An ENTJ man externalises the means completely and the end not at all. He builds a machine that outlives him, and a machine has no purpose inside it, and thirty years on the machine is running perfectly in the service of something he never intended.
The first board meeting I sat in with Ivar Kilmartin, he spent nine minutes taking apart a report that his own company had been producing, in the same format, since 1994.
It was the quarterly supplier scorecard. He built it. Everyone in that room knew he built it.
“What’s the third column.”
Isak Nordvik said it was depletions per point of distribution.
“I know what it is. What is it for.”
Isak said it told us which suppliers were pulling their weight in the accounts we’d placed them in.
“That is what it says. It is not what it is for.” He had not raised his voice and he did not for the rest of the meeting. “I put that column in because we were carrying nine suppliers who were not calling on their own accounts and I wanted a way to make them do it without a fight. It worked and they all do it now and they have done it for twenty-two years. That column has been measuring a solved problem since about 2001 and we run the portfolio off it.”
Vikram Sethi asked what he would replace it with.
“I don’t know,” Ivar said, and that sentence is the whole of this piece, and I did not hear it for six years.
Ivar founded Kilmartin Distributing in 1979 at twenty-one, with a truck, a route, and nine accounts. He is a wide, still, very quiet man who is never the loudest thing in a room and is usually the reason the room ends where it ends. Small talk is not something he does and he does not apologise for it. He reads a document once. He has said the sentence what is it for to me perhaps two hundred times and not once have I heard him say it unkindly.
Kilmartin is a wine and spirits distributor in Copley — about two hundred and forty people, a warehouse, sixty trucks, and a book of roughly four hundred suppliers, from the multinationals down to a woman in the next county who makes nine hundred cases a year. In 2016 a larger group bought us.
I came in with that acquisition, as the buyer’s integration lead. I have been president of this division since 2019.
This building and the group both have a sentence for Ivar Kilmartin. It was admiring for about three years and then it turned into something else, and it has come out of my own mouth in the room where it decided things.
Nobody’s harder on that system than the man who built it.
The annual portfolio review is in three weeks. Isak has circulated a target for supplier count. And there are two pages on my desk from an analyst who is twenty-seven, which I have now read six times.
Vikram has one move for closing an item once Ivar has spoken.
“Noted. Where does that leave the number?”
The Book
A distributor’s book has two kinds of things in it and they are not the same business.
The first kind is the big houses. Everybody carries them. They are volume, they are margin dollars, they come with marketing money and quarterly programmes and a regional manager who will fly in, and roughly four cases in five that leave that warehouse are them.
The second kind is the small ones. Three cases a month. A cider from two counties over. A distillery run by two brothers. The woman with nine hundred cases. A slot in the warehouse, a line on a truck, and about nine minutes of a salesperson’s week each.
Examine any one of those on its own and it is a loss. Not marginal — an actual loss. The slot costs what a slot costs, the pick costs the same as a forty-case pick, and the salesperson’s nine minutes could have gone somewhere with volume in it.
Now here is the part that cannot be put in a column.
The reason a beverage director at a good restaurant takes your salesperson’s call is the second kind. Nobody has ever agreed to a meeting because you carry the third-largest vodka in the country; they can get that from anybody, including your competitor, who is also calling. They take the meeting because your person brings them something they have not seen, and they do that four times a year, and after six years of that your person is somebody whose call gets returned.
And then, on the strength of that returned call, the third-largest vodka in the country gets placed.
The access is created by the small book and spent on the large one. Every dollar of it lands in a line item belonging to somebody else.
Anselma Ruiz makes about nine hundred cases a year two counties over and has been on this book since 2004, which is longer than three of our five largest suppliers.
I met her at a trade tasting in my second year and asked, because I was new and it was a fair question, how she had got on the book.
“I called them,” she said. “Everybody else wanted a thousand cases and a marketing budget. Mr. Kilmartin came out and walked the rows and asked me what I’d be making in ten years.”
I asked what she’d told him.
“I said I didn’t know. And he said fine, and put me on for six.” She was pouring for somebody else while she talked to me. “Six cases. That’s the whole order. Twenty-two years.”
She is on the current cut list. She scores in the bottom decile and she has scored in the bottom decile every year since we built the model.
The warehouse runs a slot for every item and the two-case order costs the same to pick as the forty. Sal Vaccaro has run that floor since 1991 and can tell you, without looking, which twelve slots have not moved in a quarter, and has never once suggested cutting one.
I asked Sal Vaccaro about it once, standing at the pick line.
“You know which twelve haven’t moved,” I said. “Why don’t you ever say anything?”
“Because I know what they’re for.”
I asked him what they were for.
He thought about it and then said, “You’d have to see it from a truck,” which is no kind of answer, and which took me three years to understand is the same answer Ivar gives, in a different vocabulary, by a man who left school at seventeen.
I do not drink. Not for any reason worth writing down. I have been in this industry for nineteen years and it comes up at every dinner and I have a line for it that works about half the time.
The Machine
What Ivar does is not management, which is a word for keeping something running.
He builds apparatus.
He takes an intention and constructs a thing in the world that will keep producing the intended result after he has stopped paying attention to it. Not a decision, not a policy — a mechanism, with parts, that runs.
I have been through the operating documents of this division and I can list them. The supplier scorecard, 1994. The route structure, which he rebuilt in 1997 around drive time rather than geography and which is still the model the whole group uses. The commission plan, 2001, which pays on placements in the first year and volume thereafter and which is the reason this division’s salespeople open doors instead of milking accounts. The warehouse slotting rules. The credit policy. The midweek order cutoff. The supplier review calendar.
Every one of those still runs. Not a modified version. The thing itself, in 2026, thirty years on, in a company he no longer operates, inside a group that bought him.
The second thing supplies the intention.
He looks at a market and sees a position in it, ten and fifteen years out, and works backward. Sentiment had nothing to do with the small book and it was no favour to anybody. It was a strategy, formed sometime in the middle eighties, that access to on-premise accounts would become the scarce thing in this business as consolidation ran, and that the way to own access was to be the house that carried what nobody else would carry.
He was right. He has been right about this market for forty years running.
Six years I sat across the table from that before the obvious part of it landed on me.
A machine has no purpose inside it.
He externalised the means completely — thirty years of apparatus, running perfectly, transferable, documented, teachable. And he externalised the end not at all. The end stayed where it came from, which is inside a man’s head, and it has no derivation, and it never got written down anywhere because the only form it could take was a machine, and a machine cannot hold it.
Ilka Rondeau built the same kind of apparatus at a distributor two states east and her route model is used in nine markets. What gets said about Ivar is that he built this. What gets said about Ilka, when her company gets talked about, is that they have a very good system there. What that same machine-building gets a woman running a house like this, and who is presumed to have authored it, gets its own page opposite this one.
Two hundred and forty people. Nine on the division leadership team. Three are women.
What I Cut
Between 2019 and 2021 I took the supplier book from four hundred and sixty items to three hundred and ten.
How that was done matters, because it was not crude and nobody was cavalier about it.
We built a model. Contribution margin per slot, adjusted for pick cost, weighted for growth trajectory and supplier support. Every item scored. We reviewed the bottom quartile line by line with the sales leadership in the room, and about forty of them we kept on judgement despite the score, and I put my name on all forty of those exceptions.
Ivar sat in on one of those bottom-quartile sessions, at my invitation, and said almost nothing for two hours.
At the end I asked him for anything he wanted to add.
“The forty you’re keeping on judgement,” he said. “How did you pick them?”
I said the sales leadership had flagged them.
“On what basis?”
I said on judgement.
“That’s the whole of what I’d have said too,” he said. “I’m just noting that we’re calling it judgement when it’s forty and a factor when it’s four hundred.”
The result was very good. Gross margin per case up nine per cent. Warehouse throughput up. Days of inventory down by nine. Two of the group’s other divisions have since run the same exercise off my model.
Nobody cut anything out of ignorance. Every single item we dropped was examined, individually, against a defensible standard, and each one of them failed it.
That is the thing I would like a reader to sit with, because I have been sitting with it since the spring. The analysis had nothing wrong with it. It was correct on every line. The book had a property that was not on any line.
Put It in the Standard
In 2020, halfway through the rationalisation, Ivar came to my office.
He did not object to the model and he did not defend any particular supplier. He said the exercise would work and that in about six years we would find we could not get into the accounts we used to be able to get into.
I asked him to help me build it in. I meant it. I was not managing him.
“Tell me what to weight,” I said. “Give me the factor and I’ll put it in the model.”
He said there wasn’t a factor.
I said there had to be something — that if carrying a small supplier produces access, then access produces placements, and placements are measured, so somewhere there is a relationship we could estimate.
“You can’t attribute it,” he said. “The placement lands on the brand that got placed. The thing that opened the door was three years ago and belonged to somebody who isn’t in the account anymore.”
And I said the sentence.
“Then put it in the standard and I’ll adopt it. Anything you can write down, I’ll build.”
He sat there for a while. Ivar is not a man who fills silence.
Then he said, “I can’t write it down. Whether it is there is a separate question.”
And I said — reasonably, professionally, in the voice of a person doing her job — that I could not run a two-hundred-million-dollar book off something nobody could state.
He said, “No. You can’t.”
And he got up, and he has never raised it again in six years, and in each of those six years he has come to the portfolio review and sat through it and asked one or two questions and voted with the board.
I have replayed that exchange more than anything else in my career. Every word I said was correct. The request I made was a completely fair request that was structurally impossible to satisfy, and when he could not satisfy it, I took that as the answer, and so did everyone else in the room afterward when I described the conversation.
Two Territories
In 2022 the group ran a pilot.
Two of our nine territories went to a further-reduced book — the group’s proposed national core list, about two hundred and forty items, no local exceptions. The other seven stayed as they were.
The stated purpose was to test warehouse and route efficiency at a tighter assortment, and on those measures it worked. Cost per case in the two pilot territories came down about six per cent and stayed down.
Eighteen months in, somebody noticed a variance.
New on-premise placements of the large brands — the national houses, the ones the whole exercise exists to sell — were down in the two pilot territories against the seven control territories. Not dramatically. Down about fourteen per cent on new accounts opened, against a market that was flat.
Isak asked for an explanation of the variance and the explanation that came back was salesperson turnover in one of the two territories, which was true, and which accounted for maybe a third of it.
The rest went into a footnote and the pilot was declared successful and the group is now proposing the core list nationally, which is what the target Isak circulated is.
Rafferty
He is twenty-seven, came to us out of a supply chain programme, and does the group’s commercial analytics for this region.
I gave him the variance to close out in the spring, as a housekeeping item, because Isak wanted the footnote resolved before the national rollout.
He came back with two pages I did not ask for.
What he had done was take eleven years of account-level data across all nine territories and build a lagged model: for each on-premise account, the number of small-supplier items placed in that account in years one through three, against the probability of a large-brand new placement in years four through six.
The relationship is there. It is not enormous and it is not noise. An account that took two or more small items in a three-year window is something like sixty per cent more likely to take a new large-brand placement in the following three, controlling for account size, cuisine type, and salesperson tenure.
He wrote a paragraph at the end saying the mechanism is presumably relationship access and that he could not test it directly and that somebody should.
I asked him what made him look at it that way.
“The pilot’s a natural experiment,” he said. “You cut the small book in two places and left it in seven. That’s the cleanest design anybody’s going to get. It’d be a waste not to use it.”
I asked whether anybody had suggested it to him.
“No. Mr. Kilmartin said something at the review last year about doors and I didn’t know what he meant, so I went and looked.”
Ivar Kilmartin has been saying that sentence in that building since 1985 and could not write it down, and a twenty-seven-year-old with eleven years of transaction data and a different set of instincts wrote it down in about nine working days, twenty-two years after the thing it describes stopped being defensible.
The Profile
From 1996 to 2014, every candidate for a sales position at Kilmartin took a personality questionnaire.
Ivar bought the licence in 1995 after a run of bad hires. Sixteen boxes, four letters, a scored profile, and a target pattern derived from the five best salespeople this company had at the time.
They hired against it for eighteen years. Candidates who did not fit the pattern were, in the language of the hiring file, not a profile match, and there is a stack of those in the archive, and I have been through it.
The group’s employment counsel made us stop in 2014, which is the correct outcome and which I would have insisted on myself, because using an unvalidated personality questionnaire to screen job applicants is a liability and it does not measure what it is being asked to measure.
Ivar took it himself, in 1996, to validate the profile. His sheet is filed with the candidates because that is where it went.
It says ENTJ.
The target pattern they hired against for eighteen years does not resemble it at all. The pattern was built off five salespeople in 1997 and every one of the five sat in the two boxes you would expect a good salesperson to sit in, and the man who founded the company, built every system in it, and picked the strategy that made it worth buying would have been screened out of his own sales force.
The material worth reading sits at the back of the manual. Four letters, it says, are an order of operation: one habit in the lead, a second feeding it, and a last pair that turn up only when a man goes after them on purpose and gives them time. The order, for him, gets laid out where the type is handled alone, from the sixteen the manual was printing off.
At the head of that order is the construction of apparatus — the turning of an intention into a mechanism that runs without him. Executing well is a different and also rare thing and this is not it. It is the making of something that keeps producing the result after the maker stops looking at it.
Supplying it is the picture: a market position at fifteen years, arrived at whole, without a derivation, correct across four decades at a rate nobody has computed.
The last pair are what would have saved the second thing.
One is the immediate concrete present. Ivar has never been good in a room, on a call, at a table with a customer. He knows it and says so. Sal has closed more accounts in a year than Ivar closed in ten and Ivar has said that out loud at a sales meeting.
The other is a private language of value — the register in which a person says this is the kind of house I set out to build, and that matters to me, and here is what I mean by it. He has never had it. So when the machine turned on the purpose, the only register available to him was an argument about outcomes, and an argument about outcomes requires exactly the attribution the purpose does not have, and so he lost, correctly, on the merits, in the only language he owns.
That is the whole shape of it. He exported the means completely and the end not at all, and thirty years on the means are running perfectly in the service of something he did not intend, and the only person who knows what they were for is the one man in the building whose objection sounds like a founder defending his legacy.
I built the model, so this part is mine. If an employer puts a personality questionnaire in front of you as part of hiring, that is a screening instrument, screening instruments carry legal obligations, and it is a fair question — in writing — to ask what is being measured, whether it has been validated for the job, and how the result is used. And from my side of the table: when you are about to cut something whose value cannot be attributed line by line, the thing to ask for is a counterfactual rather than a better line — a pilot, a holdout, two territories — proposed in writing before the cut, because afterward nobody will fund the measurement and the absence of evidence becomes the evidence. Temperament has nothing to do with either of those. The people for them are your own attorney, and, for the second one, whoever signs your analytics budget.
The manual itself said in 1998, in its own front matter, that the instrument had no place in selection, and that a person answering it again a month later would come out somewhere else roughly one occasion in three.
We used it for selection for eighteen years, and the first person it would have rejected paid for the licence.
Three Weeks
The review is in three weeks and the target is a hundred and ninety items, which is a cut of a hundred and twenty from where we are.
Rafferty’s two pages are on my desk and they are not enough on their own. The effect is real and the mechanism is inferred and the group will want the mechanism tested, and testing it means holding a book open in some territories for four years, which is a real cost against a certain saving.
This spring took two things apart.
The board meeting in 2019. I have put that one into two group offsites as a story about a founder who audits his own work — that column has been measuring a solved problem since 2001 — and it always lands, and I have used it to make a point about intellectual honesty. What that room contained was the only person alive who knew what any of those instruments were for, going through them one at a time, in public, looking for the ones that had come loose from their reasons. There was no defence of anything in it. He was doing maintenance nobody else in the company was equipped to do, and we heard it as a man being hard on his own system, which is a compliment, and compliments do not require an answer.
Then my office in 2020, and put it in the standard and I’ll adopt it. I have carried that for six years as the moment I was fair to him. It was fair. It was also a request that the purpose of a thirty-year strategy be rendered in the language of the machine built to serve it, made to the one man who had no other language, and his failure to do it went into my account of him and into everybody else’s.
What that sheet left out. He has been estranged from a brother who was in this business since 2004 and will not say why; money has surprised two accountants of his; and he goes to the funeral of every retired Kilmartin driver, in three states, and has never once mentioned doing it.
His board seat runs out in the autumn. He has not asked to be renominated and nobody has asked him whether he intends to.
I called him about Rafferty’s pages, because he should hear it from me.
He listened to all of it and did not interrupt, and then asked one question, which was how many years of data it took.
I said eleven.
“Then we could have had it in 2007,” he said.
The people in these essays are composites drawn from long observation. The theory is not.