A year without sprints: nine engineers, 36 services, and a WIP cap of eight graded B

Little's Law does the work the grade does not. A cap of eight items in progress against 14 completed standard tickets a fortnight, taking a fortnight as ten working days and the queue as roughly steady, puts the average ticket in progress about six working days ^[1]. The 11 the team found sitting in Doing on the first Tuesday puts it near eight ^[2]. A limit does not make anyone type faster. It sets how long a half-finished thing is allowed to sit there, and those two days per ticket are the whole prize.

Then look at what the 11 was made of. Three were waiting on a reply, two had not been touched since the previous week, and one was a task to investigate service mesh options ^[10]. Five of the eleven slots were held by work that needed a message sent rather than an engineer's afternoon, which leaves six genuinely in hand, under the cap they had supposedly blown through ^[3]. The limit did not find an overloaded team. It found that "active" had been defined to include waiting.

That is also where the unbudgeted chores come from. When the board refuses a new start, the next move is usually outside the team: the engineer on the OpenTofu module went and asked the service owner for the missing variable list, and someone else confirmed with a vendor that a release had been withdrawn before closing a stale upgrade card ^[11]. Blocked went onto the board on day two precisely because Doing had been hiding that category of work ^[9]. None of it appears in the throughput count, and the team's own phrase for it, chores nobody put on a capacity plan [c5c], is the accurate one.

The claim about how disagreements get argued is thinner than the flow numbers, and the B grade concedes it [c5a]. Expedite has a test with a number in it, live production impact or a security deadline under five days out, and fixed date has to be a date imposed from outside the team ^[7]. The account presents that as what they expected of classes of service, not as something they went back and measured. What did get measured is the check-in, 25 minutes down to about 11 ^[12], which across nine people is roughly 21 person-hours a fortnight returned if everyone attends ^[6].

The cap itself is still the weakest number on the board. Eight came from headcount minus the people they assumed would be in meetings, a derivation the authors disown in writing ^[6], and it leaves fewer standard slots than engineers ^[4]. Re-deriving it from completions is harder than it sounds: 23 tickets in October against six in January is a spread of nearly four to one ^[13][5], and an average across that is something to argue about capacity with, not something to set a limit from.