Stafford Brief — 2026-11-05

One find, and it lands on your own desk rather than on the field's. This morning you re-spined asset 12 — the gate everything else waits behind — on the claim that nobody has assembled these five conditions in a form an operator can actually use, and you named Martins & Coetzee's Table 1 as "the asset's single best exhibit." I downloaded that PDF this afternoon and read the table. It does not say what we have been recording. The same search handed over a better exhibit, a quantified one, and a counterargument that is tougher than the one we had been defending against.

The 12:00 scan said today's substance already — the HBS Copilot paper, the coordination-displacement pairing, Goldberg & Puranam. Read its brief for those. This is the delta layer and it is almost entirely about the gate asset.

📖 TODAY'S READ

CFIR's own authors asked 134 experienced users whether their framework is usable. Only 16% said a non-specialist could use it. — The updated Consolidated Framework for Implementation Research based on user feedback, Damschroder, Reardon, Opra Widerquist & Lowery, Implementation Science 17:75, 2022-10-29.

They surveyed the 334 corresponding authors of every published paper that used CFIR and got 134 completed responses from people who had actually worked with it — over 65% on two or more projects. Then they asked the question directly. Easy for implementation researchers: 50%. Easy for non-researchers: 16%. One wrote, in the paper: "CFIR is far too complicated and difficult to use."

The intersection is the one you need: this is the nearest fully-integrated prior art to your framework, measured on exactly the axis you staked the gate asset on this morning, by the people who built it. You wrote that the usability claim was "currently unsupported." It is supported now, and by a harder instrument than you had — a number, in the field's own flagship journal, from its own authors, on its own users.

What to take from it: your spine is right and your exhibit was wrong. Swap one for the other. Full entry: knowledge-base/damschroder-updated-cfir-user-feedback-2022-10.md.

🥊 CHALLENGE

Your single best exhibit fails on four counts, and any reviewer with the PDF open finds them in a minute. Verified verbatim today; full working in knowledge/thesis-challenges.md 2026-11-05.

1. The table compares six models, not seven. Both this morning's correction and the 11-04 file say seven. There is no seventh row.
2. "Not one of the seven is rated usable" is false — and the exception is the row that matters. Burke–Litwin's evaluation reads "Fits most criteria", "Of practical usefulness", "Well-researched", "Empirical evidence of addition of value." The one model this table endorses is the complex one. And its lone limitation, which we have been quoting as bare "Complexity", reads in full: "Complexity (as in relation to the intricacy of organisational phenomena)." That parenthetical is a defence of the complexity — it says the model is complicated because organisations are.
3. It is a selection rationale, not an assessment. Models were "too simplistic for the purpose of the current study." The section opens by quoting: "there is no best diagnostic model. The model used depends on the situation and the practitioner's style and lenses."
4. The ratings are not even these authors' own. Table 1 is credited to five secondary sources. It is a compilation assembled by people who had already picked Burke–Litwin, to explain why they picked it.

And the count cuts against you. "Too simplistic" lands on Weisbord (six boxes), the McKinsey 7-S (seven elements) and the Galbraith STAR (five points). Five patterns and four forces sits on the "too simplistic" side of this table, not in a gap in it. You cannot win this axis by being shorter than the prior art. Shorter has been tried three times and rated inadequate — by the very compilation the asset wanted to cite.

The harder objection, which today also installs. Told plainly by 134 of its own users that the framework was too complicated, CFIR's authors added constructs: 37 in 2009 → 48 constructs plus 19 subconstructs, growing in every domain except one. They also print a user arguing the complexity gets blamed for implementation research being hard in itself. So "mine is clearer" is no longer facing silence. It is facing a documented, dated decision against parsimony, taken by an expert panel with user data in hand.

What it would take for this to stop mattering: the asset argues altitude and audience instead of count. CFIR is 48 constructs for researchers; Burke–Litwin is 12 variables with a 96-item questionnaire; yours is five patterns and four questions for an operator running it unaided. Lead with who the instrument is for and the count becomes a consequence rather than the claim. That argument is already sitting in point 2 of this morning's amendment — it just has to become the spine instead of the support.

⚙️ FOUR FORCES CONCLUSION

Momentum, and this is the day's one piece of good news — the claim held up against the updated prior art, not just the old one.

The base has said since 2026-10-11 that CFIR has no construct for things decaying after launch. That was the 2009 version. Today I checked the 2022 revision against the authors' own published construct list: none of the 48 constructs, and none of the 19 subconstructs, concerns sustainment, sustainability, maintenance or anything continuing after implementation. The nearest, Adapting, is about changing the thing while you are still rolling it out.

Why this is worth more than the 2009 absence. That revision existed because users told the authors what was missing. They closed gaps all over the framework. They did not close this one — and the same paper cites three sustainment papers in its reference list. So it is a deliberate scope decision, not an oversight. The field's most-used diagnostic looked straight at post-launch decay, had the literature to hand, and left it out.

The bound, because it matters for what you may write. The claim is not "nobody has a sustainment construct" — a sustainment literature plainly exists. The claim is: the integrated transformation-level diagnostics have no sustainment construct, while a sustainment literature exists that has not been integrated into them. Narrower, true, checkable. This partly discharges your 2026-11-11 sustainability check and replaces its search with three named references.

💭 OPEN QUESTION

If simplicity is the claim, who is the judge — and have you ever let one rule against you?

Every usability verdict found today is self-assessed by someone with a stake. CFIR's users rate CFIR. Martins & Coetzee rate the model they chose. Your four diagnostic questions have never been put in front of a leader who was not already in a room with you.

And here is the uncomfortable part. This morning's amendment proposes exactly the right instrument — have the client's own leaders identify the operative failure mode unaided, and record whether leaders asked separately converge. That is a test with a losing outcome, and you should name the losing outcome now, before you have data. If three leaders asked separately pick three different breakpoints, the framework is not simple to understand — it is simple to remember, which is Kotter's property and precisely what the McAdam standard indicts.

So: what convergence rate counts as a pass? Pick the number before the first engagement, not after. Everything in Tier 0.2 that gets decided after the data arrives is a rubric you cannot falsify, which is the thing you built the whole order to avoid.

🧱 ASSET SUGGESTION

No new asset, and one declined. master-plan.md hard stop 1 is unconditional and nothing today needs an exception. Highest number in the ledger stays 18.

Declining the "asset ten" the 12:00 scan registered this morning — the individual-coordination-falls / project-coordination-rises displacement claim. The scan cannot read master-plan.md and does not know about the freeze. The claim is good and it is genuinely yours; it belongs inside asset 2 (The Flattening Bet), which already argues the flattening bet, not in a nineteenth entry. Routed there. The constraint travels with it: same platform, same tool, same vendor, never independent corroboration.

Today's work is a single amendment to asset 12 — exhibit swapped, counterargument installed, content claim narrowed. Full text in knowledge/asset-suggestions.md 2026-11-05b.

SOURCES & THREADS

Threads — none due, none checked. The scan confirmed nearest due are 2026-11-09 (The Regime Fork) and 2026-11-10, and the standing 10-29 correction against early checks holds. Nothing today bears on an open thread.

Exploratory query — aimed at the gap you created this morning, not at the scan's gap list. Your correction conceded the usability claim is "answerable, cheap, and currently zero-evidence," and nothing in the gate ships without it. First query (inter-rater reliability of organizational diagnostic frameworks) was null on target — the field has healthcare error taxonomies and software usability models, but nobody has measured whether practitioners agree when using an organizational diagnostic. Second query reframed from the measurement to the instrument most likely to have been measured, and hit. Same genre move the scan recorded today. Logged in knowledge/query-log.md.

Goldberg & Puranam — two of the scan's four untried routes run, both failed. The Stanford GSB publication page and the INSEAD publications page are both metadata-only landing pages that link back to the blocked Springer article. Two routes remain (a later open deposit, or the authors directly). Still unquotable; the scan's discipline on this holds.

No candidate added. cfirguide.org is the framework's own documentation site, not an outlet that publishes findings — it is a reference, and logging it as a monitored source would inflate the ledger.

One correction owed to the base, routed below — the 11-04 KB file's description of Table 1 needs the same four fixes.

Routing

Strategy — the one thing to act on, and it is a swap, not a rewrite. Asset 12's spine from this morning survives intact. Replace its named exhibit: Martins & Coetzee Table 1 out, Damschroder et al. 2022 in. 16% of 126 experienced users say the nearest integrated prior art is unusable by a non-specialist, reported by its own authors — that is the sentence the asset has been missing since 2026-09-17, and it is better than what it replaces. Then rewrite the usability argument around altitude and audience rather than count, because the count argument loses to the table you were going to cite.

Strategy — and this is the part worth sitting with. This is the fourth instance in a week of a ledger recording something stronger than the source supports: the case file drifted on Claim 1 (11-04, caught by the scan), this ledger drifted on the novelty claim (11-05, caught by you), and today the exhibit installed to replace it drifted on its first day. Three of the four were caught within 48 hours, which is working. But the common cause is the same one your in-place Winning condition: / Discharged: proposal addresses: a ledger that cannot mark an item discharged cannot keep its own claims aligned to what it has actually read. The structural fix is worth more than any of the four corrections.

PROPOSED-UPDATE — knowledge-base/martins-coetzee-burke-litwin-diagnostic-framework-2009-09.md, the Table 1 paragraph in ## Summary and the Table 1 sentences in ## Relevance to Brandon's Positioning. The file says Table 1 is "a peer-reviewed comparison of seven organisational diagnostic models." Replace with:

Table 1, "Comparing diagnostic models," compares six models — Weisbord, Nadler–Tushman congruence, Freedman Swamp, McKinsey seven-S, Galbraith STAR and Burke–Litwin. It is a model-selection rationale rather than a comparative usability assessment: the body states the alternatives were "too simplistic for the purpose of the current study" or "too complex to address the aims of the study," against Burke's three selection criteria, and the section opens by quoting Jones & Brazzel that "there is no best diagnostic model." The ratings are not the authors' own — the table is credited to Cummings & Worley (2005), French & Bell (1999), Jones & Brazzel (2006), Van Tonder & Roodt (2008) and Van Tonder & Dietrichsen (2008). Burke–Litwin is rated positively ("Fits most criteria", "Of practical usefulness in cross-cultural research setting", "Well-researched", "Empirical evidence of addition of value to outcomes of an organisational diagnostic process"), and its sole limitation reads in full "Complexity (as in relation to the intricacy of organisational phenomena)" — a defence of the complexity, which the body then makes explicit. Table 1 cannot be cited for the proposition that the genre rates none of its models usable.

PROPOSED-UPDATE — knowledge/paper-2-case-file.md, gap ranking. Add: the operator-usability instrument is now partly supplied (Damschroder et al. 2022, 16%/50%) and the remaining gap is narrower and sharper — nobody has measured whether practitioners asked separately converge on the same diagnosis using any organizational framework. Today's first query returned null on exactly that and the nearest neighbours are healthcare error taxonomies (Fleiss' kappa 0.43–0.72 across two taxonomies) and software usability models. This is the measurement Tier 0.2's proposed second instrument would produce, which makes Tier 0.2 the only route to it in the near term. Also: two of the four untried Goldberg & Puranam routes are discharged negative (Stanford GSB, INSEAD) — gap item 1 stays first but is cheaper to state than to solve; two routes remain.

PROPOSED-UPDATE — knowledge/emerging-patterns.md, candidate (not opened). The Prior Art Chose Complexity: given direct user feedback that a diagnostic framework was too complicated, its maintainers added constructs rather than reducing them. One member only — CFIR 2009→2022 (37 → 48 constructs plus 19 subconstructs). Opening condition: a second diagnostic or determinant framework whose documented revision, in response to recorded usability complaints, also increased its element count. Burke–Litwin has had no comparable revision, so this needs a genuinely different framework and not a second reading of the same two.

Governance / Market — nothing today. No window opened; the Uber window has six days and nothing here adds to it. Today's find is 2022 and carries no clock.

OPS — research-evidence branch claude/modest-faraday-0bxhno: one KB file copied (damschroder-updated-cfir-user-feedback-2022-10.md) and one entry staged (2026-11-05-implementation-the-updated-consolidated-framework, pending/2026-11-05). Pushed to that branch only; promote-claude-branches folds it into main once scripts/verify.sh passes. Note on the staging path: --from-kb mis-parsed this file twice — it read the "considered and declined" lines as live breakpoint tags and would have staged momentum_mirage and process_friction as evidence. The KB file was reworded so those sections no longer lead with an exact tag name, and the entry was then re-staged from an explicit payload with a proper instrument block and real key findings. Any KB file that names a breakpoint in order to decline it is a staging hazard; the converter cannot tell affirmation from refusal. Worth a note in the scan's own ingest discipline.