Wanting to Be Known, Selectively
In e-commerce, more traffic still mostly means more revenue. It sounds like something you would have said in the early days of online retail, when banner ads were novel and checkout funnels were short enough to memorize.
Today we speak in more elaborate language with attribution models, lifetime value, cohorts, marginal efficiency etc but the old arithmetic hasn’t gone away. If more people walk into the shop, more of them tend to buy something.
This simple idea sits in the background of product meetings and marketing reviews. We don’t argue with it directly. We just keep building things that assume it. Pages designed to catch search intent and navigation designed to make the right product easier to find.
All of it is, in one way or another, an attempt to invite more people in and gently guide them toward a decision.
If that sounds mechanical, the reality is anything but. The closer you look at how people browse and move through an online store, the more human and contradictory it looks.
What Filters Really Tell You
When a page reflects their preferences, maybe a color they like or a category they’ve browsed before, they linger.
And yet the same customers are deeply reluctant to hand over the raw material that makes this recognition possible.
We will enjoy a personalized homepage and still decline cookies. We want the benefits of intimacy without the obligations of disclosure. We want to be known, but only in the ways we approve.
So much of practical e-commerce work lives inside that contradiction. The fantasy of a fully mapped customer stitched together across sessions and devices into a coherent narrative belongs mostly to conference slides and vendor demos.
In ordinary life, you work with fragments.
Working With Fragments
This is where apparently dull things like product categorisation and faceted navigation begin to look more interesting.
They are often treated as housekeeping, something you configure once and rarely revisit unless something breaks. But in practice they are among the few places where customers tell you, in plain behavioral terms, what they care about.
“Only in my size.” “Under this price.” Each click narrows the story, which is why allowing users to combine multiple filtering options can make product discovery far more intuitive.
In an ideal world, all of this would be meticulously tracked. You would see exactly which filters were used and with what outcome.
But many businesses live far from that ideal because tracking is expensive to implement and maintain. What exists is often partial and/or delayed.
But even so, the task is to understand market demand. If you have data, you use it. Customer service teams hear the same questions again and again.
Keyword research then acts as a kind of further validation. It tells you whether the language you’re hearing internally maps onto the language people use in the wider world. It helps prioritize. It introduces a note of humility, too. Some things that feel important inside a business barely register in search demand, while other phrases, awkward, unglamorous ones, carry enormous weight.
Out of this usually comes a shortlist, a handful of attributes or subcategories that seem both meaningful to customers and significant in search behavior.
The temptation at this point is abundance. If five look good, why not twenty? The spreadsheet makes it all feel possible. But restraint is important. It is often better to implement a small number, observe what happens, and let the next iteration be informed by real behavior rather than enthusiasm.
Why Small Changes Carry More Weight Than We Expect
None of this is glamorous work. What makes it interesting is how directly it reflects changes in the customer journey.
I have been surprised more than once by how small changes alter behavior. A feature that staff mention constantly barely appears in search data. These moments are useful precisely because they puncture certainty. They remind you that your mental model of the customer is always guesswork, assembled from incomplete evidence and personal bias.
The personalization aspect never fully goes away, but let people reveal what matters through their actions rather than through forms.
Treat the signals they give (clicks, searches) as tentative hints rather than definitive truths. Not everything needs to be known to be useful.
Most organizations operate under constraints. Perfect tracking is rare. Learning to work around missing data by speaking to frontline staff and sampling behavior is your best workaround.
Many systems rely on flows of attention and participation. Many depend on people feeling understood without feeling exposed. If nothing else, it seems worth noticing how much of our work is really about interpretation, watching what people do, guessing gently at why they do it, and adjusting the environment accordingly.