Introduction to Mail Recommendations

What are Mail Recommendation strategies?

Raptor offers several recommendation modules for email, each built around a different way of selecting and personalizing which products to show. Picking the right one for each email, whether that's a newsletter to your full list, a basket recovery email, or a re-engagement send, determines how relevant your recommendations feel to the subscriber. This article walks through the five most commonly used strategies, when to use each, and how they compare. For the full technical reference covering every mail module on the platform, including ones not covered here, see Mail modules: Recommendations.

Prerequisites

A live product feed connected to the Recommendation Engine, delivering at minimum a product ID, price, image and category for each product. See Setting up a Product Catalog for Recommendation Engine.
Tracking implemented and validated for visit, basket and purchase events, and search events if you plan to use search history as a signal. See Tracking Events & Parameters reference .
Your Email Marketing Integration and REAID synchronization set up, so each subscriber can be recognized. See Getting started with your E-mail Personalization.

🔍 Note: Build your product catalog and tags around the use cases you want to cover first, and validate the data feed, stock handling and layout on real test profiles, including profiles with no history, before going live.

How a mail recommendation module is built

Every mail module is built from four blocks, configured in the same order:

Candidate set strategy: how the pool of possible products gets filled, from category or global popularity, trending, or the subscriber's own journey.

Rerank weights: the settings that decide what order candidates come out in, based on cross-sell, look-alike, buy history, basket, visit and search behaviour. See Module adjustment and tuning for the full list of weights and what each one does.

Backfilling: the fallback that fills the list when a subscriber's personalized data is empty or too thin, usually with popular products.

Output & selection: formatting only, layout, product fields returned, and any filters. Doesn't affect which products get chosen.

🔍 Note: Candidate set strategy and backfilling decide which products can appear. Rerank weights decide what order they come out in. Output & selection only affects how they're rendered, never which products are chosen.

💡 For precise technical definitions of these terms, see the Glossary at the end of Mail modules: Recommendation Strategies.

Choosing the right strategy for your email

Start from what you know about the subscriber, and what the email needs to do.

No usable behavioural signal (cold segment, non-opener): use Trending Products, or Trending Products by Category if you know the category.
Want to remind subscribers of products they already looked at: use Recently Viewed Products.
Want to grow the basket with complementary items: use Cross-sell.
Want to personalize to the subscriber's taste: use Personal Recommendations.
Want to spotlight a category or a season: use Trending Products by Category.

Comparison at a glance

Criterion	Trending by Category	Trending	Personal	Recently viewed	Cross-sell
Need for individual data	Low	None	High	Medium	High
Resistance to cold start	Good	Excellent	Low	Low	Low
Primary objective	Category context	Mass send / fallback	1:1 personal	Reminder / re-engagement	Basket / average order value
Sensitivity to product catalog	Medium	High (long tail)	Medium	Low	Medium
Priority KPI	Category click-through	Click-through / coverage	Click-through + conversion	Return rate	Average order value / incremental revenue

1. Trending Products by Category (`GetPopularItemsInCategoryMail`)

Principle: surfaces the most popular or trending products within a category relevant to the subscriber, either their affinity category or a category you set for the campaign. Answers "what's working in the category they care about?"

How it works: the candidate pool is built from the targeted category's products, ranked by popularity (aggregated visits and purchases) or by trend. Personalization comes from the choice of category, not a 1:1 subscriber profile. Falls back to overall category popularity if there's nothing better to show.

Strengths: works even without much purchase history. Reads as more relevant than site-wide popularity, since it's contextualized to a category. Suited to non-openers and new subscribers.

Limitations: does not personalize within the category itself. Depends on how well your catalog's category tree is structured. Can repeat the same best-sellers if popularity is weighted too heavily.

👀 Use case: thematic newsletters, a seasonal spotlight on one category, cold-start subscribers, or catalogs with a clean category tree.

🔍 Note: not suited to a poorly categorized catalog, or when you need personalization within a single category, use Personal Recommendations instead.

2. Trending Products (`GetPopularItemsMail`)

Principle: the most popular or trending products across the whole site, all categories combined, the "wisdom of the crowd" with no individual data involved. Also serves as the fallback for every personalized module.

How it works: global ranking by popularity (visits and purchases over a 7, 30 or 360-day window) or by trend. No profile required, the same selection shows for everyone, filled on open.

Strengths: does not depend on subscriber data, so there's always something to show. Suited to cold start and mass sends. Works as a fallback for other modules.

Limitations: shows the same products to every subscriber, with no personalization. Can crush the long tail if popularity is weighted too heavily. Can feel dated if the trending period is too long.

👀 Use case: non-openers, new subscribers, "best-seller" blocks, or a safety net at the bottom of the email.

🔍 Note: on an engaged subscriber base where personalization is possible, avoid using this as the main block, it doesn't use the data you already have.

3. Personal Recommendations (`GetUserItemRecommendationsMail`)

Principle: recommends products to each subscriber based on their own behaviour (visits, basket, purchases). This is the mail equivalent of the Personal Shopping Assistant module used on your website, see Personal Shopping Assistant, and forms the core of a personalized CRM programme.

How it works: built in two stages, a candidate set that adapts to the subscriber's journey, then reranked by weights based on their profile (cross-sell and look-alike combined). Rendered on open; falls back to popularity if the profile is empty.

Strengths: provides the highest level of 1:1 relevance among the five strategies. Uses all available behavioural data. Increases click-through rate and conversion.

Limitations: needs enough history to be reliable, typically after about a week of tracking. Reasons at parent-product level, so colour or size variants aren't distinguished. More complex to tune than the other strategies.

👀 Use case: an engaged subscriber base, signal-rich e-commerce, main newsletters, or the core of a CRM programme, especially for profiles with more than 3 purchases.

🔍 Note: not suited to a large cold-start audience, single-purchase subscribers with little signal, or a very small catalog.

4. Recently Viewed Products (`GetUserItemHistoryMail`)

Principle: reminds the subscriber of products they themselves recently viewed, so their browsing continues in the email. Effective for re-engagement.

How it works: replays the subscriber's own visit and basket history, sorted by recency. No discovery algorithm involved. Key filters exclude anything already bought and keep only what's still available.

Strengths: shows products the subscriber is already interested in. Simple and easy to read. Effective for re-engagement and browse abandonment.

Limitations: provides no discovery or novelty. Empty if the subscriber has no history. Can resurface bought or out-of-stock items if filters aren't set correctly.

👀 Use case: re-engagement sends, browse or abandoned basket, "you viewed..." blocks, and other behavioural triggers.

🔍 Note: not suited to acquisition or cold-start sends, or when the goal is to broaden a subscriber's taste rather than remind them of what they already saw.

5. Cross-sell (`GetUserCrossSellingItemsMail`)

Principle: suggests products that complement what the subscriber bought or added to their basket, "customers who bought X also took Y", personalized to their own basket or purchase. A direct lever on average order value.

How it works: applies item-to-item related products to the subscriber's purchased or basket items. Uses a negative basket weight so the module doesn't re-suggest what's already in the basket. Falls back to popularity when there isn't a strong complementary match.

Strengths: increases average order value and incremental revenue. Works well post-purchase and for basket recovery. Monetizes existing purchase behaviour directly.

Limitations: needs a base of purchases or basket activity to work from. Weak without clear complementarity between products. Requires clean, well-structured categories.

👀 Use case: post-purchase emails, basket recovery, upsell and average-order-value campaigns, and bundles.

🔍 Note: not suited to cold start, a catalog without complementarity logic, or single-purchase subscribers with nothing in their basket.

Notes and limitations

🔍 Note: these five strategies are the ones most commonly used across mail campaigns. See Mail modules: Recommendation Strategies for the full reference of every mail module available, including brand-, content- and Merchandising-based modules.

⚠️ Warning: changing a module's rerank weights changes the order products come out in for every email using that module. Test changes with real subscriber profiles, including profiles with no history, before rolling out.

🔍 Note: see Recommendation module settings by industry for starting-point weight values by business type.

FAQ

Q: Can I use more than one strategy in the same email?
A: Yes. Many emails combine a personalized block (such as Personal Recommendations or Cross-sell) with a Trending Products block as a safety net for subscribers without enough history.

Q: What happens if a subscriber has no browsing or purchase history?
A: The module falls back to backfilling, usually overall or category popularity, so the email doesn't render with an empty recommendation block.

Q: Which strategy should I start with if I'm not sure?
A: Personal Recommendations for your main, engaged newsletter list, and Trending Products as the fallback for non-openers and new subscribers. See Recommendation module settings by industry for guidance by business type.