A key area that SAPinsiders highlight as one of the top business challenges: evolving customer expectations and demand. This has emerged as a top business driver in many of our research reports, including the most recent report, SAPinsider Process Automation and SAP S/4HANA. To meet the rapidly evolving customer expectation and demands, organizations need to understand their customer better, and this is where customer intelligence (CI) comes into play. In simple terms, CI is all about gathering and analyzing customer data to understand your customers better. Considering the increased focus in this area, we have a research report focused on CI scheduled to be published in Dec 2022. This article discusses the advantages of building CI capabilities and explores the key elements required to build this capability. But as with every other capability that incorporates "intelligence", analytics also plays a big role in the customer intelligence area. Many CI tools already leverage advanced ML-based algorithms, like SAP CX suite of solutions. Even if the tool you use does not leverage ML algorithms, you can build customized Machine Learning (ML) algorithms that leverage your customer intelligence data inexpensively using open source tools. In this article, we share some areas where ML algorithms leverage, or can leverage, data from CI tools.

Machine Learning Algorithms Leveraged on Customer Intelligence Data

A significant portion of the "intelligence" aspect of customer intelligence is powered by analytical methods. In this article, we discuss some examples of some of the areas where ML algorithms are being used or can be used, with customer intelligence data. Customer segmentation is a broad term for various analytical approaches that help you categorize or segment your customers in different buckets. It can be as simple as the amount of money they spend with you over a period of time. In reality, however, it is a combination of attributes that defines a customer segment (like salary, age, education, etc.). An example of an analytical method that can be used for segmentation is cluster analysis. As the name suggests, this ML algorithm creates clusters of data points, creating clusters of similar data points. So if the algorithm is fed a dataset with three attributes, say salary, age and education, it will create clusters of customers that are similar to each other in terms of these attributes. Attribute analysis: This type of analytics basically evaluates how a customer values specific attributes of a product. Let us use an example of laptops. When we buy a laptop, we have different preferences. Some would value aesthetics (slim profile), some are very specific about the display (OLED) whereas many like me are fixated with the RAM and CPU. Obviously, manufacturers use this to determine a price point for their products- Will customers pay more for a certain attribute? A method used widely for this type of analytics is conjoint analysis. The term itself means analyzing multiple features jointly. The gist of this method is that you can infer the value of a product feature by experimenting with a combination of features and evaluating data points of consumer ratings for these combinations. While traditionally a statistical method, ML algorithms have significantly enhanced the prowess of conjoint analysis tools. Another ML algorithm that works best on historical data for this type of analysis is linear regression. In simple terms, linear regression models use historical data and then use that to predict which product attributes will impact product sales the most. Demand elasticity: Demand elasticity relates closely with attribute analysis. It determines how "elastic" the demand is to change in certain attributes, like price. Just like it suggests, the term elasticity here pertains to how the demand changes. And the simplest example is that demand for certain products decreases when prices are raised. But as you can imagine from this example, this also depends on product attribute perception. While it does not happen, if Apple increased the price of a new product a couple of weeks after launch, chances are people will still buy it. But if you manufacture table salt and raise your prices while your competitors do not, your demand may go down. This was an extremely simplified example but the idea is that analytics methods in this category help us evaluate the impact of changes in certain attributes on demand. Price and advertising elasticity of demand models are widely used and our good old friend regression analysis comes in handy here as well. Customer Lifetime Value: Customer Lifetime Value(CLV) is an approach to determine the expected monetary value of a customer to the company during their expected lifetime of association with the company. The idea here is simple and similar to Return On Investment(ROI). If you invest to acquire a customer, you expect a return. Using the traditional CLV formula, companies can calculate what they can expect from a customer during their lifetime, which helps them determine how much they should be investing in acquiring these customers. ML algorithms can bring a whole new flavor to this analysis. The current CLV approach makes many assumptions to determine CLV. On one hand, ML algorithms can help bring more science around these assumptions by analyzing historical data. On the other hand, we can use ML algorithms themselves to make CLV predictions, in conjunction with other algorithms, like cluster analysis.