Good to Great — Jim Collins: COMPLETE

The design, quoted openly. The research team — more than twenty people over five years — started with the universe: the 1 435 companies that appeared on the Fortune 500 lists between 1965 and 1995. They then screened, with no theory behind them (the opposite of hypothesis-driven research: they let the data speak), for companies showing a transition point — a shift from persistent mediocrity to lasting greatness — measured as cumulative stock returns of at least THREE times the general market over FIFTEEN years after the shift.

The funnel: 1 435 companies, of which 126 moved on in the first cut, nineteen after tougher requirements — and after sector waves had been cleaned out (entire industries flying high did not count as proof of the company's greatness, but of the industry's) eleven remained. Eleven companies that had beaten the market by an average of around seven times over fifteen years after their leap — Circuit City a full 18,5 times.

The eleven: Abbott, Circuit City, Fannie Mae, Gillette, Kimberly-Clark, Kroger, Nucor, Philip Morris, Pitney Bowes, Walgreens and Wells Fargo. Note the sample's two methodological choices, which are gold for the analyst. First, greatness was measured EX ANTE in the share price, not in reputation — companies were not chosen for being famous or admired but for having delivered a measurable impossibility. Second, a transition point was defined: the leap is an EVENT in a time series, not a gradual honorary title.

That makes the question falsifiable: what happened in these companies at this date that did not happen in comparable companies? That is the comparison set's task — the next chapter. The research work: 87 interviews with key people, tens of thousands of pages of articles, everything coded without any prior theory of what would emerge — Collins calls the process chaos to concept.

The sample's pedagogical point for AK1A: the definition of good is the analysis's first decision. Collins defined GREAT as a number before he knew which companies were involved — three times the market over fifteen years after a leap — and then let the definition sort. Most investors do the opposite: they start with companies they like and then look for measures that glorify the choice in hindsight. The difference is the difference between research and halo — words the controversy chapter returns to, because Collins's design too has a built-in problem: selecting companies ON the outcome makes every pattern you subsequently find suspect.

Both halves of that insight — definition before outcome, and distrust of patterns in outcome-selected samples — are the course's first and most transferable straitjacket.

AKM1 bridge: AKM1 is the same design idea turned into a tool: twenty variables are scored on the same scales every quarter BEFORE any conclusion is drawn — the definition before the names, the measurement before the story.

The AK1TS lesson is the same: deterministic rules, the same data gives the same answer. The course's first line: research design is not a methods chapter — it is the discipline of analysis itself.

The direct comparisons: one company per great company, chosen from the same industry, with the same opportunities, comparable resources and starting positions — but which never made the leap. Upjohn versus Abbott. Silo versus Circuit City. Great Eastern versus Fannie Mae. Warner-Lambert versus Gillette. Scott Paper versus Kimberly-Clark. The Great A&P versus Kroger. Bethlehem Steel versus Nucor. RJR Nabisco versus Philip Morris. Addressograph versus Pitney Bowes. Eckerd versus Walgreens.

Bank of America versus Wells Fargo. The difference in outcome was not subtle: over the same period when the eleven beat the market by an average of around seven times, the comparisons performed in line with — or below — the market. Same industry, same winds, different results: it is the contrast that makes claims like level 5 and hedgehog meaningful: they were formulated as what SEPARATED the groups, not as what was present in one of them.

The third group — the six non-durables: Burroughs, Chrysler, Harris, Hasbro, Rubbermaid and Teledyne. They DID make the leap from mediocrity to greatness — but did not hold it for fifteen years. What they prove is the study design's finest point: transforming and SUSTAINING are two different problems with two different sets of causes. Rubbermaid is the group's icon: in the early 1990s celebrated — at one point ranked as the most admired company in the USA — only to then collapse into what Collins later calls the doom loop and be sold to Newell in 1999.

A company that was EXTREMELY good at inventing and launching — but whose transformation rested on adrenaline rather than discipline. The group foreshadows the whole course's dark chapter: how many of the eleven later fell (chapter 12) — the fallen in practice land in this group, just after the publication date.

And luck? Collins tested it explicitly — one of the project's most impressive moves. The question: were the eleven simply luckier? The answer in the data: no, they did not have more luck — they received roughly as much good and bad luck as the comparisons. The difference lay in what they DID with luck: they capitalized on the good and recovered from the bad, while the comparisons squandered the good and were crushed by the bad.

(Ten years later Collins formalized the idea in Great by Choice, 2011, as return on luck.) For the analyst this is the very keyhole: characteristics at a leap are one thing — but the luck-exploitation causal link is not something you can see in the rearview mirror without lying to yourself, which is precisely the objection the critics later raise: chapter 13 takes it to the maximum.

AKM1 bridge: AKM1's moat variables V13-V15 are at bottom a comparison question — best in the world means nothing without a number two. So always score V13-V15 AGAINST a named comparison set, not in absolute terms, and let V07-V09 be the arbiter: whoever is best must show it in margins (V07 gross margin, V08 EBITDA margin, V09 ROE) over time.

The AK1TS line: contrast measurement, not contemplation — measure both, and the difference is the analysis.