I have spent more hours than I would like to admit inside a Pmax dashboard, watching an algorithm decide who gets shown what. And here is the thing nobody in martech wants to say out loud: the algorithm does not have taste. It has data. Those are not the same thing, and anyone who has seen the power of curation actually work already knows it.
This is not only a luxury conversation, even though luxury makes the clearest case study. Every retail business, whether it sells couture or basics, has people who already know things the customer does not. What pairs well. What lasts. What is actually worth buying instead of what just photographs well. Good curation is how a customer discovers that expertise, and that is exactly why taste is a stronger nudge to purchase than a generic recommendation will ever be. It carries someone's judgement, not just a probability score.
For decades, that judgement was delivered by people who curated obsessively. What goes in the window. What the editor puts on page one. What the visual merchandiser decides you see the second you walk in. Now most of that decision has moved online, into a recommendation engine that was never trained to understand desire, only behaviour. This is not a case against martech. I have built my career on it. But it is a question every retail brand needs to sit with, luxury and essentials alike: when a machine takes over curation, does curation disappear, or does it just get better disguised?
Algorithm is not an editor
Retail has always run on tastemakers, luxury most visibly of all. Someone with a trained eye decided what made the cut and what did not, and that decision was final. It was not about what would sell the most units. It was about what deserved to be seen, and what the customer would only find because someone with expertise put it in front of them.
Scroll through any shopping app today and you are looking at a different kind of decision maker. An algorithm reads your browsing history and your past purchases, compares you to millions of people who behave the same way, then serves up whatever you are most likely to click. That is not curation. That is optimisation, and it is optimising for a completely different goal: conversion, not meaning. The next shift makes this more interesting. AI shopping assistants are starting to build a customer's shortlist before that customer ever lands on a brand's website. If that becomes the norm, brands are not just competing for attention anymore. They are competing to be legible to an algorithm that decides who gets a shot at that attention in the first place.
Abundance kills want
Desire runs on friction. Hype, waitlists, limited drops, the fact that you cannot always get what you want immediately. All of it stirs want. Martech runs on the opposite instinct: strip out every bit of friction between a customer and checkout. Reconciling the two is the real job right now.
There is also a difference between predicting desire and creating it. An algorithm is very good at telling you what you already want, based on what you and people like you have already done. I like to say “I don't know what I don't know.” An algorithm cannot say that. If it only ever shows a customer what they already like and are already familiar with, that customer will never discover something they might have liked just as much, maybe more.
That has never been how great curation worked. The entire point of a great creative director or editor was making you want something you did not know existed yet. Influencers still do that today. Websites optimise instead. Optimisation cannot manufacture desire. It can only refine what has already worked before.
This gets riskier when a trend spikes for the wrong reasons. Algorithms optimise for engagement, not prestige, so if a product goes viral because people are mocking it, the system reads that outrage as relevance and pushes it further. A brand that chases the spike is trading long-term equity for a short-term metric, and fashion was never meant to work like that. Fashion is supposed to be a long-term record of culture. Algorithms are built for right now. Chase them too closely and years of heritage get flattened into a string of disconnected viral moments.
Not a Zero-sum game
Before this turns into a simple algorithm-bad, human-good story, it is worth being honest about the limits of human curation too. No single creative director, however brilliant, can manually predict what a customer in Seoul wants versus a customer in Sao Paulo versus someone scrolling at midnight in Delhi. That scale of personalisation was never a realistic human job.
Pure manual curation also creates a bottleneck. Every extra layer of human decision making between a product and a customer is one more thing slowing the journey down, and today's customer has close to zero patience for a slow journey. The nostalgia for one all knowing tastemaker is understandable. It is just not a real solution. It is a different kind of failure.
Man & Machine make Magic
The answer is not picking a side between the algorithm and the human. It is deciding what each one is allowed to touch.
Let the machine own the background: discoverability, personalisation at scale, matching the right customer to the right product from a catalogue too large for any person to hold in their head. But the brand still has to build that catalogue with intention. That means turning heritage, craftsmanship and aesthetic codes into something an algorithm can actually read, sometimes literally, through structured data like Digital Product Passports that carry a product's story alongside its size and price.
Think of it less as one curator trying to serve everyone, and more as a brand building an entire universe of curated rooms. The algorithm's job is recognising which customer belongs in which room and guiding them there. The human's job is making sure every room is actually worth walking into. That is identity led curation. Thebrand keeps the aesthetic authority. The machine handles the routing.
Taste is the real flex
Here is where this is heading. As more of the market becomes fully automated (fast fashion already runs almost entirely on optimisation), the brands that hold on to unapologetic, non-optimised human judgement will stand out simply by being different. Not everyone can move at algorithmic speed, and not everyone should want to.
The martech tools I use every day are very good at execution. Targeting, personalisation and scale: they have that covered. What they cannot do is decide which ideas are worth pursuing in the first place. That judgement call, deciding what deserves to be seen rather than just what is likely to be clicked, is the one thing that will not get automated. For retail, luxury and essentials alike, that might end up being the most premium thing left to sell.
Related Articles
