The Loop
What changed, and what do we change?.
A score is only worth taking if it changes what you do next. The Loop asks whether any of it worked, and changes the plan when it didn’t. The quarter has four boxes: ask, fix, ask again, and record. Record is the step most teams skip, and the one this step exists for.
Change nothing in the sample mid-year. Read ranges, not just numbers: a quarter whose range sits wholly above the old score is starting to prove something.
Split the loss by stage again and see whether its shape has changed. A fix that’s landed drops off; the next biggest lever moves up.
Three fixes a quarter, no more, each with its moments, its stage and a predicted range written down before it ships. The record is the asset.
The builds find moments the map missed. New questions are held as hypotheses until interviews confirm them.
Kestrel’s third-quarter note
The range now sits wholly above the January score of 18, for the first time. The mining mix-up: gone from every answer, as predicted, within three scans.
Kestrel is invented.
Where your craft is
Honesty is a craft. Dropping a fix that did nothing, in writing, is what makes the next prediction worth believing.
You should be, a little. 1,818 answers a month from the machines, the deal record read moment by moment, levers reranked every quarter, three fixes predicted and recorded, every build’s Loop table, and a yearly reset.