Standup Duration and Decision Rate: Cutting 14 Minutes to 6

Twelve people attended a standup that lasted 14 minutes and produced 0.8 decisions per session, measured over three weeks. The same team later ran a six-minute version that produced 1.9, which suggests the extra eight minutes were doing something other than work.

The team maintained a booking system for dental clinics: two mobile clients, one API, 12 engineers and a product manager spread across four time zones.

What the 14 minutes contained

A volunteer timed the segments for fifteen sessions. Status reports took nine minutes and forty seconds. Problem discussion took three minutes. Nothing else happened, and the final minute was usually spent deciding whether to continue the discussion elsewhere.

Nine minutes of status reporting to twelve people is 108 person-minutes a day, which is two working days a week spent describing work rather than doing it.

The first cut

The team moved status to a written message before the call and reduced the call to blocking issues only. Duration fell to four minutes in the first week, then crept back to nine as people used the space to raise questions that were not blockers.

The creep is the normal behaviour of any meeting with a fixed slot. A shorter agenda expands to fill the room.

The timebox that held

The second attempt added a hard timer and a rule that the call ends when the timer sounds, even mid-sentence. Anything unfinished goes to a written thread. The timer was unpopular for three days and then became the reason people arrived prepared.

Duration settled at six minutes, with a standard deviation of 70 seconds across twenty sessions.

Why six and not three

Three minutes was tested and produced a higher rate of items deferred to written threads, which then took longer to resolve. Six minutes allowed the two or three live problems to be described aloud, which is faster than typing when the problem is tangled.

The team concluded that the value of a short call is not speed but selection: a short slot forces people to choose the two things that matter.

What replaced the status

The written message used three lines with ticket references, and it was read before the call by 10 of 12 people on average. Two people never read it and said so, which the team accepted because their work rarely intersected with the others.

Reading took about 90 seconds per person, which is a genuine cost and still less than nine minutes of listening to the same information.

The cost to the quiet people

Three engineers who rarely spoke in the fourteen-minute version spoke less in the six-minute one, because there was no slack. Their blockers surfaced in writing instead, which was acceptable, and two of them said they preferred it.

The format lost the accidental conversation that happens when a call runs long. Nobody could point to a decision that came from one, so the loss was recorded and not acted on.

The measurement method

Decisions were counted as items where the team agreed to change scope, sequence or ownership, recorded in the ticket system on the same day. Status updates that changed nothing were not counted, which is the distinction the exercise depended on.

The count of 0.8 against 1.9 is small enough that a bad week could shift it. Three weeks on each side is not a rigorous sample, and the team treats the direction as informative rather than the number.

Two quarters later

The call still runs at six minutes and has occasionally been skipped entirely when no blocker was posted by the deadline. Skipping happened four times in eight weeks, and no one reported a problem on those days.

Fourteen minutes produced 0.8 decisions a session. Six minutes produce 1.9, and the eight minutes saved per person per day are now spent on the work the meeting was describing.