Stafford

← 2026-10-27 · all briefs · 2026-10-29 →

Stafford Brief — 2026-10-28 (Wednesday)

OPS. Clean run. Started from main at 663bfcb; both repos synced before work began. Yesterday's run merged, so there is no OPS flag to carry. Ship disposition is on the last line.

What this brief is. The 12:00 scan was a big day and then Brandon answered all three open decisions at 14:05, declining all three and setting a rule in the process. None of that is repeated here. This brief does one thing: it takes the rule Brandon set this afternoon and points it at the evidence base instead of at the two sources he declined. The rule lands somewhere nobody looked — in his own published paper — and it costs him a figure he has been using for fourteen months.

📖 TODAY'S READ

The number Paper 3 uses to replace the "70% of transformations fail" figure is executives grading their own homework, and the article it comes from publishes no method at all.

Paper 3 retires the 70% because it traces to a 1993 estimate the authors called unscientific and a 2011 review found nothing underneath it. Then it offers what "the evidence does support": McKinsey's finding that organizational redesigns succeed less than 25% of the time among companies following few of its nine design rules, and 86% among those following all nine. That sentence does the work the retired figure used to do.

I read the study today. It is the first time anyone in this system has. McKinsey Quarterly, June 2015, Aronowitz, De Smet and McGinty — reachable the whole time from a third-party copy of the PDF, which matters below. Here is what the 86% is, in McKinsey's own words rather than mine:

So one executive reports how well their organization followed the rules, and the same executive, in the same questionnaire, reports whether it worked. Both halves of the correlation come out of one person's account of their own work. And the article carries no methodology section whatsoever — no sample size, no fielding date, no sampling frame, no response rate, no definition of success, no scoring rule. The "1,323 companies" that secondary sources attribute to this study appears nowhere in it.

Why today and not some other day: this is Brandon's own rule, pointed at Brandon. At 14:05 he declined two pieces on one ground — "people can say whatever they want... there are no specifics to validate what they claim" — and the scan wrote that into the standing rules as covering the whole maturity-ladder genre. It was then applied to the two sources declined and to nothing already in the base. I swept the index for the same genre: 33 entries. Most are harmless. One is in print under his name.

The so-what, and it is a decision rather than a citation. The breakpoint intersection is Momentum Mirage, operating inside the instrument that is supposed to detect it — the field's most-cited measure of redesign success records declared victory, not measured movement. The fix is cheap and Paper 3 already carries it: lead with Flyvbjerg's 16,000+ projects — 8.5% delivered on time and on budget, 0.5% delivering the promised benefits. That is administrative outcome data from real projects, nobody is scoring their own work, and it is precisely the kind of validation this morning's rule says to look for. Then name the McKinsey spread for what it is — the field's self-assessment of its own methods — in one sentence. The structural argument does not depend on the 86% and never did. Full entry, including the one real defense of the figure: knowledge-base/mckinsey-getting-organizational-redesign-right-2015-06.md.

🔭 THINKER TO WATCH

Aaron De Smet, McKinsey senior partner (People & Organizational Performance) — and the point is that he is not on the field map, which is a hole rather than an oversight.

De Smet co-wrote the 2015 article above. He is also the recurring author behind the McKinsey organizational material this base cites more than any other single source — the State of Organizations work, the redesign rules, the operating-model line — and he published the successor to the 2015 piece on 2025-06-25: "The new rules for getting your operating model redesign right." The field map has 3,996 lines and nearly four thousand lines of thinkers in Brandon's territory, and the man who wrote the number in Brandon's paper is not in it.

What he argues: that redesign success is a function of following a documented, learnable set of rules, and that the rules can be refreshed as conditions change. Where Brandon differs, sharply and in a way worth saying out loud: Paper 3's whole argument is that the rules are already available and still not used, so the gap is structural — accountability, tenure, career calculus — not informational. De Smet's response to a persistent failure rate is to publish better rules. Brandon's is to say that publishing better rules has never once closed the gap. Those are opposed positions, not adjacent ones, and the differentiation is cleaner against De Smet than against any academic in the base.

Second figure, smaller but genuinely useful: Aarn Wennekers, The Board Room Voice (Substack), who published the only critique of the 2025 successor I could find — 2026-01-29. His line is worth keeping: "High certainty in complex systems is usually a signal, not of mastery, but of over-smoothing." His diagnosis is not Brandon's, and the contrast is the useful part: Wennekers says the rules cannot work because organizations are complex systems that resist planning. Brandon says the rules do work and the system rewards not doing them. Same observation, opposite mechanism — and Brandon's is the one that implies a remedy.

🏢 CASE IN THE WILD

The organization is McKinsey, and the case is what happened to its own number.

Not a client story — the firm itself, because the sequence is a clean instance of the mechanism and it is dated. In 2015 McKinsey published nine rules and reported 86% success for companies following all of them. On 2025-06-25 it published a refreshed set of nine rules. Per McKinsey's own email digest, the refreshed nine score 97% success — and the original nine now score 59%.

Read that again as a measurement claim. The same construct, from the same firm, measured eleven years apart: 86% becomes 59%. Meanwhile the new product scores 97%. Either organizations got dramatically worse at executing the old rules, or what moved was the instrument, the framing and the thing being sold. The second reading is the one a hostile reader reaches in about four seconds, and this base already has a hard rule built on exactly this shape — the NBER working paper whose headline numbers changed three times in fifteen months.

Hold the bar hard here, because I could not read it. mckinsey.com, mckinsey.com.br and the ~/media/ PDF paths all returned HTTP 503 to this environment today — the fifth consecutive scan, and now established as covering the whole domain family rather than one site. Every successor figure in this section is a search summary of a page nobody in this system has read, and none of them may be published. What is established, from multiple independent routes including a dated third-party critique: a successor exists, dated June 2025, and Paper 3's August 2026 update does not account for it.

The open question this leaves, and it is Brandon's to answer with access I do not have: does the 2025 successor measure success any differently than the 2015 version did? If it is the same self-report design at a higher number, the successor is not a threat to his argument at all — it is the best evidence for it anyone has produced, because it shows the field's own success metric inflating 11 points while nothing in the world changed.

⚙️ FOUR FORCES CONCLUSION

Momentum, and specifically what we are willing to count as evidence that it exists.

Momentum stewardship is reinforcement: the organization keeps going because progress is visible and real. The 86% figure is reinforcement built on declaration — executives who say they followed the method also say it worked, and the field has cited that loop for eleven years as proof the method works. Paper 2's stewardship table says Momentum fails when activity substitutes for movement. The measurement instrument for the field's own best practice does exactly that, and it took the author's own skepticism rule, applied this afternoon to someone else's survey, to make it visible here.

For the case file this is one line, not a revision: when the base judges an instrument, "did the respondent author both sides of the relationship?" belongs next to "what is the observation window?" in the standing instrument check. One question, free to ask, and it would have caught this in 2015.

💭 OPEN QUESTION

Which other number in your published papers would fail the rule you set this afternoon — and would you rather find them, or have a reader find them?

Not rhetorical, and the scope is small enough to finish. The six papers' cross-paper evidence anchors are listed in one file. A first pass today says the 86% is the worst case, and that two others are worth a look on the same test: "81% of boards raised expectations, 30% can reconfigure quickly" (KPMG — can reconfigure quickly is a self-assessed capability) and "C-suite 56% vs middle management 27%" on strategic clarity (McKinsey — self-reported, though the gap between two groups answering one question survives much better than either level does, and the 29-point collapse is the part the paper actually uses).

Here is the distinction worth keeping, because the rule applied flatly would cost too much. Self-report is worthless for "are we a mature organization" and perfectly good for "do I personally feel motivated" — a person is the only possible authority on their own motivation. The Bain figure in Paper 2's anchors (33% feel personally motivated to adopt the new operating structure) is fine. What fails is a respondent certifying an organizational fact, and what fails worst is a respondent certifying both an organizational fact and its consequence. Three tiers, one sweep, probably two hours.

🥊 CHALLENGE

Wednesday falsification sweep. The sweep was pointed at this layer's own finding, at full strength — and it failed to falsify.

The target. The 12:00 scan had already spent its Wednesday slot on AI gains depend on complementary organizational change and returned an honest null. Re-running it would have been the same query twice. So this slot attacked the claim I had just formed: that the 86% is self-classified on both sides and would fail Brandon's own rule. The sweep hunted for the opposite — evidence that the 2015 study used an independent or external measure of success. It found the article instead, and the article answers in McKinsey's own verbs: felt, declare victory, owned up to, acknowledged. The claim survived contact with its own best counter-search.

The one thing that genuinely cuts the other way, recorded so it travels with the critique. McKinsey says it "tested more than 20 common approaches and found that upward of half of them weren't correlated with success," naming competitor benchmarking as specifically disconfirmed. That is a real defense. If executives simply flattered themselves across the board, every practice would correlate with success about equally. Half of them not correlating means the instrument distinguishes between practices. Its limit: that tells you executives' shared beliefs about what good practice looks like differ — not that their outcomes differ. If leaders broadly agree that "aligned early" counts as doing it right and "benchmarked competitors" does not, this exact pattern appears with zero real variation in outcomes. The defense narrows the attack. It does not defeat it.

Where the objection is wrong, stated first because the ledger should say so before Brandon does. His rule as written forbids using a self-classified share as a claim about how many organizations are in a category. Paper 3 does not do that — it cites a conditional (success covaries with adherence) and uses the spread to argue the method is available and unevenly used. The rule does not formally reach it. But the reasoning behind the rule reaches it harder than it reaches the share, and that is the uncomfortable part: a share from a self-classified tier is merely uninformative, while a correlation between a respondent's account of their method and their verdict on their own result is exactly what self-presentation alone predicts. The share is weak evidence. The conditional is evidence-shaped. And in Paper 3 it is load-bearing.

Full entry, including the defense, the bounds and what would defeat it: knowledge/thesis-challenges.md, 2026-10-28 — The replacement-figure objection. Status OPEN: the method half is effectively conceded and awaiting Brandon's decision on what Paper 3 runs on instead; the citation half needs the 2025 primary, which needs access this pipeline does not have.

🧱 ASSET SUGGESTION

Asset 12 just found its opening paragraph, and no new asset is proposed — master-plan.md hard stop 1 is unconditional and this respects it.

Asset 12, "What the Five Breakpoints Owe," is the Tier-2 gate that nothing in the differentiation class ships before, and the plan specifies it as "the Paper 3 seventy-percent move at corpus scale." The strongest possible version of that move is performing it on his own replacement number, under a rule he set himself. The essay's hardest structural problem has been how to concede without sounding defensive. This solves it: open with the author retiring his own load-bearing figure in public.

The sentence it can now carry: "I retired the 70% because nobody could check it. The number I replaced it with is executives grading their own work, published without a method, and I used it for fourteen months. Here is the one in that family that survives, and here is why the argument does not depend on any of them."

That is the essay's thesis performed instead of asserted, in four sentences. Full spec, including the repair that keeps the structural argument intact: knowledge/asset-suggestions.md, 2026-10-28. And the honest part: asset 12 now has an opening paragraph, a concession section, a prior-art section and a repair specified, and zero words written. Eighteen proposed assets, zero shipped. Research has not been the binding constraint on this asset for two weeks.

🪟 WINDOW FLAG

None, and the reason is worth one line rather than silence. The successor article is sixteen months old and nothing is breaking. Under the discipline Brandon applied to the ONS material at 14:05 — keep the evidence, the clock has run out, the argument has no clock — this is an asset input, not a window. Manufacturing urgency on a June 2025 article would be exactly the error the rules name.

🔍 SOURCES & THREADS

Threads — none due, none checked. The scan confirmed the earliest due dates are ASML (11-01) and Project OT (11-02) and checked nothing early. Nothing owed here, and nothing in today's find opens a thread: the open item is a retrieval problem Brandon can solve and the pipeline cannot, which is a chase, not a story.

Chase — partially discharged after 57 days, and the route lesson is the valuable half. The McKinsey "new rules" chase has been open since 2026-09-02, attempted on at least ten consecutive scans, every route returning 503, and on day ten the scan's own recommendation was hand it to Brandon or formally abandon it. Ten scans attacked mckinsey.com and its regional mirrors. The 2015 predecessor was readable the entire time from a third-party PDF mirror found by searching the article title (coreinternational.com; veruspartners.net carries another). Today also establishes that the 503 covers mckinsey.com.br and the ~/media/ paths — domain-family-wide, which is a sharper statement of the coverage hole than "mckinsey.com is down," and it means a third-party mirror is not a workaround but the only route. The successor article itself remains unread and its figures uncitable; the chase stays open, narrowed to "find a mirror of the 2025 article or get it from Brandon."

Candidates — 1 added, 1 field-map candidate. Added: third-party PDF mirrors of McKinsey content as a retrieval channel (not an outlet — a route), cadence irregular. Field map: Aaron De Smet (absent, and he is the author behind the base's most-cited organizational source) and Aarn Wennekers / The Board Room Voice (cadence continuous, 1 signal).

Exploratory slot — spent on the Wednesday falsification sweep above. Confirmed unspent before searching: zero KB files, zero index entries and exactly one citation line across both repos for Aronowitz / getting organizational redesign right / nine design rules. Recorded for what it says about the base's priorities: Ewens & Giroud has been verified three times in seven weeks, and the instrument behind the number in Brandon's own published paper had never been looked at once. Full log: knowledge/query-log.md, 2026-10-28.

Staging note. The entry tripped the pipeline's retired_figure_70pct flag — a false positive: the entry is about the retirement and quotes the figure to name it. Flagged here so a later reader does not treat the warning as a defect. The entry was also submitted to verify-entry.sh ahead of its 24-hour auto-merge, since its load-bearing content is a criticism of a citation in a published paper and that deserves a human read.

Routing

PROPOSED-UPDATE → research-evidence/knowledge/papers-reference-index.md, §Key Evidence & Citations Across Papers

Under the retired-70% block, add:

Use restriction on the replacement spread (added 2026-10-28). Of the four components that replace the bare 70%, only Flyvbjerg & Gardner (16,000+ projects; 8.5% on time and on budget, 0.5% delivering promised benefits) rests on administrative outcome data. The McKinsey redesign spread (<25% → 86% by design-rule adherence) is a 2015 McKinsey Quarterly survey in which executives self-classify their own rule adherence and, in the same questionnaire, deliver the verdict on their own result — McKinsey's own wording is that executives "felt that the organizational redesign had succeeded" and were six times more likely to "declare victory." The article publishes no methodology section: no sample size, fielding date, sampling frame, response rate, definition of success or scoring rule. The widely attributed "1,323 companies" is not in it. BCG's 70%-short-of-objectives and Bain's 88%-short-of-ambitions are also self-reported outcomes against self-stated ambitions. Lead with Flyvbjerg; cite the McKinsey spread only as the field's self-assessment of its own methods, named as such. Per the standing self-classification rule (publication-windows.md §Rulings 2026-10-28), a self-classified adherence tier crossed with a self-assessed outcome is the most exposed form of that genre, not an exception to it. Successor note: McKinsey published "The new rules for getting your operating model redesign right" on 2025-06-25, revising the nine rules; secondary reporting puts the original nine at 59% and the refreshed nine at 97% in that survey. Those figures are not citable (every mckinsey.* route returns HTTP 503 here), but the dated existence of a successor the Paper 3 August 2026 update does not acknowledge is established. Entry: knowledge-base/mckinsey-getting-organizational-redesign-right-2015-06.md.

PROPOSED-UPDATE → research-evidence/knowledge/emerging-patterns.md, Confidence Without Substrate

This morning's scan brief proposed re-reading the lane as "confidence is reported at whichever level the respondent is not accountable for" on the strength of a second anti-member, and flagged it for Friday. Recommend not adopting that re-reading, and the reason is Brandon's 14:05 ruling, which the proposal predates by two hours. The proposed version still treats the direction of a self-report as informative — confidence inflates where accountability is absent. Brandon's rule says the direction is not the issue: the low tiers are nearly free to claim and the high tiers are unevidenced, so neither direction carries information about organizational state. A lane built on where self-reported confidence points is built on the thing the rule says cannot bear weight. Proposed instead: keep the pattern as a record of how the genre behaves — explicitly a pattern about instruments, not about organizations — and require any member promoted to ESCALATED to carry at least one measure the respondent could not author. Current members do not clear that bar, which is the honest state and worth recording as such.

PROPOSED-UPDATE → research-evidence/knowledge/source-candidates.md

  1. McKinsey "new rules" chase — status change, not closure. Open since 2026-09-02, ten-plus identical 503s, recommended for hand-off or abandonment on day ten. Partially discharged 2026-10-28 by a route nobody tried: a third-party PDF mirror of the 2015 predecessor, found by searching the article title. Record three operational facts: the 503 covers mckinsey.com, mckinsey.com.br and mckinsey.com/~/media/ (domain-family-wide, fifth consecutive scan); third-party mirrors (coreinternational.com, veruspartners.net) serve McKinsey PDFs and are the only working route from this environment; and the 2015 predecessor is now read and filed, so the chase narrows to the 2025 successor alone.
  2. ADD — third-party PDF mirrors as a retrieval channel. First seen 2026-10-28, surfaced by the chase above. Cadence irregular, no expiry clock. Not an outlet and not a source of signal on its own: a named route for the publisher this base cites most and cannot reach. Note the hazard — a mirror can be stale or altered, so record the mirror host alongside the citation and treat the figures as primary only when the document is internally complete, as today's was.
  3. ADD — Aarn Wennekers, The Board Room Voice (Substack). First seen 2026-10-28, via the only critique found of McKinsey's 2025 successor article (2026-01-29). Cadence continuous, 1 signal, expiry 2026-12-27 absent a second. Worth probation: a board-facing voice attacking consultancy redesign claims on complexity grounds — adjacent to Brandon's territory with an opposed mechanism, which is the useful kind of neighbor.

PROPOSED-UPDATE → research-evidence/knowledge/field-map.md

ADD — Aaron De Smet, senior partner, McKinsey People & Organizational Performance. First seen as a field-map candidate 2026-10-28. Rationale for adding him at first contact rather than on probation: he is not a new find, he is a long-standing absence. He co-authored the 2015 redesign article behind Paper 3's replacement figure, the 2025 successor that revises it, and recurs across the McKinsey organizational material this base cites more than any other single source. Position: redesign success follows from a documented, learnable and periodically refreshed rule set. Brandon's differentiation: Paper 3 argues the rules have been in print for a decade and are still not used, so the gap is structural rather than informational — publishing better rules has never closed it. Opposed positions, not adjacent ones, and the sharpest differentiation available against a named practitioner rather than an academic. Watch for: whether the 2025 successor's success measure differs from the 2015 self-report design — if it does not, the 11-point rise in self-assessed success is evidence for Brandon's argument rather than against it.

MERGED afa6f65 main. Research-evidence: branch claude/modest-faraday-9556hm pushed (1 KB file, 1 staged entry submitted to verify-entry.sh); its promote-claude-branches workflow folds it into main once scripts/verify.sh passes. Brief emailed to brandon.freitag@gmail.com.