A weekly playable iteration loop
A weekly playable iteration loop keeps changes to one variable at a time, running from Monday's data read to Friday's live launch.
The cadence
A playable team that ships one polished build a month is optimising for the wrong thing. The unit that matters is the iteration rather than the launch, and a weekly cadence keeps that unit small enough to learn from without burning a full production cycle on every guess.
Monday is a read day, not a build day. The team pulls the prior week's numbers, meaning completion rate, average session length, click-through past the end card, then picks one variable to change. Not three. One. A build that changes tutorial pacing and end card copy and difficulty in the same week produces a result nobody can attribute to anything.
Tuesday and Wednesday are variant days. Two or three variants of the chosen change go into build, each isolating that one variable at a slightly different setting: a tutorial three seconds shorter, one five seconds shorter, or the tutorial removed entirely.
Thursday is QA, and under deadline pressure it is still not optional. A playable that crashes on one in twenty devices in a target market produces data that looks like a creative failure when it is really a technical one, and teams that skip Thursday QA to hit a Friday deadline usually pay for it with a week of misread results.
Friday is launch, into the smallest test budget that will return a readable signal by the following Monday.
Naming and retirement
Version naming should encode the one variable that changed, not a sequential number nobody can decode six weeks later. A build labelled tutorial-3s beats v14. When a mechanic has been through four or five iterations without a completion-rate improvement worth the production time, as this desk covered in its piece on the playable production bottleneck, retire it rather than iterate again on hope. That call belongs on the Monday read, against the data, not in a hallway conversation after a bad week.
This cadence is a rhythm rather than a rulebook, and it survives contact with a real production calendar better than heavier processes do, precisely because each week's commitment is small enough to abandon if the read says to.
