Hardware stores knew this in 1968.

There is a hardware store in a small town in the north of England that has been open since 1968. The owner, who took it over from his father, can tell you, without consulting anything, what kind of fixings hold up the timber on a particular kitchen extension three doors down from the pub. He sold them to the previous owner of that house in 1994 and remembers because the previous owner came back twice that year to ask about the same job.

The store has, on a clipboard behind the counter, the names of about a dozen people who are waiting on a particular kind of brass hinge that the supplier has been unable to source for six months. He calls them when it comes in. Not all of them buy. He calls them anyway.

The owner does not have a CRM. He has a relationship with each of the few hundred people who walk through the door, and a memory in which all of those relationships sit at once.

This is, when you describe it on paper, an extraordinary system. It costs almost nothing to run, it requires no software, and it produces customer outcomes that nine out of ten contemporary retailers cannot match.

The original promise of retail software, in the 1990s and early 2000s, was that it would give every retailer the working memory of that hardware store, and the patience of its owner, and apply both to a customer base of any size.

The promise was not unreasonable. A computer is, in principle, very good at remembering specifics. A computer should be able to know, on behalf of a tired founder shipping orders at midnight, that the woman buying a serum tonight is the same one who left a one-star review six months ago about a different product, and to handle that knowledge gracefully.

That is not what happened.

What happened, mostly, is that retail software became extremely good at counting things and extremely poor at remembering anything specific. The tools that ship today produce charts. They produce funnels. They produce cohorts and conversion rates and median order values. The actual people who left the actual sentences disappear into the aggregate.

A modern e-commerce dashboard knows almost nothing the hardware store owner knows.

The thing that is missing is simple, and unfashionable, and not yet a category. It is the willingness, on the part of a tool, to read the customer's sentence and remember it. To take the actual language a person used about an actual product, and carry that forward, into the next piece of marketing, into the next reply, into the next page about that product, into the next ad.

It is the difference between a system that knows you bought a hydration mist and a system that knows you bought a hydration mist because your skin was tight in the morning under SPF. The first is a database row. The second is a reason. The reason is the only thing that lets you do anything intelligent next.

The reason your customer bought is the most valuable signal you have. It almost never makes it into the software.

The hardware store owner does, every day, a particular sequence of small acts. He writes things down. He remembers what he wrote down. He uses what he remembered, the next time the same person walks in, without making a show of it. The customer feels recognised, not surveilled, because the recognition is in service of the customer, not in service of an upsell.

This is the loop. Write down. Remember. Use, quietly, on the customer's behalf.

The tools we have built for the last twenty years break this loop at the use step. They write down. They remember. They show the merchant a chart. They never use what was remembered to do something kind for the customer or something useful for the store, on either's behalf, automatically.

That is the part we are quietly working on.

We are not nostalgic about the hardware store. We are not trying to recreate it. We are trying to take the thing it does well (the loop) and run it at a scale and a quality the hardware store could never run it at, in software, on every store on the open web.

That is the engine.

You will hear about it in the summer.