A running record of the claims our program designs rest on, and the sources behind them
Every leadership programme you write stands on somebody else's work. You name Kotter in the change module, Edmondson in the team module, Kirkpatrick in the measurement plan — and somewhere in your client's building is a person who has read two of the three.
This page is the record of what happens when we check those names before we use them.
I am publishing it for a plain reason. We sell design documents whose main promise is that the scholarship in them is real. That is an easy thing to claim and a hard thing to demonstrate, and the only demonstration anyone should accept is the working. So here is the working: what we asserted, who it belongs to, what the source actually said, where we read it, and what we could not confirm.
I have presented a slide with a number on it that I could not have sourced if asked. Nobody asked. That is not a defence, it is the reason the habit survives — and it is why this page has a column for the things we could not close.
Three verdicts, and they mean exactly what they say.
Verified — we read a primary source, the author's own organisation, or the publisher's record of the work, in the month shown, and the sentence we use stays inside what that source supports.
Limited — we confirmed that the work exists and where it was published, and could not reach the full text. We quote only what we read. Anything beyond that is absent from the row rather than softened.
Cut — we intended to use the claim, could not close it in this pass, and removed it. The row stays on the page with the reason.
The last category is the one that makes the other two worth anything.
What we say: the eight steps, as Kotter's own firm currently names them, are create a sense of urgency, build a guiding coalition, form a strategic vision, enlist a volunteer army, enable action by removing barriers, generate short-term wins, sustain acceleration, and institute change.
What the source says: those eight names are read verbatim from Kotter Inc.'s own methodology page. The same page states: "He evolved the linear 8 Step from Leading Change to the 8 Accelerators outlined in his 2014 book, Accelerate and the firm's 2021 book, CHANGE."
Why we print this rather than the version in your slide deck: most eight-step lists in circulation are the Leading Change wording, presented as current. The author's own organisation describes that linear form as something he moved on from. A design that presents "Kotter's eight sequential steps" in 2026 is presenting a version its own author has revised — a small error, and one a client who has bought Kotter's consulting will notice.
What we did not verify: we did not read Leading Change or Accelerate in this pass. This row is what the author's firm states today, and nothing more.
What we say: we do not use it.
What the source says: Mark Hughes reviewed the figure directly in Do 70 per cent of all organizational change initiatives really fail?, Journal of Change Management, volume 11, issue 4, 2011, pages 451–464 (record). The abstract is unambiguous:
"A 70 per cent failure rate is frequently attributed to organizational change initiatives, raising questions about the origins and supporting evidence for this very specific statistic. This paper critically reviews five separate published instances identifying a 70 percent organizational change failure rate. In conclusion, whilst the existence of a popular narrative of 70 percent organizational change failure is acknowledged, there is no valid and reliable empirical evidence to support such a narrative."
Separately, and worth knowing: Kotter Inc.'s current methodology page — the page the figure is most often bounced off — states no percentage failure rate at all. We asked that question of the page directly.
What we did not verify, and will not imply: we did not read Leading Change, so this page does not tell you John Kotter never wrote the number. It tells you that a peer-reviewed review of five published instances found no valid empirical basis for it, and that his firm's current statement of the model does not carry it. Draw your own conclusion; just do not put the figure in a business case, because the person across the table can find Hughes in one search.
What to say instead: name the specific failure you are addressing in this organisation, with the client's own evidence. It is a better argument anyway, and it cannot be fact-checked out from under you.
What we say: the distinction between technical problems and adaptive challenges comes from Ronald Heifetz's work with Donald Laurie.
What the source says: from the article text — "Adaptive work is required when our deeply held beliefs are challenged, when the values that made us successful become less relevant, and when legitimate yet competing perspectives emerge." The article sets out six principles for leading adaptive work: getting on the balcony, identifying the adaptive challenge, regulating distress, maintaining disciplined attention, giving the work back to people, and protecting voices of leadership from below.
The trap, and it is a real one. This article has two lives. ERIC's bibliographic record gives the original: Harvard Business Review, volume 75, issue 1, January–February 1997, pages 124–134. The copy most people find and read is the December 2001 Best of HBR reprint, pages 5–15, reprint number r0111k — which is the copy we read this month.
Those are different page numbers for the same words. A citation reading "1997, p. 7" or "2001, p. 128" has mixed the two, and it is checkable in about forty seconds by anyone with library access. If you print a page number, print the one from the copy you actually opened, and say which printing it is.
What we did not verify: the technical/adaptive material in Heifetz's books is cited elsewhere in our work and is not re-verified in this pass, so it does not appear on this page.
What we say: psychological safety is associated with team learning behaviour, and learning behaviour is what connects it to performance.
What the source says: Amy Edmondson, Psychological Safety and Learning Behavior in Work Teams, Administrative Science Quarterly, 44(2), 1999, 350–383 (full text). The definition, at page 350: "a shared belief held by members of a team that the team is safe for interpersonal risk taking." The finding, in her words: "As predicted, learning behavior mediates between team psychological safety and team performance."
The specifics matter more than the headline. The study covers 51 work teams in a single office-furniture manufacturer of roughly 5,000 employees. 427 team members responded (86%). The external ratings come from 135 of 150 managers who received the teams' work (91%) — those are internal colleagues, not independent observers. And Edmondson states the limit herself: "The cross-sectional survey design prevented a demonstration of causality and also limited my ability to explore dynamic issues."
Why we print all of that: an AI draft will hand you "research shows psychological safety drives team performance." Two things have quietly gone — the single-site sample, and the mediator. The mediator is the part a programme designer actually needs, because it tells you what your programme has to produce. If your design targets psychological safety and never specifies the learning behaviours it is supposed to unlock, you have built the first half of the mechanism and hoped for the second.
What we say: the breakeven point is a useful frame for a transition programme. The six-month figure attached to it is not a measurement.
What the source says: Michael Watkins, The First 90 Days, Harvard Business School Publishing, 2003 (full text of the first edition). The definition: "The point at which new leaders have contributed as much value to their new organizations as they have consumed from it."
And the number, in full:
"When 210 company CEOs and presidents were asked for their best estimates of the time it takes a typical midlevel manager in their organizations to reach the breakeven point, the average of their responses was 6.2 months."
That is the mean of 210 executives' subjective estimates. It is not an observation of when managers broke even. Watkins says so plainly; the secondary summaries that circulate a clean "six months" are what strip the survey off.
How we use it: as a planning frame with the qualifier attached, in the sentence, every time. A number with its qualifier removed is a different number.
What we say: six leadership styles, drawn from a large executive sample, with climate as the mechanism between style and results.
What the source says: Daniel Goleman, Leadership That Gets Results, Harvard Business Review, March–April 2000, reprint R00204 (publisher record; full reprint text). The sample is "a random sample of 3,871 executives selected from a database of more than 20,000 executives worldwide." The six styles, in the article's own words: "Coercive leaders demand immediate compliance. Authoritative leaders mobilize people toward a vision. Affiliative leaders create emotional bonds and harmony. Democratic leaders build consensus through participation. Pacesetting leaders expect excellence and self-direction. And coaching leaders develop people for the future." On the mechanism: "climate accounts for nearly a third of results."
The negative finding, which is the reason this row is here. The claim that emotional intelligence accounts for 90% of what separates high performers is attached to Goleman's name constantly — including by drafting tools. It is not in this article. We asked the full text directly. What the article does report is that leaders with a critical mass of six or more emotional-intelligence competencies outperformed those without, with specific figures attached to specific divisions.
Worth one more note on precision, because it illustrates the whole problem: HBR's own summary page for this article says "3,000+ executives." The article says 3,871. Neither is wrong; they are different roundings of the same study, and if you cite the summary you have cited a number the article does not print.
What we did not verify: whether the 90% figure appears in any other Goleman publication. We can only tell you it is not in the one it is usually hung on.
What we say: the origin is Paul Hersey and Kenneth H. Blanchard, Life Cycle Theory of Leadership, Training and Development Journal, May 1969, pages 26–34, written while both were in the Department of Management and Organizational Behavior at Ohio University.
What the source says: we read the original article on ATD's own asset server — the publishing body's archive of its own journal. The central proposition: "As the level of maturity of one's followers continues to increase, appropriate leader behavior not only requires less and less structure (task) but also less and less socio-emotional support (relationships)."
The negative finding: the phrase "situational leadership" does not appear in the 1969 article. The authors write about situational determinants and about style varying with the situation, but the name the model is now sold under came later — and the two founders subsequently built separately branded models on top of it.
Why this matters in a design document: if your client's existing leadership material uses one of the later branded versions, match their terminology rather than blending it with the 1969 vocabulary. Merging the two under one name is the error, not using either.
What we say: there are four Kirkpatrick levels. Level 5 is not one of them.
What the source says: Jack J. Phillips of the ROI Institute, in his own paper updated February 2007: "The ROI Methodology adds a fifth level to the four levels of evaluation, which were developed almost 40 years ago (Kirkpatrick, 1975)." And on what it measures: "At Level 5, Return on Investment (the ultimate level of evaluation), the measurement compares the program's monetary benefits with the program costs." (source)
Why this row exists: AI-drafted measurement plans regularly contain a heading that reads "Kirkpatrick Level 5: ROI." Two models, two authors, one label — and a client who knows the difference will stop reading carefully at that point and start reading suspiciously. If you want to see what a measurement chain looks like when each link is named and owned, our Level 3 measurement plan worked through end to end is published in full.
A detail we are not resolving here: note that Phillips dates the four levels to 1975 in his own citation, while the four articles are usually dated 1959–60. We have not reconciled that in this pass, and there is a documented dispute about the framework's original authorship that deserves more room than a ledger row gives it.
What we say: Robert Kegan and Lisa Lahey, The Real Reason People Won't Change, Harvard Business Review, November 2001, reprint R0110E.
What we read: the publisher's record and its visible abstract: "It's a psychological dynamic called a 'competing commitment,' and until managers understand how it works and the ways to overcome it, they can't do a thing about change-resistant employees."
What we did not read: the full text. It is behind a paywall we did not pay, and two hosted copies failed this month — one returned 404, the other timed out on its robots file. So the fuller account of competing commitments as subconscious rather than merely unspoken, and everything about big assumptions, is absent from this row. It is a distinction we think is load-bearing for module design, and we are not going to assert it from a source we have not currently got open.
Three claims we intended to use and removed.
Kluger & DeNisi (1996) on feedback interventions. We hold a detailed internal verification of this from early September. A prior verification is not a current one — that is a rule we adopted after a citation we had recorded as checked turned out to be wrong — and no fetchable primary source or publisher record surfaced in this pass. Out until it does.
Locke & Latham (2006) on goal setting. Same position, same reason.
70-20-10. Not cut for doubt — the evidence against the popular framing is good — but the review we rely on could not be retrieved in this pass, and the claim is handled properly elsewhere in our writing. A half-sourced row here would be worse than no row.
Neither of these is caught by asking "does this source exist?", which is what almost all guidance on checking AI output tells you to do. Why that test fails in L&D is set out separately.
The reprint locator. A famous article gets reprinted, repaginated, and given a new reprint number. Both printings are real; their page numbers are not interchangeable. Heifetz and Laurie above is the clean example — 1997, pages 124–134, versus 2001, pages 5–15. Never print a page number from a copy you did not open.
The keyword false positive. You search the right book for a distinctive phrase from your quotation, the phrase appears, and you mark the citation verified. But the phrase can appear in a different sentence in that book while the sentence you quoted comes from another book entirely. This one cost us a correction: two quotations we had recorded as verified were traced to the wrong title, because the search term was present in the book we expected. Confirm the sentence, not the keywords. A keyword match in the right book is the most dangerous possible result of a citation check, because it returns a false positive with high confidence.
The commercial instinct is to publish only the verified rows and let the page read as a clean bill of health. That would be a worse page and, we think, a less persuasive one.
A document in which every claim is apparently settled science is one a careful reader stops trusting at the first claim they know is shakier than that — and then they doubt the whole thing, including the parts that were solid. Publishing the limits is what makes the rest legible. It also happens to be the only honest description of the work: checking attributions is an ongoing pass with a residue, not a state you reach.
Our twelve-week first-time manager design is published in the same spirit — free, in full, with the grounding stated module by module, including the modules where the honest answer was practitioner judgement rather than evidence.
Rows are re-verified rather than carried forward. When a row is re-checked, the month changes; when a source moves or a claim cannot be re-closed, the row moves to Limited or Cut rather than quietly disappearing. Nothing on this page was verified by a previous version of this page.
If you find a row that is wrong, we would rather hear it from you than from a client.
Most citation problems are cheaper to prevent than to fix. By the time a claim is in a design document it is load-bearing, and removing it means rebuilding a module. The work goes much faster when the diagnosis at the front of the engagement is sharp enough that you know which frame you are actually committing to before you write anything down.
If you want to test that end of it, the Diagnostic Conversation Coach is free, needs no signup, and takes about ten minutes. It runs the discovery conversation with you and shows where a client's stated problem diverges from the real one. If it does not sharpen the way you run a discovery call, nothing else here will interest you either.