The buyer who shipped twelve AI tools and is still a buyer.
Every couple of years, somebody offers me the other job.
The shape is roughly the same. A coffee, a senior person leaning back, a sentence that begins with you should be running. What follows is some version of analytics, data, technology, transformation. The seat being offered is the one you would expect a category buyer who learned to code in his thirties to want next. Head of something on the systems side. A CTO-shaped chair. A merchandising analytics director seat with a bigger team and a smaller fabric collection.
I have not taken any of them.
It is not for lack of options. The people who run those seats inside large retailers are mostly people I respect. They are well paid. The work is interesting in its own way. Some of them are friends. I have helped some of them write the spec for the work I would be hired to do.
I have stayed a buyer. This essay is about why.
The boring story most people expect
The standard arc for a buyer who can also write code is: they learn to write SQL well enough to argue with the analyst. They pick up Python on holiday. They build an Excel macro for the team. The macro becomes a workflow tool. The workflow tool becomes a system. The system needs maintenance. The maintenance is more interesting to them than the line review. They negotiate their way onto the data team. Two years later their business card says Head of Merchandising Analytics and they do not buy fabric anymore.
They are good at it. The data team likes them. The buying floor lost an operator who could read both the swatch and the regression. The new seat has a higher base, more direct reports, and quarterly reviews with people who use the word velocity without irony.
This is the boring story because it happens often, and because it is the story senior leaders push when they are trying to retain a technical buyer. The reasoning is sound, on its surface. You are wasted on the buying floor. You should be running something bigger. Let us promote you out of the noise.
I have heard this exact pitch four times. Once from a sister-company CEO. Twice from people sitting on the boards of retailers I had been quietly helping. Once from a recruiter who had not bothered to read the deck I had sent him about why I do not want what he was offering.
The pitch sounds like a promotion. It is actually a relocation away from the work.
Why the floor is where the work is
A retail business has two kinds of work. The work of operating the business. And the work of building the systems the business runs on. Most large companies separate these two kinds of work into two kinds of seat, and then spend the rest of their organisational lives trying to glue the two seats back together at the edge of every meeting.
The operating seat is the one with the customer in the room. The buyer is in front of the swatch. The merchandiser is in front of the open-to-buy. The store leader is in front of the cash register. These are the seats where the brand actually meets the world. They are also the seats where, after fifteen years, somebody hands you a different chair and tells you that you are too smart for this.
The building seat is the one with the engineer in the room. The data team writes the dashboards. The platform team writes the pipelines. The analytics director writes the strategy memo. These are the seats with the higher base, the cleaner promotions, the conference invitations. They are also the seats where, in most retail organisations, the customer is a row in a database and not a person walking into a store.
There is a small number of people who can sit in both seats at once. They are not interesting because they are a hybrid in the abstract. They are interesting because of where they put their attention. The hybrid who has chosen the operating seat keeps their attention on the customer. The hybrid who has chosen the building seat keeps their attention on the system. The two will look superficially similar on a deck. They are doing very different work.
I am the kind of hybrid who has chosen the operating seat. The systems are in service of the customer, not the other way around.
What changes when a buyer writes the system
There is a thing that happens when the buyer is also the person who writes the system. It is a quiet thing, and most retailers do not notice it has happened.
The spec is honest. A spec written by a buyer who has been on the floor for a decade does not have padding. It does not ask for a screen that nobody on the floor will use. It does not specify a metric that a buyer cannot actually see. It does not ship a workflow that requires somebody to copy a value from one tab to another. The spec is what a tired buyer at 9:30 on a Monday actually needs.
The model has the right priors. A model built by an engineer who has read the documentation will read the dataset literally. A model built by a buyer who has read the dataset and sat through three hundred line reviews will know which signals are real and which signals are weather. The buyer who has been overruled by a model and proven right at the next quarterly review knows where to put the buyer-veto-as-feature into the spec.
The feedback loop is fast. A standard cycle of floor complains → product manager listens → engineer prioritises → release plan → ship → wait for feedback takes months. A buyer who can write the tool herself cycles in days. Sometimes hours. The compounding effect over five years is not small.
The political economy is different. When the buyer writes the tool, the buyer owns the failure. The data team is not blamed. The engineer is not blamed. There is no inter-team blame surface for the failure mode to fall through. The buyer fixes the tool, ships the fix, and the floor sees the next release on Wednesday. The whole organisation gets faster at this kind of work because there is one accountable person instead of three.
The customer wins. The buyer is closer to the customer. The system is closer to the buyer. The system is therefore closer to the customer. This is the entire pitch, in one sentence, and it is the reason I have not taken any of the other jobs.
What I have given up
I would be lying if I said the choice has been costless.
There are people I came up with who are now running larger teams than I am. There are titles I do not have. There are conference panels with my peers on them where I am in the audience instead of on the stage. The seats I have refused had bigger base salaries than the seat I have stayed in. The version of me who took the merchandising analytics job four years ago has a different LinkedIn.
There is also the social fact that most senior leaders do not understand what I do, and a few find it threatening. A buyer who writes the tools is not a category the org chart knows how to evaluate. The HR function does not have a band for this. The compensation committee does not have a comparator. When the next round of leadership development comes, the buyer who writes the tools is sometimes left off the list because nobody can decide whether to send them to the buyer-track training or the tech-track training. They are between bands. The compromise is that they get neither.
These costs are real. They are also smaller than they look. The seats I have refused offered larger compensation and less interesting work. The titles I do not have were never going to be the thing that mattered. The conference panels were not where the actual work was happening. The shop floor was.
What I am not
I want to be careful here, because the framing of this essay can be misread.
I am not arguing that everybody who is in an analytics or data role should be a buyer. Some of the best operators I know are full-time analytics directors, full-time CTOs, full-time data scientists. They have made a different choice and they are doing different work. The work is good work. I am not denigrating it.
I am also not arguing that staying in the operating seat is harder than the building seat, or more virtuous. It is not. The work is differently hard. The trade-off is real.
I am arguing two narrower things. First: that the integrated operator is rarer than people think, and that retailers who have one should not treat them like a candidate for relocation. Second: that the integrated operator should be careful about taking the promotion-out-of-the-floor, because the promotion is usually a one-way door, and the work the seat does is more interesting than the seat that opens after.
A buyer who has shipped twelve tools and is still buying fabric is in the rarer place. That is the whole argument.
The compound interest of staying
The single most underrated reason to stay on the floor is the compound interest.
Year one as a buyer who codes: the macros work. Year three: the macros are now small tools that the team uses. Year five: the tools are the system the category runs on. Year seven: the system has a name on the floor. Year ten: the buyers who have joined the team in the last two years think the system is just the way buying is done here and they do not remember a time before it. Year twelve: the system has become the floor itself.
This is the thing that the lateral move into a data role costs. The compound interest stops. You have left the surface that was compounding and started a new surface, in a new org, with a new charter, and the compounding has to restart. Some people are willing to do that. I have not been.
The longer you stay, the larger the tool surface gets. The larger the surface, the more interesting the next tool you write. After twelve years, the tools are no longer about saving you three hours a week. They are about how the category itself operates. The next one will be about how the customer relationship itself operates. The one after that, I do not know yet. The point of staying is to find out.
What this is for
I write essays like this because I am sometimes asked to defend the choice and I am tired of saying the same things in private. The argument lives now on the web, where future versions of me can refer people to it instead of repeating the conversation.
It is also for the next buyer who is sitting in front of an Excel macro and wondering whether to learn the next thing. The answer is yes. The corollary is that the Excel macro and the next twelve years of macros after it are not preparation for leaving the floor. They are the floor.
And it is for any retail leader reading this and wondering whether a buyer-who-codes belongs in an operating seat or a building seat. The answer, in my experience, is that the seat that gets the most out of that person is the operating seat with the budget to also build. The integrated seat is more interesting than either pure seat. It is also harder to staff. The people who can sit in it are not common, and the ones who can are not always available, because they have made the same choice I have made and they are not actively looking.
That is the essay. The work continues this Monday. Always up for a conversation about a tool you are thinking of writing, or a seat you are thinking of refusing, or anything else on the floor.