Architecture

Open

Configuration Steps

1. Deploy to a partition the Product_Recommendation and Product_Recommendation_Eval logics. The logics can be renamed but keep the format <new logic name> and <new logic name>_Eval.

2. Set up an Optimization model with the Product Recommendation logic and save it. Give the model a name, e.g., CrossSell_1.

   Open

   

   • Minimal # of Transactions and Minimal Count_Product_Baskets set the statistical thresholds for creating recommendations. Customers with lower number of total transactions and products occurring in the lower number of baskets will not be given a recommendation.

   • # calculated co-products sets the number of recommendations given for each customer and each product.

   • Customer and Product Name can be on different level of hierarchy, then recommendations will be calculated on this level.

   • Product Id needed to find the product in Product Master.

   • Basket ID is used to determine which products are bought “together”. If not available then Invoice Date is a good proxy.

3. Go to Calculation/Step1 and click “Run”. The model will fail for the first time.

4. In Analytics > Data Manager > Data Loads find your model of type Calculation and uncheck the Allow distributed calculation checkbox. Then click the Save button.

5. Run the model. It will fail again.

6. Go to logs and find the hash of the executed queries. If hash is not available, contact support and ask to set “log category DMSqlQuery to DEBUG”. Copy the hash values for all failed queries and in the Product_Recommendation logic in element Query data set the hashes accordingly. Note that if a partition resides on two nodes, the logs with Hash can be redirected to the other node.

7. Run the model. If successful, the Company Parameter tables are produced with the name format RecomendationsPerCustomer<$modelName> and RecomendationsPerProduct<$modelName>. Observe the precalculated recommendations in the tables.

8. In Administration > Configuration > Advanced Configuration Options create "quoteProductRecommendationsConfig" entry with this value:

   { "recommendationsEnabled": true, "modelName": "CrossSell_1", "allowDuplicates": false, "categories": [ { "categoryName": "customer", "defaultMaxResults": 5, "enabled": true }, { "categoryName": "customerSegment", "defaultMaxResults": 5, "enabled": true }, { "categoryName": "productHistory", "defaultMaxResults": 5, "enabled": true }, { "categoryName": "productRelations", "defaultMaxResults": 5, "enabled": true } ] }

   
   Note that modelName should point to the PO model with Product_Recommendation logic, e.g. CrossSell_1. To use a different Product Recommendation model put here a different model name.

9. Quotes with product recommendations are available in Unity view of the type: https://qa.pricefx.eu/unity/develop/#/quotes, note the middle part of the address. Create a new quote, select a customer or/and select a product under Items tab. Click Add Recommended Products and discover the products recommended for the current quote.

    

Calculation Methodology

Customers Buying History

Products Bought Together