Collection strategy: how sorting drives conversion

Every catalog we open has one: a product that outsells everything around it, parked at position 50 of its own collection page. Nobody demoted it. It just got old.

We used to treat collections like folders. Drop the products in, pick a sort order, get back to real work.

So when a good product went quiet, we blamed the product page. Thin description, tired photos, the usual suspects. The copy got rewritten. Nothing. The photos got reshot. Nothing. Meanwhile the same product kept selling whenever a search result or an email pointed straight at it, which should have been the clue, and for an embarrassing while it was not.

The product page was never the problem. Nobody was reaching the product page.

Where the clicks actually go

Then we looked at where clicks land on a collection page, and the folder idea quietly died. Products in the first 10 positions capture roughly 10x the clicks of anything sitting at position 100 or beyond. Not 10 percent more. Ten times.

A collection page is a ranked list, and ranked lists starve the tail. Shoppers do not scroll past position 20 in meaningful numbers, and no amount of product-page polish rescues an item nobody sees. Our quiet best seller had not lost its touch. Its own collection had buried it, one newer upload at a time.

Position 50 is where good products go to retire illustration 1

So the question stopped being which collections to have. The real question is who decides what sits in those first 10 slots, and based on what.

Two ways to fill the shelf

Shopify gives you two architectures, and they fail in opposite directions.

Manual collections are what they sound like: you place each product by hand and remove it by hand. Everything in the list is there because a person decided it should be, which is the good part. The bad part arrives at scale. A catalog of 50,000 SKUs cannot be hand-curated across every collection, and nobody should try.

Smart collections flip the tradeoff. Define conditions (vendor equals X, price under $Y, inventory above 100 units), up to 60 per collection, and membership maintains itself as product data changes. The catch is opacity: a product can land in a smart collection for reasons that take real digging to trace, a chore we file under spreadsheet archaeology.

Our take, after enough catalogs: go smart unless curation is genuinely part of your brand. Most of the time it is not.

The setting nobody touches twice

Filling the collection turned out to be the easy half. The half that touches revenue is the order things display in, and that is where our best seller had been sentenced.

Shopify exposes nine sort orders. Newest-first quietly promises the top slots to whatever you uploaded last, forever. Flip the same collection to best-selling and the products people actually buy move into the 10 positions that take the clicks.

Same inventory, same prices, radically different revenue exposure. Most stores pick a sort at launch and never look at it again. Hand-tuning it in the Admin is a hobby, not a merchandising system.

The jacket problem

One structural limit survived the sort fix. A collection is a flat list: one set of products, one axis.

Picture a women's apparel collection sorted by best-selling. The top item is a popular jacket. Fine, unless the shopper came for dresses, in which case the jacket ranking first tells her nothing. She wants the dresses together, and a flat list cannot do that.

Dynamic grouping is the restructure: group by product type, then rank by best-selling inside each group, so she reaches the dresses before she ever sees a ranked product. Bloomreach's documentation on group merchandising, a close cousin of this pattern, makes the case plainly: shoppers who can jump straight to the group they want, instead of scrolling one long list, convert more often.

That thinking became ARRAY, our dynamic grouping system: collections rebuilt as hierarchical, context-aware groups that follow how people browse instead of the merchant's flat taxonomy.

Position 50 is where good products go to retire illustration 2

What it turned out to be worth

Documented results across clients running ARRAY: up to a 45% conversion-rate lift. We are careful with that number. It comes from active retainer clients, and the gap is that wide mostly because the starting point, alphabetical or date-created sorting, is usually that bad.

The ongoing cost rounds to zero. Smart rules decide membership from live product data, and collectionReorderProducts rewrites positions in batches of 250 on a schedule, fed by live sales, so the best sellers hold the top slots without anyone opening the Admin. The labor moves upfront, into rules and testing. Then the system just runs.

What the log says

A collection is a visibility-ranking system wearing a category's name. The first 10 slots are the scarce asset; guard them accordingly.
Sort by evidence, not by upload date. Best-selling by default; newest-first only if novelty is genuinely your thesis.
Set the rules once and let live data hold the order. Hand-curated order decays, and noticing the decay is nobody's job.

Somewhere in your store there is a product earning the top slot every day and being denied it by a default setting. It is not underperforming. It is retired.

Un-retire it.

If you want to measure the gap between what sells best and what is actually visible on your store, that is exactly what an Infrastructure Stress-Test maps.