Product recommendation emails can drive serious revenue for online stores. When done well, they feel helpful. When done poorly, they feel like spam.

The difference comes down to relevance. Recommending a winter coat to someone who just bought three winter coats is annoying. Recommending a matching scarf? That's useful.

Here's how to send product recommendations that people actually appreciate and click on.

Types of Product Recommendation Emails

Not all recommendation emails are the same. Each type works best in a different context.

Cross-Sell Recommendations

"You bought X, you might also like Y."

This is the most common type and the one most stores think of first. The key is making the recommendation genuinely complementary.

Good cross-sells:

Bought a camera → Camera bag, extra lens, memory card
Bought running shoes → Running socks, insoles, shoe cleaner
Bought a coffee maker → Coffee beans, filters, descaling solution

Bad cross-sells:

Bought a camera → Another camera
Bought running shoes → Dress shoes (unless they've also browsed dress shoes)
Bought a coffee maker → A different coffee maker

The best cross-sells solve a problem the customer didn't know they had, or make the product they already bought even better.

Upsell Recommendations

"You're looking at X, but Y might be an even better fit."

Upsells work best before the purchase, not after. Once someone has bought, telling them they should have bought the more expensive version feels bad.

Where upsells work in email:

Browse abandonment: "You were looking at our Basic plan. Here's what you get with Pro."
Pre-launch: "The Premium version of [Product] is now available."
Subscription upgrades: "You've been on the starter size for 3 months. Want to try the family size?"

"People Also Bought" Recommendations

Social proof combined with product discovery. These emails show what other customers with similar buying patterns purchased.

This approach works especially well when you have a large catalog. It helps customers discover products they wouldn't have found through browsing.

Personalized Picks

Curated recommendations based on the customer's browsing and purchase history. "Hand-picked for you" style emails that feel personal.

These require good data and either manual curation (for smaller stores) or algorithmic recommendations (for larger ones).

Category and Collection Emails

Instead of recommending specific products, recommend categories or collections. "New arrivals in your favorite category" or "Our spring collection is here."

This works well when you're not sure exactly which product to recommend but know the customer's preferred category.

Timing and Triggers

When you send a recommendation email matters as much as what you recommend.

Post-purchase (21-30 days): The sweet spot for cross-sells. They've received their order, used the product, and are still thinking about your brand. Read our post-purchase guide for the full sequence.

Replenishment window: For consumable products, send recommendations alongside a reorder reminder. "Time to restock? Here are some other products our customers love."

Browse abandonment: If someone viewed products but didn't buy, follow up with those products plus related recommendations. This requires tracking browse behavior on your site.

Seasonal transitions: When seasons change, recommend relevant products from your catalog. "Getting ready for summer? Here are our picks based on what you've bought before."

After a review: If someone leaves a positive review, they're in a good mood about your brand. Follow up with a recommendation email while the positive sentiment is fresh.

Writing Recommendation Emails That Convert

The copy around your recommendations matters more than people think.

Lead with value, not selling. "We thought you'd like these" feels different from "BUY THESE NOW." Frame recommendations as helpful suggestions.

Explain why you're recommending it. "Because you bought [Product]" or "Based on your recent purchase" gives context and makes the recommendation feel less random.

Include social proof on each recommended product. Star ratings, review counts, "bestseller" badges. Let other customers do the selling for you.

Limit the number of recommendations. 3-4 products per email is ideal. Too many choices leads to choice paralysis. If you show 12 products, people click on none of them.

Make the CTA specific. "Shop running accessories" is better than "Shop now." Specificity creates a clearer expectation of what they'll find when they click.

Use real product images. Lifestyle photos work better than plain product shots on white backgrounds. Show the product in context.

Segmentation for Better Recommendations

Generic recommendations sent to your whole list will always underperform. Segment first, recommend second.

By product category purchased: Someone who buys skincare gets skincare recommendations. Obvious, but many stores still send the same recommendations to everyone.

By price sensitivity: High-AOV customers can see premium recommendations. Price-conscious buyers should see value picks or products on sale.

By purchase frequency: Frequent buyers can receive recommendations more often. Occasional buyers should get them less frequently so each one feels special.

By lifecycle stage: A new customer who just made their first purchase needs different recommendations than a loyal customer who's bought 20 times. New customers should see bestsellers and safe picks. Loyal customers can see new arrivals and niche products.

Sequenzy's smart segments let you build these audience groups based on purchase data from your Shopify store and target recommendations accordingly.

What Not to Do

Don't recommend products they've already bought. This seems obvious but it happens constantly. Check purchase history before sending.

Don't recommend out-of-stock items. Nothing kills a sale faster than clicking through to a product page that says "sold out."

Don't send too many recommendation emails. Once a week maximum. More than that and you become the "buy stuff" email brand. Mix recommendation emails with other content (stories, tips, behind-the-scenes).

Don't use generic "you might like" language for everyone. If you have purchase data, use it. "Based on your recent purchase of [Product]" is 10x more engaging than "Check out these products."

Don't forget mobile. Product images need to look good on phone screens. Buttons need to be tappable. Keep the email scannable.

If You Don't Have AI-Powered Recommendations

Not every email platform has built-in product recommendation engines. If yours doesn't (Sequenzy included, honestly), you can still send effective recommendation emails manually.

Manual approach:

Segment your list by product category purchased
For each segment, curate 3-4 complementary products
Send a targeted recommendation email to each segment

It takes more work than algorithmic recommendations, but it can actually perform better because you're using human judgment about what products go well together. An algorithm might suggest statistically popular items. You can suggest items that tell a cohesive story.

For stores with smaller catalogs (under 100 products), manual curation is often the better approach anyway.

Measuring Recommendation Email Performance

Click-through rate: How many people click on your recommended products? 3-5% is good for recommendation emails.

Conversion rate: Of those who click, how many buy? This tells you if your recommendations are relevant.

Revenue per email: The total revenue generated divided by emails sent. Compare this to your other email types.

Which position converts best: If you show 4 products, track which position (1st, 2nd, 3rd, 4th) gets the most clicks and purchases. The first product almost always wins, so put your best recommendation there.

Getting Started

If you're not sending any recommendation emails:

Identify your top 3 products and think about what complements each one
Segment your recent buyers by what they purchased
Send a simple "based on your recent purchase" email with 3 complementary products
Track clicks and purchases
Iterate based on what performs

You don't need fancy AI for this. Start with human-curated recommendations and see how they perform. You can always add automation later.