Customers who bought this item also bought – Salesforce Analytics
When analyzing consumer data, we instinctively segment customers and accounts using their purchase history. Historic sales data allows us to easily find customers that have purchased Product A and label them as Product A owners. The challenge has been handling instances where customers purchase Products A and B. If the data is structured around the product with each product serving as a separate field, those customers would appear under both Products A and B. Alternatively, if the data is structured by sales record then there would be multiple rows of purchases for the same customer. Neither option yield ideal results. The objective for this post is to demonstrate a solution to this problem and applications of this data using the Salesforce Wave platform.
The most important step in this this analysis is ensuring that the data transformation works properly. Rather than structuring the data around the sales records or products, the data can be structured around the product owners. The sales record data can be added to the product owner data where the owner is the key and the product field is added to the list of owners. It is vital to use a tool that can handle multi-value lookups like Salesforce Wave. The result will be a list of product owners in one field and then all of their purchased products listed in a single record. Keep in mind that owner-level filtering can be done before or after joining the data, while purchase-level filtering should be done prior to the join. For example, to filter on purchases made this year, filter on sales that occurred this year and then perform the join. After completing and validating the transformation results, the data can now be visualized to generate important insights into customer purchasing.
Multi-Product Demo Dashboard
There are a number of approaches to visualizing multi-product purchase data. This post aims at covering two complimentary Salesforce Wave dashboards that can solve multiple use cases. The first of the two dashboards is the Multi-Product Demo Dashboard, which allows users to filter their list of owners by selecting products from a series of owned products lists. The user first selects a product from the leftmost product list and the lists on the right begin to filter and show adjusted counts of owners. Meanwhile, the rest of the dashboard adjusts accordingly. The queries return a count of owners that own all of the selected products, a list of those owners, detailed records of those purchases, and a list of those owners’ open opportunities. Users can open the dashboard and freely explore the options or use it with a list of products in mind. This dashboard can also be customized to add even more products if necessary. After identifying the multi-product owners, users can target those accounts for future opportunities or even view open opportunities that might require follow-up.
2-Product Demo Dashboard
While the Multi-Product Demo Dashboard is set up to support a greater number of products, the 2-Product Demo Dashboard is capable of providing more powerful insights. The dashboard is centered around a cartesian product heat map that shows the concentration of multi-product owners. The dark green colored blocks show the highest cross-product ownership between Products 1 and 2. While this visualization only shows two products, it offers the deeper analysis by showing the relationship between two products relative to the others. Upon selecting the intersection of two products (Product 1, Product 2) the rest of the dashboard filters on the list of owners that own those two products as well as those products themselves.
Similarly to the Multi-Product Demo Dashboard, the 2-Product Demo dashboard shows the count of owners, the list of those owners, the purchase records, and the list of their open opportunities. The additional features on the left side of the dashboard show the count of total owners of each product, the percentage of owners of one product that purchased the other (not the same as the percentage of owners that own both), and finally the percentage of product owners that purchased one product and then the other. While most of this information is great to know, perhaps the most influential pieces are the percentages.
By knowing the percent of ATVs owners that also own Motorcycles, analysts can begin to understand the correlation between the two products. If many customers purchase both products, but 80% of ATV customers also purchased Motorcycles, then maybe ATV customers should be considered Motorcycle leads. Additionally, if users know that 90% of those ATV owners purchased an ATV and then a Motorcycle, that further strengthens the case for marketing Motorcycles to the remaining and future ATV owners. While the correlation in this example may seem obvious, not all scenarios will be. The data modeling approach and visualization demos in this post can help simplify the analysis and uncover leads that might have otherwise gone unnoticed.
 The data used in this analysis is visualized using two custom Salesforce Wave dashboards. Also note that the data shown is fictitious and most numbers being used are completely random.
 80% is not meant to imply a confidence interval or significance.
 90% is not meant to imply a confidence interval or significance.