Algorithm Studio is an Experience OS app that empowers both marketers and data scientists to create custom algorithms that address unique business needs. A user-friendly interface enables you to quickly generate algorithms to use in your recommendation campaigns.
Step 1: Define the factors to consider in your algorithm and their relative impact
- Go to Algorithm Studio › Create Algorithm.
- Name your algorithm. Keep in mind that the algorithm will be used in Experience Email or recommendation strategies, so keep the name short and descriptive. For example, “Trending Products”, “Trending by Views”, or “Bestselling Offline”.
- Select up to 4 ranking factors.
Note: Only product feed columns reported with type "number" are available in the Product Attributes section. A product is scored and ranked as part of an algorithm with a product attribute factor only if it has a valid number value in all its number-type fields.
- Define each factor's relative weight, and set the factoring direction (high to low to prefer products with high factor values and vice versa. For example, in the case of Return Rate, the lower the value, the higher the product should be ranked). These define how your products are ranked in the algorithm.
Here's how these definitions affect your algorithm: Suppose you select only Purchases. This creates a Best Seller algorithm. If you select Purchases, Reviews Score, Return Rate, and Sale Price – this creates a Top Picks algorithm that considers multiple product performance and attribute factors.
Step 2: Define the product performance lookback window
This determines how far back the algorithm counts product interactions as relevant for the purpose of computing your algorithm. For example, to create a “3 Days Trending” algorithm, use a 3-day data window.
Step 3: Select an audience to factor by
Learn what your audiences like and apply it to your recommendations: Select an audience (with a minimum of 1000 users) whose behavior you want to consider for your algorithm.
Note: Works only for non-session audiences
Step 4: Preview the results
Before saving your algorithm, preview the results to see how the algorithm ranks your products. Depending on the size of your product feed and the volume of your data, the preview calculation might take up to a couple of minutes.
Step 4: Save your algorithm
When you save the algorithm, its status is Pending. Depending on the size of your catalog, the initial processing of the algorithm can take up to several hours. The algorithm is recalculated every time your Product feed syncs, or for API Product feeds – every 12 hours.
Step 5: Use the algorithm in your strategies
After the algorithm is processed for the first time, it appears in the algorithms drop-down in the Create Strategy screen, where you can see all the algorithms you've created in the Algorithm Studio as well as the built-in Dynamic Yield algorithms. Select an algorithm and apply merchandising rules and filters as needed, depending on your use case.
Limitations:
- You can create up to ten algorithms, but only five of them can have an audience-based product performance setup.
- The Product Performance by Audience section includes only audiences with at least 1000 users.