Blog
Twelve on the call is not attendance
A Zoom participant count hides empty seats. How to run a visible clock, take attendance by role, and treat a muted required seat as a finding.
1 October 2026, 14:02. Zoom says 12 participants. The facilitator pastes the scenario into chat and says the room will walk through containment. Dana K, the legal seat the plan names, is in the participant list with camera off. At T+38 Priya S isolates prod-db-01. At T+41 Sam O asks whether the customer draft can go. Marcus T looks at Dana's tile and asks if legal is on. Silence. Someone types "Dana is on the call" in chat. The packet later reads "Legal attended."
That sentence is a pointer, the same way a Slack thread is a pointer. A join link proves who arrived. It does not prove who sat the seat when the inject that needed them landed.
A weak attendance line reads:
Present: 12 participants (Zoom). Legal, comms, engineering, and security all joined. Good discussion.
A strong one reads:
T+00 roll call by role. Present and speaking: Marcus T (incident commander), Priya S (technical lead), Sam O (comms). Dana K (legal) joined at 14:01, camera off, did not unmute on the T+41 notice inject. Seat treated as empty from T+41; Sam held the draft. Follow-up: Dana names a legal backup who will speak on the next run, by 2026-10-17.
Both describe the same ninety minutes. Only the second can be checked by someone who was not on the call. The first costs you the finding: on a real night Dana is in a client meeting, the 24-hour Northwind clock started at 09:12, and nobody practised the reassignment the plan needed.
Take the roll twice
The invite-list post is who has to be asked. Remote facilitation is whether they actually sat the seat. A T+00 roll call by role, spoken out loud, is the start. It is not the finish. Call the role again at the inject that names them. Presence at 14:02 is not presence at T+41.
Camera-off is fine. Mute through your inject is not. If Dana does not unmute when the notice inject lands, say her name. Wait. If she still does not own the call, write silent, treat the seat as empty, and reassign so the clock keeps moving. Do not let chat fill the seat with "legal would probably say wait." That sentence has no owner and no time.
If the plan cannot name who sits legal on a remote bridge, stop and run the free plan self-check before you book the next one. An unnamed seat will be silent for the same reason an unnamed seat is empty in the room: nobody was assigned to speak.
Put a clock on the screen
Remote rooms lose the wall clock. "We'll give you a few minutes" is not T+18. Share elapsed exercise time on the screen the whole session. Speak the inject time out loud when you drop it. Stop talking long enough for the owner to speak the choice and for the scribe to write it.
A weak facilitation line reads:
We'll give you a few minutes to discuss containment, then we'll move on.
A strong one reads:
T+18 is on the shared timer. Priya S has four minutes of exercise time: isolate prod-db-01, isolate the file share only, or keep both up for payroll at 06:00. At T+22 the scribe reads back: Priya isolated prod-db-01 at T+21, payroll risk accepted, backup path unknown.
The first produces chat. The second produces a decision someone can sample. The injects post is how to write the prompt. This is how to run it when nobody is in the same room. If an inject does not produce a named decision in about eight minutes, the inject is the problem, not the bandwidth.
Chat is not the decision log
Decisions that live only in the meeting chat are gone the moment the call ends. Require the owner to speak the choice out loud. The scribe writes four things: elapsed time, seat, decision, what was still unknown. Then they read that line back. If the owner does not confirm it, it did not happen.
Do not paste the chat transcript into the packet to make up the difference. Do not attach the recording and call it the minutes. The packet-hygiene post is why those files fail sampling. Facilitation is how you avoid needing them: the facts get written while the tiles are still on screen.
That is the attendance and decisions fields in our free evidence template. Fill them during the session and download a formatted record. It sits with three other formats in the template library. A sample packet shows the finished artifact with the same seats, including the legal stall when the seat is empty. Ransomware with a customer clock is one of the scenario families we actually run if you need a remote session that forces those two injects.
A well-run remote session proves that named people practised decisions under a concrete scenario, and that you kept a checkable record of who spoke and who did not. It does not prove the 02:14 bridge will be this meeting client, that the isolate button works, or that last night's backup of prod-db-01 restores inside four hours. Those are technical tests. Keep them off this call.
Bottom line: run a clock the tiles can see, take the roll by role at the inject that needs them, and write a muted required seat as empty. Twelve on the call is not attendance.