Six ways to use Data Science to drive your cross-sell and upsell activity

Share this post:

Getting more from what you have is the key tenet of the cross-selling and upselling mantra. By running deeper analysis on your customer information, you get more from your data. You then take these new insights and get more revenue from your existing customer-base by offering them relevant products and services. It’s a win-win process, when it works.

Increasing customer lifetime value works for both the business and the customer. More revenue (with lower marketing spend) for the business and more relevancy and loyalty from the customer. Yet, achieving this sweet spot requires complex analysis of the customer, their actions and their needs.

Here are six ways data science can help drive additional customer value:

Analyse & segment your customers

Before an analysis of customers for cross selling can be done, the dataset must first be divided into segments or cohorts, based along shared attributes, such as average spend, age, location or gender.

To successfully segment a dataset requires detailed data mining techniques to correctly decide how and where the segments should be created. Once created, cohort analysis enables the organisation not only to view which customers might buy more, but also to understand what the might buy, and when.

Understating these basic facts about your segments will allow you to begin to test and implement changes to increase any segment’s value.

Demo: how to segment your customers with SPSS Modeler

Modelling for uplift

One of the key goals in this analysis is to implement net lift modelling. This technique develops targeting or predictive analytics tools that not only identify who can spend more, but also the likelihood of whether they will do it or not.

Uplift modelling is a technique that uses a randomised scientific control to both measure the effectiveness of a marketing action and also to build a predictive model. This model predicts the incremental response to the marketing action.

Increase customer lifetime value and grow wallet share with predictive analytics.

Pick the next best product

Once a key audience has been identified for cross or upsell, a compelling offer must be selected. Next best product to recommend models are the foundation of cross-sell targeting analytics. These encompass triggers, segmentation, regression models and optimisation.Such models provide answers to the what (product), whom (customers), when (timing) and how (channel) of this exercise. Further consideration in the model also needs to be given to inter-purchase time, especially in retail.

Learn more: five ways to maximise customer value

Market Basket Analysis

Using historic analysis of customer data can highlight if a certain combination of products purchased makes an additional purchase more likely. This is called market basket analysis (also called as MBA). It is a widely used technique to identify the best possible mix of frequently bought products or services. This is also called product association analysis.

Association analysis is mostly done based on an algorithm named Apriori Algorithm.  The Outcome of this analysis is called association rules and can be implemented into marketing activity to trigger upsell and cross-sell actions.

Combine forecasting with predictive analytics and decision optimisation to create insights and turn them into actions

Clustering

By using machine learning techniques, multivariate data can be mined and processed to identify groups of customers who display closely matched activities and traits. These commonalities to make them likely to behave in a similar fashion to one another when presented with an offer or incentive for increased spend.

Cluster analysis is not one particular algorithm but the general task that needs to be solved. The appropriate algorithm to use in each case depends on the individual dataset and intended use of the outcomes.

Video: Get started with Watson Machine Learning on IBM Data Science Experience

Deep Learning and Deep Neural Networks

Deep learning is a part of the wider area of machine learning. The main differentiator between the broader set of machine learning and deep learning is that deep learning applies a greater level of learning on the technology than some more task-specific techniques found in other areas of machine learning.

Deep learning removes more of the human element of cross-sell and upsell analysis and allows a system to learn and implement its own findings from the underlying data.

Read more about deep learning on IBM Data Science Experience (DSX).

 Conclusion

Identifying and executing a campaign to increase the lifetime customer value will add to your bottom line. Using some data science techniques in your analytics will help you identify the best customer to target, with the best offers, at the best time. If you want to learn more about how we help organizations grow revenue with analytics, please get in touch with me at schristoffersen@dk.ibm.com.