Product feeds are the basis for recommendation campaigns, social proof tests, defining user affinity profiles, creating audiences and so much more. Therefore, setting up you feed correctly is critical to driving uplift.
Read on to discover:
- What is a feed in Dynamic Yield
- Which type of fields1 are required and why
- Why should you enrich your feed with additional attributes
What is a Data feed in Dynamic Yield
Data Feeds are files used to ingest data into the Dynamic Yield platform. They contain the product catalog, list of variations to be used in campaigns or content available on your site.
Today we will focus on the Product feed.
What type of data should be in your Product feed
Think about what the data is going to be used for and based on that, add relevant fields (attributes).
Affinity is calculated on the attribute level. Every time a user engages with any product on your site or app, we are enhancing their affinity profile based on the attribute values engaged with.
What are the attributes that will help you create more meaningful Affinity profiles? Below are some recommended examples from different industries.
- Jewelry: Type of stone, Material, Size, Type (necklace, bracelet, earring etc…)
- Fashion: Gender, Color, Brand, Material, Style
- Travel: Destination, Departure month, Origin, Number of travelers, Type of Board
- Electronics: Brand, Season, Size, Energy level, Weight
- Beauty: Skin tone, Brand, Type, Color
- Furniture: Color, Brand, Room, Price range, Size measurements
PDP of a Jewelry site, depicting the relevant attributes to add to the feed
Dynamic Yield recommends specific products based on the algorithm defined, and availability in your feed. As such, the feed must include certain information like the SKU, stock level (in_stock), name, price, url, image url and categories.
How should the feed be set up in order to make sure that the most relevant items are recommended? We will focus on 3 fields: GroupID, Categories and Keywords2
The GroupID signals to Dynamic Yield that certain unique SKU’s are part of the same group, and therefore should not be recommended side by side. We always recommend one representative of the group.
For example, the same shirt in different sizes and colors. Each of the unique types will have their own SKU, however they will be under the same GroupID; Red Size S, Red Size M, Red Size L, etc…
Items with the same GroupID have different size and color combinations
This field is especially pertinent when using the Similarity algorithm. The similarity of items is based on their keywords and categories. Categories should mirror the breadcrumbs of a site, to give the full path that this product is defined by.
Feed categories #1: Women > Women clothing > Womens shirts > Womens short sleeve shirts > Womens polo shirts
Feed categories #2: Women > Womens shirts
Popularity algorithm (strategy) with the option of filtering items based on the category/ parent category. Parent category is the second from the right, and category is the full string. In the first example, Womens short sleeve shirts is the parent category.
A recommendation widget using the Similarity algorithm based on feed categories #1 would return the following results:
Whereas using the second feed, the results would be:
In the former, the results are specific to style whereas in the latter, general shirt category was used.
Much the same as the categories, the Similarity algorithm relies on the keywords to return relevant product recommendations. The more keywords, the more specific the algorithm can get.
When collecting view and purchase data on an item, we are generally going to want to amalgamate data from all items within the same group. For example, the views and purchases of the shirt we mentioned earlier, available in multiple sizes and colors, should all be collected together as one “grouped” item. The GroupID will manage this process.
Once determining the relevant attributes to add to your feed, and adding them, the next step is to use this information in your campaigns; create audiences, target experiences, define campaigns, etc... For example;
Creating Audience based on Page type and Product properties:
Targeting by page and product type:
1 Fields in the feed can be synonymous with attributes. Different fields can be defined as attributes when referring to Affinity Profiling. The values within the field are the attribute values (for example; attribute = Color, attribute values = red, green, blue, etc...)
2 Keywords is not a mandatory field in the feed, however, the GroupID and Categories are
- Click here to read more about feeds
Please sign in to leave a comment.