You Are Leaving Money on Every Order. AI Can Fix That.
The average e-commerce order value sits around $80. Most small businesses do nothing after the customer clicks “add to cart.” No suggestions. No bundles. No “you might also want” nudge. Just a checkout button and a hope that the customer does not notice the shipping cost.
Cross-selling and upselling are not new ideas. The problem has always been execution. Manual recommendations are slow. Static “related products” sections are generic. And the person running the store is already busy with orders, support, and inventory.
AI changes the math. It can look at what is in the cart, what the customer has bought before, and what similar customers tend to buy together, then serve a recommendation in real time. Not a random grid of products. A specific suggestion that makes the customer think, “Yes, I actually do want that.”
How AI Cross-Selling Actually Works
There are three types of recommendations that increase average order value:
- Complementary cross-sells: Products that go with what is already in the cart. A candle buyer gets offered a wick trimmer. A skincare buyer gets offered a cotton pad. These are not random. They are the items your best customers already buy together.
- Tiered upsells: A better version of the same product. The customer is buying the 100g jar. AI suggests the 250g jar at a lower price per gram. The customer is choosing the basic plan. AI shows the premium plan with the features they actually need.
- Bundle offers: A group of products sold together at a discount. Not a random “3 for 2” on everything. A specific bundle built from purchase data that matches what people actually want together.
The key difference between AI recommendations and a static “you may also like” grid is timing. AI shows the right suggestion at the right moment: on the product page before they add to cart, on the cart page before they check out, or in the confirmation email after they buy.
Setting Up AI Cross-Selling Without Writing Code
Most e-commerce platforms now have AI recommendation engines built in or available as plugins. Here is how to set them up on the three most common platforms:
Shopify: Use Shopify Search and Discovery (free) to create merchandising rules, or use a plugin like Wiser, Rebuy, or LimeSpot. Each one connects to your store data and starts generating recommendations within 24 hours. The setup wizard asks what kind of recommendations you want (frequently bought together, related items, recently viewed) and where you want them to appear.
WooCommerce: Install Cartflows or YITH WooCommerce Recommendations. Both pull from your order history and product tags to generate suggestions. YITH lets you set rules like “show products from the same category” or “show products customers also bought.”
Squarespace: Squarespace has built-in related products, but they are category-based and not very smart. For AI-driven recommendations, use a third-party tool like Nosto or Clerk.io that connects via a code snippet in your header.
Regardless of platform, the setup process follows the same pattern:
- Install the recommendation engine or plugin.
- Let it run for 48 hours so it can learn from your store data.
- Choose where recommendations appear: product page, cart page, post-purchase email.
- Start with “frequently bought together” as your first recommendation type. It is the most accurate and has the highest conversion rate.
The Three Numbers That Tell You If It Is Working
After you turn on AI cross-selling, watch these three metrics:
1. Average order value (AOV): This should go up within the first week. If your AOV was $78 and it moves to $84, that is a 7.7 percent increase. For a store doing 100 orders a month, that is an extra $600 in revenue with zero additional marketing spend.
2. Items per order: If customers go from buying 1.3 items per order to 1.7, the cross-sell is working. This is a clearer signal than revenue because it tells you the recommendation itself is convincing, not just the price.
3. Recommendation click-through rate: Most tools show you how many people saw a recommendation and how many clicked it. If your click-through rate is below 2 percent, the recommendations are not relevant enough. Above 5 percent means they are hitting the mark.
Check these numbers after 7 days. If AOV has not moved, change your recommendation type from “related items” to “frequently bought together” and try again.
Common Mistakes That Kill Conversion
Too many recommendations at once. Showing 8 products in a “you may also like” section is noise. Show 2 to 3 targeted suggestions instead. One complementary product and one upgrade option. That is enough.
Recommending products the customer already owns. If your AI engine does not check purchase history, it will suggest things the customer just bought. This feels lazy. Make sure your tool connects to customer order data, not just product tags.
Ignoring the cart page. Product page recommendations are good for discovery. Cart page recommendations are good for urgency. The customer has already committed to buying something. A relevant add-on at this stage has the highest acceptance rate.
Using the same recommendation everywhere. “Related products” on the product page should be different from “frequently bought together” on the cart page. The context is different. The customer is in a different mindset. Your recommendations should reflect that.
Start With Cart Page Recommendations
If you want one place to start, make it the cart page. Add a single row of 2 to 3 complementary products below the cart summary. Use “frequently bought together” as your recommendation type. Give it 7 days.
You do not need a complex AI setup to start. You need a single recommendation row in a single place, measured by one metric: did average order value go up?
If it did, expand to the product page and post-purchase email. If it did not, change the recommendation type and try again.
Want more like this? Join MOKU Club for free. Weekly resources, early access to new guides, and occasional templates you can actually use.
