Concepts referenced in this article, defined.
Run rigorous A/B tests and personalize every visit on Shopify or any storefront โ no engineers required.
Product recommendations are one of the most powerful personalization tools available to D2C brands โ and one of the most commonly implemented badly. A "you may also like" widget showing random products is noise. A recommendation showing exactly the right complementary product at the moment a customer adds something to their cart is a revenue driver. Here's how to do it right.
Most D2C stores have some form of product recommendation. Most of those recommendations underperform because:
They show the same products to everyone. Bestsellers are fine for new visitors who don't have a browsing history. For a returning customer who's already bought your top seller โ showing that same bestseller as a recommendation is useless.
They're placed where no one looks. Product recommendations buried at the bottom of a page below reviews, where 80% of visitors never scroll, have no commercial impact regardless of how well-personalized they are.
They recommend competing variants instead of complements. Showing a visitor who's about to buy the 50ml version that the 100ml version exists is a useful upsell. Showing them a completely different product in the same category is competition with their current decision.
The logic is wrong. "Frequently bought together" works well for genuine complementary products. It fails when it recommends things that happen to be bought together for unrelated reasons (e.g., two products people often buy for completely different family members in the same order).
How it works: Surfaces products that other customers have bought together. Best for: Complementary products where the pairing has clear logic (cleanser + moisturizer, supplement + shaker bottle, hair oil + shampoo) Use on: Product pages, cart page Limitation: Requires transaction history. New catalogs or low-volume products lack data.
How it works: Recommends products with similar attributes (category, ingredient, price range, tag). Best for: Discovery โ when a visitor is exploring options and hasn't committed Use on: Category pages, search results, "you might also like" on product pages Limitation: Can surface products that are too similar and create decision paralysis
How it works: Recommends products based on what this specific visitor has viewed or interacted with. Best for: Returning visitors who have browsed without purchasing โ surfaces products from their interest areas Use on: Homepage (returning visitors), email retargeting, push notifications Limitation: Requires behavioral data โ doesn't work for new visitors
How it works: Combines multiple signals โ collaborative filtering + behavioral data. What other customers bought together, filtered by this user's individual preferences. Best for: Established stores with sufficient transaction and behavioral data Use on: All recommendation placements Tools: Klevu, LimeSpot (Shopify), Wiser Notify
How it works: You manually define which products to recommend with each other. Best for: New stores without behavioral data, high-margin specific cross-sells, strategic bundling Use on: Product pages for hero products, landing pages Limitation: Doesn't scale with catalog size
See also: Dynamic Content glossary | Behavioral Targeting glossary | Audience Segmentation glossary
Placement 1: Product Page โ Below the CTA "Complete the routine" or "Frequently bought together" โ complementary products that pair with what the visitor is about to buy. This is AOV optimization. Keep to 2-4 products maximum.
Placement 2: Product Page โ After Reviews "Customers who bought this also loved" โ slightly further down the page for visitors who need more engagement before buying. Serves as a second chance to convert if the primary product doesn't fully convince them.
Placement 3: Cart Page โ Last-Minute Add-On "Add this to your order" with products under a low price threshold (typically below โน300-500). At this stage, the customer has committed to buying โ a small add-on at this moment has exceptionally high acceptance rates.
Placement 4: Cart Page โ You Might Have Forgotten Products related to what's in the cart that are often needed together. "Don't forget โ this pairs with [X]." Very effective when the pairing is practical and obvious.
Placement 5: Post-Purchase Confirmation "Your next step" โ the logical next product in a routine or journey. A customer who bought an entry-level supplement sees the next-tier or complementary product here. At this moment, purchase satisfaction is high and brand trust is at peak.
Placement 6: Homepage (Returning Visitors) "For you" product feed based on browsing and purchase history. Only relevant for returning visitors โ new visitors should see bestsellers or category featured items instead.
Placement 7: Email and Push Post-purchase email with "customers who bought [X] also love" โ or a personalized newsletter featuring products from each customer's preferred category. High CTR when personalized correctly.
See also: Real-Time Personalization glossary | First-Party Data glossary | Visitor Segments glossary
Personalized recommendations vs. generic bestsellers recommendations:
The gap widens for returning visitors with purchase history. For new visitors, personalized recommendations have smaller advantage over collaborative-filtering recommendations because behavioral data is sparse.
For wellness and supplements: Stack-building recommendations work extremely well. A customer who buys a gut health supplement is a natural candidate for a probiotic recommendation. Frame it as "complete your health stack" rather than "you might also like."
For skincare: Routine-based recommendations outperform generic cross-category recommendations. "Morning routine" (cleanser + Vitamin C + SPF) and "Night routine" (cleanser + retinol + moisturizer) recommendations should be pre-built as curated bundles.
For fashion and ethnic wear: Complete-the-look recommendations โ dupatta with kurti, accessories with dress โ have high AOV impact when relevant. Personalized by past browsing style (minimalist vs. embellished, casual vs. formal).
For food and grocery: Replenishment recommendations โ "You bought this 30 days ago, time to restock?" โ have very high conversion rates for consumables. Combine with a bundle discount to increase conversion further.
For Diwali/festive: Gifting bundle recommendations surface strongly during festive seasons. "Gift-ready โ beautifully packaged, can be delivered as a gift." Make recommendations gift-context-aware during Diwali, Raksha Bandhan, and wedding season.
Don't assume more recommendations = better. Test:
Use CustomFit.ai to run these tests on Shopify without developer involvement โ set up a recommendation module A/B test in minutes.