The sort order of your category pages is a revenue decision. Most retailers are making it by accident.
When I was running category strategy for a group of independent online retailers, sort order was never on the agenda. We'd spend hours on product selection, vendor margins, photography quality. But how products appeared on the category page? That was left to the platform default. Alphabetical order. Bestsellers list. Whatever shipped out of the box.
I didn't think of it as a decision at all. It felt like plumbing. Now I think it was one of the most expensive things we never examined.
Why the Default Sort Is a Revenue Leak
Alphabetical sort is obvious in its flaw. Nobody shops alphabetically. It exists because it's easy to implement, not because it reflects how customers think. But bestsellers sort is where the real damage hides, because it feels like it should work. Bestsellers are popular, right? Show people what's popular and they'll buy it.
The problem is that bestsellers sort is a population-level view. It answers the question: what has the most people bought? It does not answer: what is this person most likely to buy?
Those two questions can produce very different answers for the same category page. Take a home goods retailer with a strong repeat-buyer base. Their Bedding category has 80 products. The top five bestsellers, by volume, are all mid-price queen-size duvet covers in neutral tones. That's fine for the median first-time visitor. But a large segment of their customers are buying for toddler beds or for guest rooms or for gift purposes, and none of those customers are well-served by neutral queen duvets on top.
Bestsellers sort optimizes for one kind of shopper repeatedly. Everyone else has to scroll past products that don't fit before they find what they want. Many of them don't scroll. They leave.
What Purchase-Likelihood Sorting Actually Looks Like
The idea is simple: rather than ranking products by total units sold across all shoppers, rank them by likelihood of purchase for this specific visitor.
That likelihood score is built from purchase history. If this shopper has previously bought twin-size bedding, products in twin-size formats should score higher for them in the Bedding category. If they've bought products in a premium price band, premium SKUs should surface earlier. If they've bought items from a specific brand multiple times, that brand's new arrivals should move up.
None of this requires complex machine learning to describe at a conceptual level. The complexity is in the real-time execution: reading the right data, scoring all the relevant products, and rebuilding the sort order on each visit before the page loads. That's the part most retailers can't do without dedicated infrastructure.
The result is a category page where position 1 is the product most likely to resonate with this person specifically, not the product most people have bought in aggregate.
Before and After: A Home Goods Example
Take a retailer in the home goods space with roughly 40,000 monthly shoppers. Their Kitchen category had 120 products defaulting to a bestsellers sort. The top positions were occupied by a small group of consistent volume drivers: a cast-iron skillet, two sets of bamboo cutting boards, a knife set. Legitimate bestsellers. But those items were bought disproportionately by first-time visitors who came in through search or advertising.
Their repeat buyers, who made up about 38% of monthly sessions, showed a completely different purchase pattern. They were buying specialty items: fermentation crocks, beeswax food wraps, Japanese ceramic tools. Those products ranked in positions 40 to 80 in the default sort because their total unit volume was low. Most repeat buyers never found them on their own.
When the category page was resorted for repeat buyers based on their own purchase history, the products they were most likely to buy moved from the middle of page three to the top of page one. Add-to-cart rate on the Kitchen category for that segment went up by 22 percentage points within the first four weeks. That's not a small adjustment. That's a structural shift in how much of the category those shoppers were willing to explore.
We are not saying bestsellers sort is always wrong. We are saying it is a population average answer to an individual question. For shoppers who look like the average, it works fine. For everyone else, it surfaces the wrong products first.
Where This Approach Has Limits
Not every product in your catalog benefits from per-shopper sort ordering. Clearance items need to move regardless of who's looking at them, so price-based or urgency-based rules often take priority there. Seasonal promotions tied to vendor commitments may require fixed positioning. New arrivals sometimes need a guaranteed window of front-page exposure to collect enough purchase data before they can participate in a personalized sort.
A good category page sort strategy accounts for all of these as override conditions. The personalization logic applies where there's room for it, and static rules take over where business requirements demand specific placement. These two systems don't have to fight. They just need to be clearly separated so they don't accidentally nullify each other.
The failure mode to avoid is building elaborate personalization logic and then burying it under so many override rules that the sort order ends up being effectively static for most visitors anyway.
The Mechanics Are More Accessible Than They Seem
The reason most mid-size retailers haven't tackled this is not that they lack the data. Most retailers who've been operating for two or more years have meaningful purchase history for a significant portion of their repeat-buyer base. The reason is that acting on that data, in real time, per page load, requires infrastructure that wasn't part of their original platform setup.
Shopify, WooCommerce, and similar platforms give you the product data and the transaction history. They don't give you a per-session scoring engine that reads that history and rebuilds the category grid before it renders. That gap is exactly what we built Curated For You to fill.
The setup process is a one-time integration with your existing store platform. After that, every category page visit triggers a purchase history read and a sort rebuild. No manual rules to maintain. No CSV uploads. The model updates as purchase history grows.
A Simple Test You Can Run Today
If you want to get a sense of what per-shopper sorting might mean for your store, start by pulling your category page analytics by new vs. returning visitor segments. Compare the exit rates. Compare the add-to-cart rates. In most stores we've looked at, returning visitors have notably higher exit rates on category pages than first-time visitors, which is counterintuitive until you remember that returning visitors have more specific tastes and less patience for a page that doesn't reflect them.
That gap between new and returning visitor performance on category pages is the revenue that better sort order recovers. For some retailers it's small. For retailers with a strong repeat-buyer base, it's the most underexploited number in their analytics dashboard.
Sort order is not plumbing. It is the first merchandising decision your shopper encounters. Most retailers are letting a platform default make it for them.