The Tableau CRM Data Prep team has been vigorously working on a new Pilot feature for generating the best product recommendations for your customers in an easy-to-use declarative manner. This feature is, appropriately named the Product Recommender (Pilot) node; which will be surfaced alongside the other Machine Learning transformations in the Data Platform’s Recipe builder. The transformation node will take in your prior customer purchase history and deliver a dataset back with recommendations of the next products to sell or resell to those customers.
In this article, we will dive into the Product Recommender (Pilot) transformation. We’ll talk about what it is, how it works, how we built it, how you can best use it, and what we’re doing next.
Product Recommender (Pilot) Transformation at a Glance
It is a very simple to use feature that takes in three columns; Customer, Product, and Rating. The rating can be an explicit value like customer satisfaction or it can be implicit like “did they reorder/subscribe”, basically you are trying to add a value for the algorithm to enhance the recommendation for preference. If you don’t have a rating then simply the quantity purchased or amount will suffice to generate an indicator of did they buy it. The node will then generate a dataset that has 10 (current pilot limit) recommendations per customer for the products they are most likely to want to purchase next.
Wait, How Does the Product Recommender Actually Work?
Intuitively we understand that the purchases you make and like define the type of “purchaser” you are, and that people that resemble your purchasing behavior and preference are likely to have purchased and liked products you haven’t. Simple enough, but there is more power in the approach that we have taken – Alternating Least Squares Collaborative Filtering – which is a very standard recommender algorithm.
Note: For more details on what and how the Product Recommender (Pilot) works check out this short video.
Ok, so the power comes in not only finding what your customers purchased, and what customers they are similar to based on products they purchased and liked, but also in what they didn’t purchase (yet) or what they purchased and didn’t like. Our implementation of Collaborative filtering boils this down to matrices for customers and products, these metrics are then combined and the result generates a scored recommendation for customers for all the products; yeah that’s a lot of data. By looking at the top N recommendations per customer we pare this down to a consumable and valuable dataset that can be leveraged operationally (next best product) or analytically (campaigns, or whitespace).
So what is the use case for this new Product Recommender transformation? First would be to offer more relevant services or products to your customers. As your Account Executive is getting ready to visit their customer, they want to know what offerings they should bring up in the conversation, not any offerings but those that are most likely to generate won opportunities.
Another usage is that you want to understand if you have products being under-represented in specific areas of your business. Some customers might be recommended something they already purchase, but if you add in the purchased quantity of their peers and see a discrepancy there might be an opportunity for recommending increased purchase quantity.
These are just two use cases that quickly can aid your Account Executives in navigating their customer meetings.
Roadmap for Product Recommender
At this point we are in a pilot so the audience is small; we have been working with key customers to make sure the feature is clean, easy to use, and valuable. As we process the feedback we will make the needed adjustments. We do have known gaps that we are working on to make sure you will be absolutely confident in deploying this feature.
Want to get involved in the Pilot?
Please reach out to your Salesforce or Tableau CRM Account Executive if you would like to be included in the pilot and try it out. Or reach out to me if you have speculative questions or feedback for the feature.
This content contains forward-looking statements that involve risks, uncertainties, and assumptions. If any such uncertainties materialize or if any of the assumptions proved incorrect, the results of salesforce.com, inc. could differ materially from the results expressed or implied by the forward-looking statements we make.
Any unreleased services or features referenced in this document or other presentations, press releases or public statements are not currently available and may not be delivered on time or at all. Customers who purchase our services should make the purchase decisions based upon features that are currently available. Salesforce.com, inc. assumes no obligation and does not intend to update these forward-looking statements.