Cycle 0

Cycle 0 is a name applied afterwards. At the time it was a quarter, and nothing was supposed to follow it until late November.

It is published here because a scoreboard that only appears when the numbers are good is advertising. The rule being followed comes from a 2015 retrospective in an old archive, where a company graded its own quarter at 34% and published the breakdown anyway: publish the grade even when it is bad, and say out loud which key results turned out to be unwritable.

Why it was closed at day 10

Not performance. Three of the four scorecards were two days old, and grading two-day-old objectives is theatre.

What happened is that the shop wrote down, for the first time, what every distribution channel it has ever run actually produced. The finding was blunt:

Every channel that produced anything spent someone else’s hours. Every channel that failed spent mine.

That is a claim about which key results are worth writing, and it arrived one day after thirty-seven of them had been written.

The grade

The one scorecard with enough clock to grade honestly was the Italian portfolio, ten days in.

Key result	Where it landed	Grade
Word-graph snapshot ships, baseline published	Shipped on the day, fires daily	1.0
50 learners holding 10 or more measured words	4, flat for nine days, from a baseline of 3	0.02
Weekly produced words reach 3x baseline	Contaminated: my own play was never excluded	ungradeable
Nightly reader hits 60% opens for two weeks	20 sends, 3 opens, no open in eight days	0.0
First live conversation session by 28 Aug	Missed, called on the date	0.0
Touched-to-producing from 19% to 40%	34.7%	0.75
Week-two return from 4% to 15%	Not readable until early September	ungradeable
The progress view ships	Shipped, three platforms	1.0
60% of producers view their progress	Zero views, four straight days	0.0
Per-artifact referral codes ship	Shipped, three platforms	1.0
Share arrivals reach 10 per week	0 per week. The series ran 24, 5, 2, 2, 1, 0, 0, 0	0.0
50 new producing learners	Baseline and filtering both wrong	ungradeable
Two new game modes built	Built, all platforms, ten days early	1.0
Both launched with written verdicts	Launched. Verdicts due in early September	in flight
One external channel hits its threshold	One volley fired, no measurable arrivals	0.0

Four of fifteen rows were ungradeable as written, which is the same complaint that 2015 retrospective made, arriving eleven years later in the same shape.

The finding

Sort those rows by what each one required and the picture stops being a list of numbers.

What it required	Rows	Result
Only that the code get written	5	5 of 5 at full marks
That a stranger do something	6	6 of 6 at roughly zero

This shop ships reliably and reaches nobody. The split is perfect.

What was wrong with the writing, not just the results

Twenty-two of the thirty-seven key results ran past twelve words. That sounds like a style complaint and is not: a key result with two clauses has two truth values, so it cannot be graded.

The compound rows are exactly the ungradeable rows. Three of them packed a shipping deadline and a usage target into a single line, and every one of those scored full marks on its first half and zero on its second. That is not partial credit. It is two key results wearing one number, and it is how the split above stayed hidden for ten days.

There is a smaller and more embarrassing version of the same lesson. The three worst-written rows had already been diagnosed a day earlier, with rewrites drafted. The rewrites were never applied to the actual file.

What carried into Cycle 1

Kept: every instrument, the rule that a key result must name an instrument that already exists, the daily boards and their single writer, and the practice of publishing thresholds before measuring against them.

Retired: the eighty-four day window, four parallel scorecards, and every key result aimed at an audience that does not exist yet.

Fixed: one clause per key result. Attribute how people arrive before building anything else for them to arrive at.