Attribute ranking: merchandising in Adobe Commerce Optimizer changes scale

Deciding which products rise to the top of a category or a results page is one of the e-commerce gestures most directly tied to revenue — and one of the most manual. In most tools, the merchandiser pins, promotes or demotes products one at a time. The method holds for a catalog of five hundred references; it collapses at fifty thousand, multiplied by seasons, markets and languages. That is exactly the lock the September 2026 release of Adobe Commerce Optimizer breaks: merchandising rules now accept attribute ranking, in beta.

What the rule can do

The principle fits in one sentence: you pick a filterable attribute — brand, manufacturer, category, any text field in the catalog —, you list one or more values, and every product carrying one of them receives the chosen action. Three actions exist: boost (the product climbs), bury (it sinks), hide (it vanishes from results). For the first two, a slider sets the strength of the effect; a rule accepts up to twenty-five conditions, and the mechanism covers the tool’s three rule types — default product listings, category pages and search results.

The most interesting point is how it coexists with algorithmic ranking. When an attribute rule and intelligent ranking apply to the same page, the attribute action wins for the products it matches; the algorithm keeps ordering everything else. Manual pinning remains a separate tool for placing a specific SKU at an exact position.

From manual gesture to company policy

For a merchant, the change is one of kind, not degree. Bury the brand whose contract is ending, hide the colours withdrawn from the market, push the higher-margin private label during the campaign: these decisions existed before, but they translated into hours of handling, redone every time a product entered or left the catalog. Expressed as an attribute rule, they apply to the whole catalog and stay true: a product added tomorrow with the right brand inherits the treatment without any intervention.

The human-algorithm hybrid also deserves attention at board level. Intelligent ranking optimises what it sees — clicks, conversions — but knows nothing of your margin structure, your supplier commitments or the stock to clear before the sales. The attribute rule is exactly that: business intent laid on top of the algorithm, without switching it off.

The traps before you click

• The rule is only as good as the attribute. Values are entered as free text and are case-sensitive: a brand spelled two ways in the catalog half-escapes the rule. The built-in test pane is there to check; PIM discipline, meanwhile, becomes a revenue topic in its own right.
• A forgotten rule costs silently. A bury set up for a campaign and never removed keeps demoting products months later. Name your rules, date them, assign an owner for their review.
• Scope is frozen at creation. A rule is born in a catalog view chosen at the start and can never move to another: the split by brand or market is planned before, not after.
• It is a beta. Test on a non-critical view, measure the effect, then roll out.

The architecture reading

What remains is to place the novelty on the map. Optimizer is a decoupled SaaS merchandising service: the existing back office — Adobe Commerce, Magento or anything else — stays in place, the catalog is ingested through APIs, and discovery (search, categories, recommendations) is consumed over GraphQL, typically behind an Edge Delivery storefront. Two consequences for a CTO. First: this is now where merchandising capabilities land first, not in the 2.4.x monolith’s admin — one more signal about the product’s direction. Second: this layer can be tried without replatforming; it is a few weeks of work on a bounded scope, not a migration. The measure of success is simple: the merchandising hours saved, and the revenue generated by rules no team would ever have had the time to apply by hand.