Boosting Telco Marketing Capabilities using Market Basket Analysis

Ever wondered how Amazon/Flipkart/Netflix knows what items/movies to suggest to you before and after you make a purchase or watch a movie?

The trick is an aspect of affinity analysis created specifically to promote sales called market basket analysis (MBA). In the B2C/ retail industry, MBA refers to an unsupervised data mining technique that discovers relationships among customers’ purchase activities. The technique is based on the theory that if you buy a certain group of items, you are likely to buy another group of items.

Example: while purchasing grocery at a retail store, if you buy Bread and Milk, you are more likely to buy Butter than someone who did not buy “Bread and Milk”

The application of market basket analysis in retailing is based on the notion that “most customers make purchases either on impulse or because there is natural affinity between the items they buy,” and that there are other items a customer would likely purchase when buying one item. Market basket analysis helps retailers discover those other items.

Let’s take one more example:

There is a very high chance of buying an HDMI cable if you just bought a smart TV , especially when offered the suggestion that an HDMI cable compliments your new smart TV.

The volume of sales made from user clicks on Amazon’s “Customers who bought this product also bought these products…” call to action links is a testament to the effect and importance of Market Basket Analysis

Now, Let us see how we can make use of this technique in B2B Context, but before we do that, let us dwell a bit on Cross-Selling…

Cross-Selling is the marketers most preferred way of generating more revenue from the existing clients. While acquiring new customers has become costlier, It has become imperative for the businesses to enhance their growth potential from their existing clients. Apart from the obvious profit from extra products sold,

It also increases the dependence of the customer on the vendor and therefore reduces churn by increasing stickiness.

This is especially important in the area of telecommunications, characterized by high volatility and low customer loyalty. Customer loyalty is quite low in this sector, mostly due to anti-monopoly measures taken by governments, as well as lucrative offers for new customers from other service providers.

Up-selling/Cross-selling assumes greater importance for cellular operators as well as telecom equipment providers since the more services a user has activated the closer he/she is tied to the company, and the harder it is for him/her to switch to another provider. Similarly, the greater part of your product/services portfolio comes from a single equipment provider, the greater your dependence (due to maintenance/ service issues and contract renewals etc.)

Although Market Basket Analysis is most often used to develop planograms and to derive shoppers insights, It is important to realize that there are some areas within Telecom domain, where we can use this technique to tune our marketing pitches. This can be done by analysis of telecom products/services purchases and using this data to develop association rules that will then help us design our product and services offerings.

Theory — Market Basket Analysis

Before I go any further, Let me briefly explain some key concepts and definitions associated with Market Basket Analysis –

• If an itemset satisfies minimum support,then it is a frequent itemset.
• Rules that satisfy both a minimum support threshold and a minimum confidence threshold are called Strong Association rules

We used this technique in the context of the transaction data that we had with the intent to

1. Identify most selling combination of products from our product/services portfolio
2. Identify what items tend to be purchased together by our customers
3. Develop purchase profiles with this information
4. Help sales teams induce add-on sales by suggesting products to other customers with similar purchase profiles (as shown below)

Implementation

Here’s an example of how the transaction wise item list that is actually an input to the algorithm looks like: (this is just an example data set)

Since, we have only 7 transactions here, it is quite easy for us to do profiling and come up with product/service recommendations.

For example, in case of Profile 1, we can observe that “CEE” and “Activation” are sold together in 3/7 transactions and in 2 out of those 3, we also sold “User Management SW”

This insight can help us develop a hypothesis that we can pitch “User Management SW” to customers who are purchasing “CEE” and “Activation” in a bundle.

When we have millions of such transactions, which is often the case, this work is made easier by Apriori algorithm that provides us with the needed association rules to develop similar insights.

The basic principle of Apriori is “Any subset of a frequent itemset must be frequent”. Let me give you a glimpse of the output that we got when we used this algorithm on sample data set, just to make it a bit easier to interpret the output.

Here I have just selected the Top 3 rules and sorted the rules basis lift value (The output is just a sample output derived after executing algorithm on an example data set in accordance with data privacy)

The interpretation is pretty straight forward:

• 75% customers who bought “Activation” and “Charging” also bought “Mediation”.
• It is 16.77 times more likely that a customer will purchase “Mediation” if he/she has purchased “Activation” and “Charging”

Insights like these give the ammunition to equip our business development teams to prepare their business pitch.

Plot of Top rules

MBA aims to find relationships and establish patterns across purchases. The relationship is modeled in the form of a conditional algorithm: IF {X} THEN {Y} : In shorthand notation, means an association rule that translates to “the items on right are likely to be ordered with the items on the left”

To interpret this plot, we just need to observe

1. Incoming arrows to a bubble — lhs
2. Outgoing arrows from a bubble — rhs
3. Bubble size — proportional to lift value

For example, Looking at rule 1, Customers who bought “Activation” and “Charging” also bought “Mediation”. Similarly, rule 2 is same as we got in our table earlier. This plot just makes it easier to interpret and understand the various associations.

Value for Business : How this analysis can be consumed?

The next logical step is to find transactions in which we had in-complete baskets and the associated revenue that could have been added to the sales tally with a high degree of certainty if we would have pitched the products/services we just identified. Presenting the analysis in terms of how it can benefit the organization and coupling it with dollar value can help us make a serious and compelling case, that is backed by data!

Summary

Telecom service providers have an invaluable asset in the form of volumes of data about customers. In the past, this information would go un-utilised because of unwieldy volumes and lack of computing power, which prevented structuring this data and using it in meaningful ways. Today, however, we have the means to tap this information and use it to unlock levers of sustained performance improvement. Breakthroughs in data storage and crunching power can transform the mountains of data into a marketing goldmine!