Tuning Recommendations for Your Store Type
Every store is different. The recommendation engine works across all product categories out of the box, but you can get significantly better results by tuning a few settings to match how your customers actually shop.
This guide covers five common store types. Find the one closest to yours and use it as a starting point — you can always refine from there using your analytics dashboard.
Settings Reference
Before diving in, here’s what each setting controls:
| Setting | What it does |
|---|---|
| Preset | Overall strategy — balanced, discovery, or conversion |
| Personalization Weight | How much to favor individual shopper preferences vs overall popularity (0 = all popularity, 100 = all personal) |
| Discovery Weight | How aggressively to surface products outside a shopper’s usual categories (0 = strict relevance, 100 = maximum exploration) |
| Recency Bias | How much to favor recently added or recently popular products (0 = timeless catalog, 100 = only what’s new/hot) |
Fashion & Apparel
How customers shop: Visually driven, seasonal, trend-sensitive. Shoppers browse by look and outfit, not just individual items. Purchase cycles are frequent but lower AOV. Customers respond well to “complete the look” cross-sells.
Recommended Settings
| Setting | Value | Why |
|---|---|---|
| Preset | balanced | Fashion shoppers convert on both similar items and cross-sells |
| Personalization Weight | 60 | Style preferences are strong — a shopper who likes minimalist basics won’t respond to bold prints |
| Discovery Weight | 25 | Keep it moderate. Fashion shoppers have defined taste, but occasional trend-based discovery works |
| Recency Bias | 70 | High. Seasonal collections and trends matter. Last season’s inventory should rank lower |
Tips
- Enable the cart widget. “Complete the outfit” cross-sells (belt with pants, bag with dress) are the highest-converting placement for fashion stores.
- Connect your reviews integration. Fashion shoppers rely heavily on fit/quality reviews. Review sentiment directly influences recommendation scores.
- Tag products by occasion (casual, formal, workout) in addition to category. The category affinity rule uses your Shopify
product_typeand collection tags — richer tags mean better cross-category recommendations. - Checkout upsell: Works well for accessories and add-ons (socks, jewelry, care products). Keep checkout recommendations in the $15-40 range — impulse-friendly items convert best at this stage.
Books & Publishing
How customers shop: Search and browse driven, rarely visual. Readers explore broadly across genres but have strong category loyalty. Purchase cycles are frequent, ASPs are clustered in a narrow range ($10-30), and backlist titles stay relevant for years. Reviews are the primary purchase signal.
Recommended Settings
| Setting | Value | Why |
|---|---|---|
| Preset | discovery | Readers explore more than most shoppers — they want to find their next read, not just something similar to the last one |
| Personalization Weight | 70 | Genre and author preferences are the strongest signal. A mystery reader and a romance reader browsing the same bestseller page should see completely different recommendations |
| Discovery Weight | 45 | Higher than average. Readers frequently cross genres (a sci-fi reader might love historical fiction). Let the engine surface unexpected finds |
| Recency Bias | 20 | Low. A book published 5 years ago is just as relevant as one published last month. Don’t bury your backlist |
Tips
- Focus on product page recommendations. “Readers who enjoyed this also loved…” is the highest-value placement for books. The similarity and category affinity rules do the heavy lifting here.
- Don’t rely on price-range rules. Most books cost $10-30, so price-based filtering adds noise rather than signal. The engine already de-weights price range for narrower catalogs.
- Tag by genre, sub-genre, and theme — not just top-level categories. “Science Fiction” is too broad. “Hard Sci-Fi”, “Space Opera”, and “Dystopian” will produce much better affinity scoring.
- Connect a reviews source. For books more than any other vertical, ratings and review volume are the most trusted signal. A 4.5-star book with 200 reviews will naturally surface above a 4-star book with 10 reviews.
- Checkout upsell: Great for bookmarks, book lights, reading accessories, or companion titles in a series. If you sell series, the complementary rule will learn to recommend “Book 2” when “Book 1” is in the cart.
Electronics & Tech
How customers shop: Research-heavy, specification-driven. Customers compare extensively before buying. Purchase frequency is low but AOV is high. Accessories and compatibility matter enormously — the right cable, case, or add-on is an obvious cross-sell.
Recommended Settings
| Setting | Value | Why |
|---|---|---|
| Preset | conversion | Electronics shoppers arrive with high intent. Prioritize relevant, high-confidence recommendations over exploration |
| Personalization Weight | 40 | Moderate. Brand loyalty exists but product category matters more — someone buying a camera doesn’t necessarily want the same brand’s headphones |
| Discovery Weight | 15 | Low. Electronics shoppers know what they’re looking for. Irrelevant suggestions erode trust quickly |
| Recency Bias | 50 | Moderate. New product launches matter, but accessories for a 2-year-old laptop are still perfectly relevant |
Tips
- Prioritize the complementary / “bought together” placement. Accessories are the money maker: cases, chargers, cables, screen protectors, extended warranties. Tag products with compatibility info (e.g., “compatible-with-iphone-15”) in collections or tags so the complementary rule can learn these pairings.
- Use the cart widget aggressively. A customer with a $800 laptop in their cart is highly receptive to a $30 laptop sleeve. Cart cross-sells have the highest conversion rate in electronics.
- Checkout upsell: Perfect for small accessories. Protection plans, cables, adapters — anything under $50 that pairs with the main item.
- Suppress out-of-stock items (enabled by default). Nothing frustrates a tech shopper more than being recommended a product that’s backordered. Electronics inventory is spiky — keep this on.
- Consider disabling the homepage widget if your catalog is small (under 100 products). Electronics stores with narrow catalogs can feel repetitive. Focus on product page and cart placements instead.
Food, Beverage & Consumables
How customers shop: Replenishment-driven with strong brand/flavor loyalty. Customers reorder favorites frequently and add variety items impulsively. Bundles and multi-packs are common. Seasonal and limited-edition items create urgency.
Recommended Settings
| Setting | Value | Why |
|---|---|---|
| Preset | balanced | Mix of replenishment (personalized) and impulse discovery |
| Personalization Weight | 55 | Flavor/diet preferences matter (gluten-free, vegan, spicy), but impulse additions are common |
| Discovery Weight | 40 | Higher than average. “Try something new” works well for food — a coffee lover is open to trying a new roast. Variety-seeking is part of the fun |
| Recency Bias | 60 | Seasonal items (pumpkin spice in fall, summer BBQ sauces) should surface prominently when they’re available |
Tips
- Tag dietary preferences and flavor profiles in your product types and collections. “Vegan”, “Spicy”, “Nut-Free”, “Bold Roast” — the more specific your tags, the better the category affinity rule can match shoppers to products they’ll actually enjoy.
- Cart is your best placement. Food shoppers add impulsively. “Customers also added…” works incredibly well when someone has hot sauce in their cart and you suggest tortilla chips.
- Checkout upsell: Ideal for small add-ons — sample packs, single-serving items, or a discounted add-on to reach free shipping. Keep recommended items under $15 for impulse conversion.
- Use the homepage widget to feature seasonal, limited-edition, and new arrival items. With
recencyBiasat 60, these will naturally surface without overwhelming the core catalog. - Review integration matters. Food shoppers trust peer reviews (“this hot sauce is actually spicy” or “great flavor but too sweet”) more than product descriptions. Connect Judge.me or Yotpo to feed review sentiment into scores.
Home, Furniture & Decor
How customers shop: Project-driven and room-based. Customers are often furnishing a room or refreshing a space, so they buy in clusters (sofa + throw pillows + rug + lamp). ASP varies enormously ($15 candle to $2,000 sofa). Aesthetic cohesion matters — shoppers want items that “go together” visually, not just categorically.
Recommended Settings
| Setting | Value | Why |
|---|---|---|
| Preset | balanced | Mix of style-matched similar items and room-completing cross-sells |
| Personalization Weight | 55 | Style preferences are meaningful (mid-century modern vs farmhouse vs minimalist) but shoppers also respond to curated rooms |
| Discovery Weight | 30 | Moderate. Shoppers want cohesion within their aesthetic, not random suggestions from other styles |
| Recency Bias | 35 | Lower than average. A dining table doesn’t go out of style in 6 months. Timeless catalog items should rank alongside new arrivals |
Tips
- Tag by room AND by style. “Living Room” + “Mid-Century Modern” is far more useful than just “Furniture”. The category affinity rule uses collection tags — a shopper browsing MCM pieces should see more MCM pieces, not farmhouse decor.
- Enable both product page placements — “similar items” (same category, different options) AND “complementary” (cross-category, same style). A shopper looking at a sofa wants to see both other sofas and matching accent tables.
- Cart cross-sells should be lower-priced add-ons. Someone with a $1,200 sofa in their cart is receptive to $40 throw pillows, not another $800 piece. The price-range rule helps here — it naturally recommends items in a complementary price tier.
- Checkout upsell: Works best for small decor items — candles, vases, picture frames, care products (fabric protector, furniture polish). Stick to items under $50.
- Discovery weight at 30 is intentional. Home shoppers who are in “project mode” get frustrated by off-style suggestions. If your store spans multiple aesthetics, keep discovery moderate so the engine stays within the shopper’s demonstrated style preference.
General Principles
Regardless of your store type, these principles apply:
-
Rich product tagging is the single biggest lever. The engine can only use signals you give it. Detailed
product_type, collection membership, and Shopify tags make every rule work better. Aim for 3-5 meaningful tags per product. -
Start with the presets, then adjust. The three presets (
balanced,discovery,conversion) are tuned starting points. Use them for the first 2 weeks, check your analytics, then adjust individual sliders. -
Watch your analytics dashboard. The click-through rate (CTR) on each widget tells you what’s working. If cart recommendations have a 5% CTR but homepage is at 0.5%, double down on cart and experiment with homepage settings.
-
The engine gets smarter with data. Intelligence Level 1 (rule-based) works immediately. Level 2 (ML) activates automatically once you have enough customer interactions. Level 3 (embeddings) is available on the Scale plan. More traffic = better recommendations = higher conversion. Give it 2-4 weeks before judging performance.
-
Checkout recommendations are highest-intent. Every shopper who sees a checkout recommendation has already committed to buying. This is the highest-conversion placement in the app. If you’re on a Growth plan with Shopify Plus, enable it.