Question and scope

This protocol estimates how much coordination work happens outside the team's tracker. Coordination work is any act that moves a request, decision, or commitment between people: a status message, a scheduling exchange, a reminder, an approval, or a short verbal agreement. The result is a share of observed coordination acts that have no matching tracker record, plus a count of distinct requests involved. It measures the gap between what the team does and what its system of record shows, not the effort of any individual.

Scope is one coordination team over a fixed observation window, usually two to four weeks. Include work the team performs for other teams and work it performs for itself, but keep the two categories separate in the report. Exclude meetings that already have an agenda and minutes, since those have their own record. Record the window dates, the participating roles, and the tracker export used for matching so the estimate can be repeated.

Why untracked coordination stays invisible

The tracker records requests that someone remembered to log, which is not the same as the work the team performed. A coordinator who answers a question in chat, confirms a date by phone, or walks a draft to a reviewer has done real work that leaves no ticket. The queue looks calm because the invisible work absorbs the friction before it becomes a tracked delay. Over time, the team appears to have spare capacity that it does not actually have.

Untracked work also distorts staffing decisions. A capacity plan built from ticket counts will understate the hours needed for coordination and overstate the time available for tracked delivery. When a new person joins, the undocumented habits are the hardest part to transfer, because no queue shows them. Measuring the gap makes the hidden load visible without turning it into a surveillance exercise.

Defining a coordination act

Define the act narrowly enough that two observers would count the same event. A useful rule is one exchange with one purpose, such as confirming a vendor date or asking a team to approve a change. A long thread that covers three topics becomes three acts, while a single reply that closes a question stays one. Write the definition into the observation sheet before the window opens and test it on a short pilot.

Decide in advance whether reading counts as an act. Opening a message, scanning a channel, and deciding no reply is needed consumes attention but produces no visible transaction. Count reading only when it leads to a decision that is recorded somewhere, and note the rule in the report. Otherwise the estimate will drift with each observer's view of what attention means.

Choosing observation windows

Pick windows that represent normal work rather than a launch week or a holiday. Two full weeks usually capture the weekly rhythm, and a third week helps when one week contains an unusual deadline. Avoid windows that overlap a major outage, since the coordination pattern will not repeat. Announce the window in advance so the team is not surprised, but do not announce which exchanges will be sampled.

Sample across the full working day instead of only the morning. Coordination often clusters around handoffs at the start and end of a day, and a morning-only sample will miss the end-of-day chase. Include at least one day when the coordinator works from a different location, because remote days often shift work into chat. Record the dates and hours observed so a reader can judge how representative the window was.

Logging without surveillance

Log the act, the channel, the time, and whether it maps to a tracker item on its own. Do not log message content, names of requesters, or anything that reads like a performance judgment. The point is the shape of the work, not who sent what. Tell the team exactly what will be recorded and show them the sheet before the window begins.

Use a lightweight log that an observer can complete in the moment or at the end of each hour. A retrospective diary written at the end of the week will smooth over the short exchanges that matter most. If the observer is also doing the work, count only a partial sample and note the reduced coverage. A rough count with a clear method beats a precise count built on guesses.

Reconciling with tracker exports

Export the tracker for the same window and match each logged act to a request, a comment, or a status change. Matching is rarely one-to-one, because a single request can generate several acts and one act can touch several requests. Define the matching rule before you start, such as matching on request identifier plus date. Record unmatched acts separately instead of forcing them into the nearest ticket.

Count an act as tracked when it appears in the system of record in a form a reader could find later. A ticket that was updated without a comment may still count if the change records the decision. A chat message that was pasted into the ticket later counts as tracked, because the record now exists. Be explicit about timing, since a note added a week later documents the act but did not make it visible when it happened.

Calculating the untracked share

Report two numbers: the share of observed acts with no tracker match, and the share of observed minutes spent on those acts. The act share shows how often the team works off-system, while the minute share shows how much time that represents. Report both with the observation window and the number of acts. A percentage without its base invites a reader to imagine a precision the sample cannot support.

Keep a separate line for acts that were tracked only after the fact. Those acts were invisible during the window but are documented now, and mixing them into either category hides the delay in the record. A high after-the-fact share points to a recording habit rather than a missing process. Show the three categories together so the reader can see the whole distribution.

Sampling and coverage

Aim for at least one hundred observed acts before reporting a share, and treat smaller samples as a pilot. Spread the sample across roles, because a coordinator and a team lead may work in different channels. If one channel dominates, sample it in proportion but report the channel mix. Note any channel the team refused to include, since an excluded channel can change the result.

Check coverage against the calendar at the end of each window. If the observer was in meetings for half of a day, the log has a gap that should be reported rather than filled by memory. Compare the number of logged acts per hour with the team's own estimate of busy periods. A large mismatch is a reason to extend the window, not to adjust the numbers.

Reporting limits

This method measures observed acts, not the full universe of coordination. Work that happens in a meeting, a phone call, or a channel nobody sampled will be missed, so the share is an estimate with a known direction of error. State which channels were covered and which were not. Do not present the result as a complete account of anyone's day, since the sample was never designed to carry that weight.

The estimate also cannot say whether untracked work is waste. Some of it is the fastest way to unblock a colleague, and forcing every exchange into a ticket would slow the team down. The useful question is which untracked acts carry commitments that later disappear. Report the estimate next to examples of commitments that were lost, and let the team decide where a record is worth the cost.

Using the finding

Start with the untracked acts that carry a promise: a date, an approval, a change to a deliverable. Those are the ones worth moving into the tracker, even if the quick chat stays informal. The dependency standup format is one place to surface commitments that would otherwise stay in chat. Add a field or a shortcut that makes logging cheap, such as a one-line command that creates a ticket from a message, then re-measure after a month to see whether the share moved.

Do not use the result to argue that the team is inefficient. The estimate usually shows that the team is absorbing coordination that the process never counted, and the fix is a lighter record rather than more control. The single-inbox measure covers a related dependency that makes untracked work harder to see. Share the finding with the teams that send work so they understand why a request may take longer than the ticket suggests, and use it to build a capacity plan that matches the real work.

Sources