise was vague, mark the commitment as unmeasurable and report it separately. A vague promise cannot have a clean miss timestamp. Record that decision in the codebook before scoring begins.

Use the end of the business day in the team's operating timezone as the miss moment when the promise names only a date. If the promise names a time, use that time. Record the timezone and the working calendar because a two-hour difference can move a Friday miss into the following Monday. Do not use the moment the owner admits the miss, since that is exactly the delay being measured.

Count a delivered item as late only when it arrived after the promised date and without a prior accepted change. A delivery on the promised date is not a miss even if the review happens later. A delivery after an accepted new date is a miss against the original promise but not against the revised one. Report both views if the team commonly renegotiates dates.

Defining the first escalation

The first escalation is the earliest recorded contact with a person who can change the outcome. That person may be the receiving team's manager, the requester's manager, or a coordinator with authority to reassign work. A message to a peer who must ask someone else does not count unless that peer is the decision-maker. Record the role and the authority basis, not just the name.

Accept as evidence a message, ticket comment, meeting note, or call log that states the commitment is late and asks for a decision or action. A status update that mentions the delay without asking anyone to act is weaker evidence and should be flagged. A calendar invite for a rescue meeting counts if it names the late item and includes the decision-maker. The test is whether the record shows that someone with authority was informed.

Do not count a routine weekly report that lists all open items, even if the late item appears in it. A list entry without a specific request for action is visibility, not escalation. The protocol should separate broad reporting from targeted escalation because the two serve different purposes. Report the share of misses that received only broad visibility. That distinction should be explained in the report.

Building the inventory

Start with the commitment log and add promises that live only in email, chat, or meeting notes. Search for language that names a deliverable and a date, such as "we will send," "by Friday," or "committed for." Check the cross-team commitment log for items that were promised but never entered into the tracker. The inventory should reflect the work the team actually promised, not only the work that followed the formal path. The search should include the coordinator's own mailbox and direct messages.

Record the requester, receiving team, owner, deliverable, promised date, and source system. Mark whether the promise was recorded in a shared system or only in a personal channel, because the latter is harder to monitor and easier to forget. Assign a stable identifier to each commitment. Freeze the inventory before reading for escalation evidence. Those four fields are enough to locate the item later.

Check for duplicate records created when a ticket and an email describe the same promise. Count the pair once and note the overlap. Treat a reopened item as a new commitment only when the deliverable or date changed; otherwise keep the original promise and note the reopening. Reopened items with new dates are a common source of artificially short delays.

Reading the escalation evidence

Search the owner's and coordinator's channels for messages about the late item, beginning at the miss moment. Record the earliest qualifying escalation and the channel where it appeared. If the first contact is a phone call, use the call log or the follow-up message that confirms the conversation. The record must be specific enough that a second reviewer can find it.

Read the message for the decision it requests, not just the problem it describes. "The vendor report is late" is a report. "The vendor report is late and we need you to approve a two-day extension or reassign the reviewer today" is an escalation. Record the requested decision alongside the timestamp. That detail explains why some delays end quickly and others continue.

If the first escalation went to the wrong person, record it and continue searching for the first contact with someone who could act. A message to a colleague who was on leave is not an effective escalation, even if it was sent promptly. The protocol measures when authority was reached, not when concern was first expressed. Record the wrong recipient as a separate field.

Handling reschedules and partial delivery

Record every accepted change to the promised date, not just the final date. A promise that moved twice before delivery has a different escalation profile from one that moved once and was met. Capture the reason when the record states one, and group reasons into categories such as capacity, dependency, or scope change. The categories explain more than a raw count of moves.

Treat a partial delivery as a miss for the undelivered portion only when the remaining portion has an independent acceptance condition. If the whole deliverable must be reviewed together, a partial handoff is not a completed delivery. Note the interpretation in the codebook. Mixed cases are where reviewers most often disagree, so resolve them before coding the full sample. State the split rule in the codebook.

Exclude commitments that were cancelled before the promised date with a recorded decision. A cancellation is not a miss, though it may still be worth counting separately as a planning change. Do not exclude a commitment simply because the owner left the team; that is a miss with a staffing cause and belongs in the population. A staffing gap is still a missed promise from the requester's view.

Sampling and verification

Use a census when the team has fewer than roughly one hundred commitments in the window. Above that threshold, draw a random sample stratified by receiving team and by month. Freeze the sample before reading escalation channels. Store the random seed and the selection method with the result so the next analyst can repeat it. A sample drawn by hand is not a random sample.

Have a second reviewer recode at least ten commitments, including several with no escalation. Compare the miss timestamps and the escalation timestamps separately, because disagreement often comes from one definition rather than both. Resolve disputes by returning to the record, not by choosing the more plausible date. Record the resolution in a short codebook log. Two reviewers should agree before the number is published.

Check timezone and working-day calculations before reporting. Two systems can record the same message on different dates when one stores UTC and the other stores local time. Count working days using one calendar and state which holidays were excluded. A delay reported in calendar days without that note will not be comparable to the next period. Keep the working calendar with the result.

Calculating and reporting the delay

Report the median delay in working hours or working days, the interquartile range, and the share of misses with no escalation at all. Show raw counts beside every percentage. A median alone hides the tail, and the tail is where the commitments that damage trust usually sit. Add the share of misses escalated only after the requester asked for a status update.

Break the result down by receiving team, owner, and commitment type. Suppress groups with fewer than five commitments so one unusual case does not look like a pattern. Pair the delay with the size of the deliverable, because a two-week silence on a one-hour task is a different problem from a two-week silence on a two-week project. The weekly blocked-work review is a natural place to act on the first case.

Report the trend rather than a single period. Delays move with holidays, staff changes, and large launches, so one month is rarely enough to judge. Show at least three periods, or state clearly that the result is a baseline. A baseline with a stated method is more useful than a comparison against an invented target. A baseline is still a useful starting point.

Limits and sources of error

The sources cited here describe internal control and project scheduling practice in general terms; they do not provide a benchmark for escalation delay. A short delay may mean the team escalates quickly, or that it only records commitments it is confident about. The measure cannot distinguish those cases without the volume of commitments and the share that were never recorded. Report both alongside the delay before drawing a conclusion about discipline.

Private channels are a persistent blind spot. Some escalations happen in a corridor or a direct message that is never logged, which makes the measured delay look longer than the lived experience. Interview a small sample of owners about how they raised late items and report the share of escalations that left no record. That estimate belongs in the limitations, not in the headline.

Selection bias is a second risk. Commitments that were never written down are also the hardest to find and may have the longest delays. If the inventory depends on a tracker, the result describes tracked commitments rather than all commitments. State the coverage rate and avoid presenting the delay as an organization-wide fact. A tracker-based inventory can only describe the commitments the tracker actually holds.

Review cadence

Review the result at a fixed point and end with one change, one owner, and one check date. A practical first change is to require a named escalation contact on every commitment at acceptance, so an owner knows exactly who to tell when a date is at risk. Repeat the measurement after two full cycles before setting a target. Keep the definitions with the result so the next reviewer can compare like periods.

Retire the measure when most misses reach a decision-maker within the team's stated service level and the tail stops moving. If the team keeps it, publish the no-escalation share and the late-escalation share together, because a short median can hide a growing tail. Link the result to the dependency standup when late items wait on another team. The purpose is to shorten the silence, not to assign blame for the original miss.

Sources