Cross-sell is the lazy cousin of upsell. Most Shopify Plus brands run it, but few run it well. The widgets are everywhere. The conversion is barely measurable.
This is the guide we walk merchants through when cross-sell is the next AOV lever. What it actually is, where it lives in the funnel, which tools win at which scale, and the common reasons it stops converting.
Cross-sell vs upsell: what the difference actually is
Upsell offers a higher-value version of what they are already buying. Bigger size. Premium tier. Subscription instead of one-time.
Cross-sell offers a complementary item that pairs with what they are buying. Cleanser plus moisturiser. Shoes plus laces. Bed frame plus mattress.
Mechanically the difference matters because:
- Upsell competes with the original choice (you replace product A with product A+)
- Cross-sell adds to the original choice (you keep product A and add product B)
Both can run on the same surface. Both can run at the same moment. But the offer language, the visual treatment, and the conversion math are all different. For the upsell side of the picture, see our upsell playbook.
The four cross-sell surfaces
PDP
"Frequently bought together" blocks on the product page. Shopper has not yet added to cart. You are suggesting items that pair with what they are looking at.
Conversion math: 1-4% acceptance. The lowest of any surface because the shopper has not committed yet.
Best for: high-consideration products where category context helps. Furniture, skincare regimens, photography equipment.
Cart drawer or cart page
Complementary product slot in the drawer. Shopper has committed to at least one item. You are suggesting an add-on.
Conversion math: 5-10% acceptance. Higher than PDP because intent is concrete.
Best for: low-cost add-ons (under $20), accessories, consumables that pair with the main item.
Post-purchase page
One-click add-on offer between checkout completion and thank you page. Shopper has paid. No payment friction.
Conversion math: 8-14% on complementary items (vs 10-15% for direct upsells).
Best for: items the shopper might have forgotten. Batteries with electronics. Care kits with leather goods.
Post-purchase email
Follow-up email 2-7 days after delivery with complementary product suggestions. Shopper has received the original item. They know what they bought.
Conversion math: 1-3% on the email itself, but compounds with repeat purchase rate.
Best for: consumables that need replenishment, accessories the shopper realises they need after using the main product.
The three approaches: rule-based, algorithmic, manual
Rule-based
"If shopper adds product A, suggest product B." You write the rules manually. Tools include Shopify metafield-driven recommendations and most third-party rule builders.
Pros: predictable, explainable, no machine-learning weirdness.
Cons: scales poorly past 200-300 SKUs because the rule set explodes.
Best for: small catalogues, hero products, manual curation of "best with" pairings.
Algorithmic (AI / ML)
Engine looks at purchase data and recommends items statistically likely to be bought together. Tools include Searchspring, Klevu, Algolia Recommend, native Shopify product recommendations.
Pros: scales to any catalogue size, learns from real shopper behaviour, handles long-tail SKUs.
Cons: needs enough data to train on (typically 10K+ orders to be useful), recommendations can drift weird if catalogue is mixed, harder to debug when wrong.
Best for: catalogues over 500 SKUs, brands with mature order volume, anywhere personalization compounds.
Manual curation
You hand-pick the cross-sell pairings per product. Often via product metafields or a dedicated app field. The front-end pulls the curated suggestions.
Pros: highest relevance, brand-curated, no algorithm weirdness, explainable.
Cons: labour-intensive, does not scale past a few hundred products without dedicated staff.
Best for: brands with strong category opinions (fashion, beauty, lifestyle), under 500 SKUs, where the brand voice matters in the recommendation.
The tool landscape
Native Shopify product recommendations
Free. Decent baseline. Uses Shopify's own ML on order data. Best as the default if you do not have a strong reason to switch.
Searchspring
Search plus merchandising plus recommendations. Strong rule builder. Best for catalogues that need search-and-recommendation as one stack.
Klevu
AI-driven search and recommendations. Particularly strong on natural-language understanding and personalization. Best for brands where shopper intent is varied.
Algolia Recommend
Add-on to Algolia search. Recommendation engine trained on click and purchase data. Best for brands already on Algolia for search (we use it on most of our Hydrogen builds).
Glood, LimeSpot, Wiser
Mid-market alternatives that do cross-sell-specific recommendations without the search-stack overhead. Lower price point, lower ceiling.
When each approach wins
Three quick rules:
- Under 200 SKUs and brand voice matters: manual curation
- 200-1,000 SKUs with order volume above 5K/year: rule-based with algorithmic fallback
- 1,000+ SKUs or fast-moving inventory: algorithmic (Klevu / Algolia Recommend / Searchspring)
Most brands at scale end up with a hybrid: manual curation on hero SKUs, algorithmic recommendations for the long tail.
Cost framework
Three approximate ranges:
Native plus light app (sub-$10M GMV): $0-100/month. Shopify recommendations plus a single cross-sell app.
Dedicated engine ($10M-$50M): $500-3,000/month for Klevu / Searchspring / Algolia Recommend. Includes implementation budget of $5K-$20K for properly tuned cross-sell rules.
Full custom ($50M+): $3,000-10,000/month in tooling plus engineering time for custom training, segment-specific recommendations, and integration into the headless storefront. For headless cart customisation specifically, see our cart customisation guide.
Common pitfalls
Irrelevant suggestions
The number one cause of low cross-sell conversion. Shopper sees a "people also bought" carousel and the items have nothing to do with what they are buying. Trust drops. The carousel becomes noise.
Fix: curate the top 50 hero SKUs manually. Let the algorithm handle the long tail. Test the carousel monthly to catch drift.
Slow load on PDP
Cross-sell widgets can add 200-800ms to PDP load if they fetch recommendations client-side without caching. Slow PDP kills conversion.
Fix: server-side render the initial recommendations. Hydrate updates client-side. Cache aggressively.
Too many slots
Three cross-sell widgets on the PDP. Five-item carousel in the cart drawer. Six suggestions on the post-purchase page. The shopper is staring at 14 cross-sells before reaching the thank-you.
Fix: one slot per surface. One carousel max. Three items max per carousel.
Not measuring incremental lift
Cross-sell tools report "items added via cross-sell." That number is not the same as incremental AOV. Some of those items would have been added anyway.
Fix: A/B test cross-sell on/off for a meaningful sample. Measure AOV difference, not "items via cross-sell."
Where to start
One: turn on native Shopify product recommendations on PDP if you have not. Free baseline.
Two: pick one of the four surfaces (PDP, cart, post-purchase, post-purchase email) that is currently empty. Add cross-sell there only.
Three: hand-curate the cross-sell suggestions for your top 20 SKUs by revenue. Manual beats algorithmic on hero products almost every time.
Four: run the test for 60 days. Measure margin-adjusted AOV. Compare to the prior 60 days. Decide whether to expand based on lift, not on "items added."
If you want help mapping cross-sell into your existing cart and post-purchase setup, book a checkout and AOV review. We will tell you which surface is leaking and what to fill first.
What is the difference between cross-sell and upsell?
Upsell replaces what the shopper is buying with a higher-value version. Cross-sell adds a complementary item. Both lift AOV but the offer language and conversion math are different.
How much can cross-sell lift AOV?
Typically 4-9% on margin-adjusted AOV across the funnel. Less than upsell uplift (which can hit 12-22% on its own) but compounds with every visit.
Which cross-sell engine is best for a $10M Shopify Plus brand?
Searchspring or Klevu if search-and-recommendation as one stack. Algolia Recommend if you are already on Algolia for search. Native Shopify if you have not exhausted the free option yet.
Does cross-sell hurt conversion?
Aggressive cross-sell (multiple widgets, many items) does. Restrained cross-sell (one slot per surface, three items max) does not measurably hurt conversion in our testing.
Should I use rule-based or AI cross-sell?
Rule-based for small catalogues under 200 SKUs and high-curation brands. AI/algorithmic for catalogues over 500 SKUs or fast-changing inventory. Most brands at scale use both.
How long does it take to set up cross-sell?
Native Shopify recommendations: 1 day. Dedicated cross-sell app: 1-2 weeks. Full engine implementation (Klevu / Algolia Recommend): 3-6 weeks including data tuning.
Where does cross-sell fit alongside upsell?
Run both, on different surfaces. Common pattern: variant upsell on PDP, cross-sell in cart drawer, complementary cross-sell on post-purchase page, both via email follow-up. Do not stack them on the same surface.
How do I measure if cross-sell is working?
A/B test cross-sell on/off for at least 30 days. Measure margin-adjusted AOV difference. "Items added via cross-sell" is a vanity metric. Incremental AOV is the real number.
A Shopify Plus Agency for Strategic Design & Advanced Engineering
Building something ambitious?
- Cross-sell offers a complementary item alongside what the shopper is buying. Different mechanic from upsell, which offers a higher version of the same thing.
- Four surfaces: PDP, cart, post-purchase, post-purchase email. The math is different on each.
- Three approaches: rule-based ("if X then suggest Y"), AI/algorithmic (Searchspring, Klevu, Algolia Recommend), manual curation. Manual still wins for catalogues under 500 SKUs.
- Native Shopify recommendations get you a free baseline. Searchspring, Klevu, and Algolia Recommend are the three engines worth comparing at $5M+ GMV.
- The biggest pitfall is irrelevant suggestions. "People also bought" carousels showing random SKUs hurt trust more than they help conversion.
- Cross-sell uplift is usually 4-9% on AOV. Less than upsell uplift but compounds across every visit.
