What should define what appears on top of each search result page?
The simple answer is “my top sellers which match the search term”, which can be done easily with Boxalino Winning Interactions platform with Best Practice Strategies like STR - Search Attribute and PCS - Smart Bestsellers.
However, these approach will not capture the way that your customers click on each specific search results page. Some of your best-sellers might have the term in their description but not be the best match for some terms. The best way to learn it is by analyzing the Click-Through and Buy-Through rates of each product displayed in the search and for each search term.
With its uniquely advanced tracking capacities, the Boxalino Winning Interactions platform is able to capture very precise information not only about the clicks, but also about the display of each product on the screen of the client (see How Boxalino tracks structured data automatically for details)
With its highly scalable Data Warehouse based on Google BigQuery (called CODW as described here), the Boxalino Winning Interactions platform is able to compute billions and billions of product displays, clicks and buy-throughs and to apply various types of machine learning algorithms to optimize the search ranking.
What is it?
The Search-Based Display $Value Self-Learning optimizes the ranking of the search results based on the self-learning of the following KPIs (computed for each search term):
Engagement (e.g. Clicks and Click-Through Rate or CTR)
Click Orders and Click Order Rate (orders following a click on the search results)
Click Revenue (total revenue generated by the product clicked and bought from the search results)
Click Margin (same, but with the product gross margin)
Display $Value (Click Revenue per Display on the screen of the client)
Display $Margin (same, but with the product gross margin)
Any other custom attributed KPIs configured in your environment
This best practice is about improving the search ranking from the self-learning of what products are more (or more frequently) clicked, bought or added to the basket than others (as well as the revenue and margin attributed to these clicks).
Best Practice Strategy
Promote products based on the self-learning of each Search Query click/display score, either directly on the number of clicks or on the conversion, average bought value, or margin generated by the clicks
Display = visual impression detected on the screen of the user before, during, and after scrolling on the page
You will need to decide if you want to use the CTR or the BTR self-learning
* the Use Cases configuration is provided as SCORER with a default weight of 2000. You will need to adjust the weight to have the desired strength in the results.
Variables
type: the name of the BigQuery generated correlation type
Cases to consider
Please ask Boxalino to receive the type, but they will look like this (it’s a string you can copy):
Per Search-Term Click-Through-Rate (CTR) self-learning boxalino_std_segmentation_widget_process_v1_ctr_products_group_id_mid_term
Per Search-Term Buy-Through-Rate (BTR) self-learning boxalino_std_segmentation_widget_process_v2_btr_products_group_id_mid_term
* products_group_id: it means that you want the statistics per Product Group. If you want, you can also use another grouping field (check the list of available fields in the Field Explorer)
** mid_term: we typically suggest considering the MidTerm period (6 months) as otherwise, the statistics do not improve fast enough
How to configure it?
You can import the JSON below directly in the Admin (use the Import button on the top right), as in this screen-shot, you will need to enter the type as indicated above.