.comThis is falkster.com, the notebook. Falkster.AI is the company.Go to falkster.ai

The Operating Cadence, One Page

Three hours total. That is the whole weekly commitment, and it is the floor, not the ceiling.

design-operating-cadence.md5 KB901 words

The template


The Operating Cadence, One Page

The whole handbook as a calendar. What happens weekly, per sprint, per quarter, and who owns each. Print it. If a practice is not on a calendar with a name next to it, it is an intention, and intentions lose to whatever is urgent.


The weekly rhythm

SlotDurationWhat happensOwner
Monday30 minPick the week's research question. One, written as a question.Designer plus PM
Tue to Wed90 minContact. Two to four user sessions.Designer
Wednesday30 minRubric review on one rotating surface. Scores prepared in advance.Facilitator
Thursday30 minSynthesis. Update the evidence register. One line to the team.Whoever ran sessions
Friday15 minConstraint check. Anything shipped this week that broke a rule?Surface owner

Three hours total. That is the whole weekly commitment, and it is the floor, not the ceiling.

The Friday fifteen minutes is the one people cut first and the one that keeps the constraint set honest. Cut the Monday slot instead if something has to go, and skip the research week rather than run it on a manufactured question.


Per feature, before anything gets made

Sequence matters. These happen in this order, and the whole model breaks when they happen after.

StepOutputOwnerGate
1Failure inventory, ten rowsDesigner, engineer, supportNinety minutes, before design
2Constraint set, three to six rulesDesignerIn the ticket, before generation
3Range spec, three columnsDesignerWorst acceptable written first
4Exposure grid placement per actionDesignerNothing lands in Q4
5Wrong-path design and copyDesignerOwned, not overflow
6Rubric for the surfaceDesigner plus oneCalibrated before launch
7Decision recordWhoever decidedFiled, linked from the work

A feature that reaches build without steps 1 through 3 is a feature where design will be doing QA at the end. That is the single highest-leverage rule in this document.


Per sprint

RitualDurationOutput
Failure inventory for anything new90 minTen rows, feeding constraints and evals
Rubric calibration if disagreement rates rose30 minRewritten tests
Wrong-path check on anything shipping20 minEvery failure state has an owner
Decision records filedOngoingOne per non-obvious call

Per quarter

RitualDurationOutputOwner
Evidence register review2 hoursStale rows marked, upgrade targets pickedRegister owner
Ownership map review1 hourBoundaries re-agreed with PM and engineeringDesign lead
Range spec re-sort1 hour per surfaceTen outputs re-sorted, boundaries adjustedSurface owner
Kill list pass90 minOne ritual killed, replacement runningDesign lead
Rubric version review1 hourWhat changed, why, and whether scores are still comparableDesign lead
Cross-team failure inventory2 hoursFailures that cross boundaries, and who owns themDesign lead

The quarterly list looks long and is about a day. A day a quarter is the cost of a practice that does not quietly drift.


What to do when you cannot do all of it

Real teams are not going to run all of this in month one. The order to adopt in, highest return first:

  1. Constraint sets on the two highest-traffic surfaces. Everything else gets better once these exist.
  2. The Friday fifteen minutes. Cheapest ritual here, and it makes the constraints real rather than aspirational.
  3. Failure inventory before new features. Ninety minutes that saves weeks.
  4. One rubric, calibrated. Start with the surface that generates the most review argument.
  5. The weekly rubric review. Only after a rubric exists and is calibrated.
  6. The research cadence. Highest ongoing cost, so it comes after the cheaper things are habitual.

Adopting in a different order mostly does not work. The review needs a rubric, the rubric needs constraints, and the constraints need something to react to.


The two numbers to watch

Percentage of live surfaces with a current constraint set. The coverage number. If it is falling while you ship, judgment is not keeping up with generation.

Rubric disagreement rate. The health number. Near zero means the rubric is trivial or people are not scoring independently. Above a third means it is not shared understanding yet.

Report both monthly. Two numbers, no dashboard.


The one thing to do this week

Put the Friday fifteen minutes on the calendar for the next four weeks, with a name next to it. It is the cheapest item on this page and the one that tells you fastest whether your constraints are real.


From "The Operating Cadence", chapter 24 of The Design Operating Model. falkster.com/design/the-operating-cadence

More from the toolkit


All templates →