Engage's product recommendations engine looks at all purchased and viewed products for a customer and from that calculates the products that might most interest them. These products can be fetched as a list of SKUs through the Engage API using the contact ID.
The more purchase and returns data available, the better the recommendations. Data such as gender or age are not considered when calculating product recommendations, although that can be inferred since certain demographics can be linked to certain purchase patterns.
Click-data behavior can also be used when available.
You can also choose to not show recommendations for customers with too little purchase behavior, since the results would not be accurate.
Learn all about the product recommendations endpoint here.
Configuring product recommendations
When you get started with Engage, you will receive questions about how you want to product recommendations to work. Below these questions are explained.
Q: Are there products that you never want to recommend?
It’s possible to exclude products that you never want to recommend such as gift cards, tobacco, gambling, etc. By default, the promotion engine will recommend all products that are in the product feed that's configured for product recommendations, so if you have products that should never be recommended, they should not be part of the feed. In this case, you'll need to have a separate product feed containing only products that you want to recommend.
Q: Are there products that you wish to promote or demote in your product line?
If you promote a product, there is a higher probability it will be recommended to your customers. This could be useful if you have your own brands that you want to promote, or products with higher margin. If you want to promote or demote, we need you to provide Voyado with a list of these products or brands on a scale of -100 (demoted) to +100 (promoted).
Q: Do you want to spread fallback products across different categories for maximum spread?
This decides if the customer should be recommended a variety of top selling products or if it’s okay that they are only recommended top-sellers from a single category. If you choose to spread the fallback products, the engine will pick top sellers from different categories, meaning that the customer can be recommended the top selling shirt, the top selling sock and the top selling pants.
If you choose not to spread the fallback products into different categories, the customer can, for example, be recommended 3 different shirts (if shirts are your top selling category).
In addition to this, you can decide how far back you want the fallback top-selling products to be calculated. We recommend 30 days but it depends on your customers purchasing cycle.
q: Do you want to alternate recommendations between send-outs?
If you alternate the recommendations, you will expose each contact to a broader range of products and not just the same top recommendations for that contact every week.
It's possible to control what is recommended to your customers. Here are two example use cases.
Company A
Customer A sells consumables like health and hygiene products.
"Our customers buy the same product a few times per year. We want to recommend the same product that the customer has bought several times, but we want to show the customer that several brands make the same kind of product by recommending this product from other brands."
By enforcing the rule "Brand and product name cannot be the same within the recommendation" we limit the output to not only show highly likely repurchases but to include some product variance.
Company B
Background: Sells a wide range of outdoor products such as tents, boots, clothing, canoes, etc.
"Our customers do not shop often but when they do, it’s an investment for them. We want to show the customer a wide range of products and make sure that we don’t recommend products from the same subcategory in one email."
In this case we should enforce a rule that ensures that subCategory differs for all products within the recommendation.
Article last reviewed
Comments
0 comments
Please sign in to leave a comment.