EliteCart

A smart recommendation engine learns what your shoppers actually buy together, then surfaces those products in your cart and on your product pages. That is powerful, but it is not the whole story. The engine knows your purchase patterns. It does not know your margins, your overstock, your brand standards, or the products you would rather never show next to a $400 jacket. That gap is exactly where merchandising rules for product recommendations come in: the business logic you layer on top of automated suggestions so the engine works toward your goals, not just statistical correlations.

If you have ever looked at an algorithmic recommendation and thought "technically relevant, but I'd never show that there," you already understand why rules matter. The good news is you do not have to choose between automation and control. This guide breaks down the three lever types every merchant should know.

Why a Smart Engine Still Needs Your Business Logic

Recommendation engines optimize for one thing: products that tend to get bought together. That is a great start, but your store has goals an algorithm cannot infer from order history alone.

Margin and inventory. The data might favor a low-margin loss leader or a product you are phasing out, when you would rather move other items.
Brand and price positioning. A premium store rarely wants its hero product paired with a clearance bin item, even if the two genuinely sell together.
Catalog reality. Some products should never appear at all: gift cards, warranties, restricted goods, or core products you reserve for other placements.

None of these are failures of the engine. They are business decisions that live in your head, not in your data, and merchandising rules are how you write them down. To go deeper on the automation side, our guide to how AI-powered upsells outperform manual product recommendations covers when to lean on the algorithm and when to lean on curation.

The Three Lever Types

Almost every merchandising rule you will write falls into one of three categories:

Eligibility - which products are allowed to show at all.
Ranking - which qualifying products rank higher or lower.
Conditions - rules that only apply in certain cart contexts.

Think of it as a funnel. Eligibility decides who gets into the room, ranking decides who stands at the front, and conditions decide which rules even apply, depending on what is in the cart.

Lever One: Eligibility (Hard Rules)

Eligibility rules are absolute. They remove any product that does not qualify, full stop, no matter how strong the purchase correlation is. These are the rules you reach for when "never" is the right answer. Common examples:

Price floors and ceilings. Only recommend products above $15 so you never upsell a single sticker, or cap recommendations at a price that fits your average order value.
Price relative to the cart. Keep recommendations within a sensible ratio of what the shopper is already buying.
Category and catalog scope. Restrict recommendations to specific product types, vendors, or collections, or exclude a "Do Not Recommend" collection for gift cards, warranties, and reserved products.
Tag and title logic. Include or exclude products by tag or by keywords in the title, handy for keeping discontinued or seasonal items out.

Because these are hard filters, use them deliberately. Filter too aggressively and you can starve a slot of good options. The goal is to remove products that should genuinely never appear, not to micromanage every suggestion. For a deeper walkthrough, see how to control which products get recommended.

Lever Two: Ranking (Soft Rules)

Ranking rules are the nuance layer. Instead of removing a product, they nudge it up or down the list, so a qualifying product can still appear while you influence how prominently. This is preference rather than prohibition. Picture a simple scale that runs from strongly dampen, through dampen and neutral, up to boost and strongly boost. You apply that scale to attributes you care about:

Boost your bestsellers. Give proven sellers a gentle lift so they surface more often without crowding out fresh discoveries.
Favor a vendor or product type. Promote a house brand or a high-margin category without locking out everything else.
Steer by price direction. Lean slightly higher or lower depending on whether you are chasing order value or attachment rate.
Promote a collection. Quietly favor a seasonal collection during a campaign, then dial it back when it ends.

The art of ranking is restraint. A light boost keeps the engine's intelligence intact while tilting results toward your goals. A heavy-handed boost on everything cancels itself out and turns your recommendations back into a static list. Price-based ranking in particular deserves its own attention, which we cover in price-aware product recommendations.

Lever Three: Conditions (Context-Aware Rules)

The first two levers apply all the time. Conditions add context: a rule that only kicks in when the cart looks a certain way. The cleanest way to think about this is "if the shopper is buying X, then apply this rule." The trigger is whatever is already in the cart. So you might say: if the cart item is a camera body, only recommend lenses and accessories above a certain price. The same shopper buying a phone case would never see that rule, because the trigger does not match.

Conditional logic is what makes recommendations feel hand-curated at scale: show premium add-ons only when the cart already contains a premium product, apply a tighter price floor for high-end items than for entry-level goods, or segment by product line without building separate campaigns for each.

Conditions keep you from writing dozens of one-off rules. Instead of a separate setup for every product line, you write context-aware logic once and let the cart decide what applies. We dig into these patterns in if-then conditional product recommendation rules.

Putting the Three Levers Together

The three levers stack and evaluate in order: eligibility trims the candidate pool to allowed products, ranking sorts what remains by your preferences, and conditions switch specific rules on only when the cart matches. A practical starting point is a small handful of eligibility rules to keep out products that should never appear, one or two gentle ranking boosts toward your bestsellers or highest-margin category, and conditions only where a specific product line needs different treatment.

Resist the urge to do everything at once. Each rule makes the system slightly harder to reason about, so add them one at a time and watch how recommendations shift. The right number is the smallest set that gets your recommendations behaving the way a careful human merchandiser would. Pairing this with a strong reward bar strategy gives shoppers both relevant suggestions and a clear incentive to add them.

How EliteCart Fits In

EliteCart's Fine-tune EliteAI™ Ultra brings all three lever types into one place. You start from a pre-trained base engine, either EliteAI™ Ultra Original for frequently-bought-together pairings (best for accessories and add-ons) or EliteAI™ Ultra CrossCategoryBoost for complements from other product types (best for broadening the basket). From there you layer your own logic on top:

Filters are your eligibility rules: hard include or exclude logic by price, price-vs-cart ratio, tag, vendor, product type, collection, or title.
Boosts are your ranking rules: a five-way scale from strongly dampen to strongly boost, applied by tag, vendor, product type, collection, price direction, or popularity.
If/Then conditions are your context rules: apply a filter only when the cart's trigger item matches an attribute you choose.

Each version trains on your own order history and catalog, so it learns from your store specifically. Every save queues a fresh training run, and nothing changes for shoppers until you assign the version as the recommendation source on a surface: your cart, product page, two-step cart, or any checkout module. You can keep up to three live versions and archive the ones you are not using.

The Help Center covers the Fine-Tuning setup and the full Filters and Boosts reference. For background on how the engine learns, start with Understanding EliteAI™, or read the feature announcement.

Start with eligibility, refine with ranking, and reach for conditions last. A smart engine finds the patterns; your merchandising rules make sure those patterns serve your business. Layer them thoughtfully and your recommendations will feel less like an algorithm's guess and more like a deliberate choice.

Merchandising Rules for Product Recommendations: A Shopify Merchant's Guide