Lived once is not enough
The method has to have been used on three separate cases, not three retellings of the same incident. If the third case needs a different table, the method is not ready and the table is rewritten before any prose.
Twelve notes sit in the live index. A note is added only after the Four-Pass Method has a table or a checklist and a written failure condition. The selection rule is use, not novelty: if a method has not been run on three live cases, it stays in draft.
Notes are edited after publish when a field name changes or a collapse condition appears. The edit is dated in the body. There is no silent rewrite that makes an old note look as if it had always been right. Retired methods are listed below the grid so the archive also records what was pulled.
Use the chips to keep one category. Without JavaScript the chips are hidden and every tile stays in view. The first two titles open as full notes; the rest expand here as summaries until they are written up at the same length.

Decision systems
Most decision logs die in the first reorg because they store names, Slack threads and slide titles instead of the call itself. This note lays out a six-field log that still works when the team is split, renamed or absorbed. The rule is simple: a stranger from a neighbouring team should be able to reconstruct the decision from the row without asking who sat in the room.

Measurement
A metric tree is only as good as the argument you can make with it in a review. This note describes a tree with one outcome, three input metrics and two guardrails, plus a diagnostic layer that never appears on the weekly slide. The test is whether someone who disagrees with you can still use the same tree to argue the other way.

Writing & docs
Decks invent product. A one-page spec written before any slide forces the claim, the non-goals and the first test into a format that cannot hide behind a transition. This note includes the page layout used here, the order of sections, and the two checks that stop the spec from turning into a press release. If the page cannot be read aloud in six minutes, it is not finished.

Review rituals
Status meetings swell because they try to be a narrative. This review uses four columns and a hard stop: what shipped, what is blocked, what was decided, and what was deliberately dropped. The last column is the one teams skip, and it is the one that keeps the board honest. The note includes the prompt for each column and the rule for carrying a row into the next week.
Tooling
Tool sprawl is usually defended with sunk cost and a screenshot of a dashboard nobody opens. This audit used a single test for every tool: which decision or measurement depends on it weekly. Three tools failed that test in March 2026. The note records what was removed, what replaced the habit, and the two-week watch for regressions. Nothing here is a stack recommendation; it is a method for taking things out.
Decision systems
Intake committees form when nobody wants to say no in writing. This note publishes a five-factor score that anyone can apply to a request before it reaches a triage slot. The score is visible, the weights are fixed for a quarter, and appeals go to a single owner role. The failure mode is gaming the rubric; the counter is a quarterly weight review, not a secret override.
Measurement
Some work resists a clean count: quality of a spec, health of a handoff, load on a reviewer. This note describes three substitutes that still behave like measurements — sampled reviews, cycle-time bands and explicit non-counts — and the language to use when a number would lie. If a metric cannot survive the question “what would we stop doing if this moved”, it does not go on the tree.
Writing & docs
A template starts as a kindness and ends as a form people paste into. This note separates templates (shape of a document) from checklists (exit tests) and explains why the second ages more slowly. Versioning rules, a retirement date on every template, and a short list of phrases that signal rot. The accompanying checklist is in the Library; the argument is free.
Review rituals
Retros turn into blame when the prompt is “what went wrong” and the room is tired. This format asks for the system that produced the result, the prompt that would have changed the path, and one experiment with an owner role and a stop date. Names of people are not written on the board. The note includes the timing, the parking-lot rule, and the two phrases that get the room back on the system.
Tooling
Notes stall when capture and publish live in the same document. This pipeline treats a working note like a build: raw log, framed problem, tested table, then publish or retire. Each stage has an exit test and a place to stop. The tooling is deliberately dull — a dated file, a table, a retired list. The point is to stop rewriting the same incident in a nicer tone.
Decision systems
Teams lock long-range decisions in the same meeting that clears a two-week block. The horizon rule splits those calls: anything whose effect sits past fourteen days needs a revisit date or it stays a hypothesis. This note shows how the rule sits on the decision log, how exceptions are recorded, and why “we will know more later” is an acceptable state if it is written down.
Measurement
Dashboards fail by addition. This note is a refusal list: vanity totals, unowned charts, metrics without a decision, and anything refreshed more often than it is used. In their place: the metric tree, a last-updated stamp, and a link to the definition. If a chart cannot name the decision it serves, it does not ship. The accompanying teardown format is in the Library.
Retired methods
These were used, then pulled. They stay listed so the same shape is not rebuilt under a nicer name.
| Method | When it was used | Why it was pulled |
|---|---|---|
| OKR cascade sheet | January 2026 | Teams optimized the sheet instead of the work. Scores moved while behaviour did not. |
| RAG status traffic lights | January–February 2026 | The colours hid disagreement. Amber became the default and blocked nothing. |
| Five-whys form | February 2026 | The form turned into a blame script. The last “why” always landed on a person. |
Four tests. A draft that fails one of them is not scheduled for a Monday.
Lived once is not enough
The method has to have been used on three separate cases, not three retellings of the same incident. If the third case needs a different table, the method is not ready and the table is rewritten before any prose.
A stranger can lift it
A reader who was not in the room should be able to fill the table from the note. If the note depends on private context, slang or a vendor nickname, it goes back to Frame. The artefact is the product, not the anecdote.
Failure is named
Every note lists where the method breaks. Collapse conditions are written as situations you can recognize, not as character flaws. If the failure section is a shrug, the note is still a draft.
No client theatre
This archive does not publish named clients, conversion rates or testimonials. Those numbers belong to other people and they decay the moment they are used as proof. The work has to stand on the table.
If you are new to the notebook, these four notes in this sequence give the load-bearing pieces. The rest can be read in any order.
The public notes argue for a structure. The blank tables, the extended versions and the monthly teardown PDFs are paid files. If you only need the argument, stay on this page. If you need the file you can hand to a team, open the Library.