.

Increasing Digital Wallet Activity Using Feature Generation

Introduction

Some reports estimate that the global mobile wallet market was worth USD 1.04 trillion in 2019 and will reach a whopping 7.5 trillion by 2027. This is not surprising in light of a Juniper Research prediction that nearly 50% of the world will use mobile wallets by 2024.

For telecom operators, mobile wallets are a lucrative revenue stream. On one hand, operators have the advantage of a large existing subscriber base that can be easily transitioned to their mobile apps. On the other, they can link their customers with select brands, giving the telco additional revenue through partnerships. Thus, operators depend on customers using their digital wallets regularly to purchase services, interact with brands and avail offers, which makes tracking customer activity vital to improve profitability.

Framework for customer inactivity analysis

Analyzing customer inactivity benefits marketing teams by helping them focus their campaigns on disengaged customers, thereby motivating action and improving revenue at a lower cost.

Since data analytics is the underlying driver in assessing customer inactivity, companies require a structured implementation framework that contains the following elements:

  1. Feature generation or selection

Bigger the dataset, the better the accuracy

Telecom operators typically have large stores of customer transactional data comprising valuable information that can be reframed as important features such as:

  • Total revenue generated by the customer – high, middle or low
  • Amount of money sent and received
  • Volume sent and/or received
  • Days of mobile wallet usage
  • Average wallet balance at the start and end of each transaction
  • Average revenue per user (ARPU)
  • Customer age group
  • Date when the wallet was last funded from the customer bank
  • Date of the last withdrawal
  • Location of the customer and the last MNO activity (if the mobile wallet is associated with an MNO operator)
  1. Customer segmentation

Identify the right customer targets

‘Total revenue generated by the customer’ is a valuable parameter as it supports the categorization of mobile wallet users into low value, mid-value and high-value customers through a percentile rank-based model. A customer generating revenue above the 75th percentile may be tagged as a high-value customer (HVC). Customers generating revenue above the 50th percentile are mid-value customers (MVCs) and those generating above the 25th percentile are low-value customers (LVCs).

This segmentation provides the baseline behaviour to tailor marketing campaigns. For instance, HVCs have a greater potential for revival and revenue generation over LVCs. Thus, targeting HVCs with attractive offers will generate higher conversion at a lower cost compared to generic campaigns for the entire pool of inactive customers. LVCs also tend to display cherry-picking behaviour, i.e., their activity depends heavily on offers, whereas HVCs tend to be more loyal with more days of usage.

  1. Inactivity identification

Inactivity insights drive better recommendations

Diagnosing the root cause of customer inactivity is important and involves identifying customer patterns as described below.

  • Understanding customer activity – Looking at the distribution of transactional data of the inactive customers, i.e., whether they sent or received money, it is possible to identify whether the customer was on the network only to receive funds (receiving customers) or send funds (funding customers) or both. Receiving customers are quick to become inactive once their funding stops. Thus, by looking at their connections, operators can identify whether the customers funding these receiving customers are also inactive.
  • Understanding customer type – A customer’s spending pattern is ascertained by viewing the difference between the average wallet balance at the start and end of each transaction. If the difference is zero, it is a ‘depleting’ customer. If the start balance is greater than the end balance, it is a ‘savings’ customer. If the end balance is higher than the start balance, it is a ‘receiving’ customer. Applying a correlation test like Chi-Square between customer inactivity and spending patterns provide insightful associations. In Subex’s experience, customers with a difference of zero (i.e., depleting customers) are highly likely to become inactive.
  • Understanding customer intent – A close look at recent customer transactions can throw light on their intent. For instance, customers that leave an end balance of zero after the most recent withdrawal (to bank or cash) indicate an intention to leave the mobile wallet platform. A combination of these parameters (wallet end balance, last funding from bank and last wallet withdrawal to bank or cash) signals a new customer feature – ‘intent to return’, i.e., whether a customer has some, high or no intent to return to the network based on his last transaction.

As explained above, a withdrawal with zero balance indicates ‘no intent to return’. Similarly, we can assume that a wallet withdrawal (to bank or cash) with an end balance greater than zero signifies ‘some intent to return’ to avail the remainder of the money. This assumption entails setting a certain threshold value as wallets with a balance of, say, 5 cents could indicate that the customer has left the network. Finally, if a customer topped up his wallet before becoming inactive, he has ‘high intent’ to return to the network. In Subex’s experience, there is a strong correlation between customer revenue segment (HVC, MVC, LVC) and customer intent (no, little and high) whereby LVCs display no intent behaviour, MVCs display little intent, and HVCs display high intent behaviour. These correlations are extremely useful to focus campaigns on the right audiences.

  • Understanding customer spends – Customer age and location are useful in cases where the mobile wallet operator is associated with a mobile network operator (MNO). Classifying customers into different age groups and mapping these to revenue generated, customer count, etc., provides a wealth of information as shown in Table 1.

Table 1: Mapping customer age to customer type using analysis of variance (ANOVA)

Customer age (in years) Customer type
12 to 18 Low-value customers
18 to 24 Mid value customers
24 to 30 Mid value customers
30 to 36 High-value customers
36 to 42 High-value customers
42 to 48 High-value customers
48 to 54 Mid value customers
Above 54 Mid value customers

 

  • Understanding customer status – ‘Last activity on the network’ is a smart and quick way to determine the reason for customer inactivity. Checking MNO-related customer data like the last SMS or voice and data usage can diagnose the root cause for inactivity. For example, wallet inactivity of a customer with no SMS, voice or data usage in the last 60-90 days indicates inactivity on the MNO. Conversely, if the customer has SMS, voice or data usage over the past 60-90 days, then their wallet inactivity may indicate they are low-value customers.
  1. Comparative analysis and target identification

Targeted marketing for customer conversion

Armed with information like baseline behaviour, customer features, and possible reasons for customer inactivity, one can perform a comparative analysis across customer revenue segments, i.e., high value, mid-value and low-value customers. The results of this analysis provide actionable insights for marketing teams to design personalized campaigns aimed at changing the behaviour of customers across each segment.

A real-world example

A well-known carrier with operations across Africa implemented the above framework to improve customer activity on its mobile portal and increase revenue. Post customer segmentation, they identified 1.7 million inactive customers across high value, mid-value and low-value segments. Applying inactivity identification and comparative analysis, the operator curated marketing campaigns for each segment. They reached out to HVCs through calls, offers and SMS. The MVCs were sent tailored offers and SMS while the LVCs only received SMS. By targeting customers based on their past behaviour, the operator was able to increase activity on the mobile app and revive 31% of its customer base into revenue-generating users. Further, the campaign boosted customer loyalty as returning customers began to engage prolifically on the platform.

Conclusion

One of the chief ways for operators to sustain revenue is keeping customers engaged and transacting on-network mobile wallets. Customer inactivity analysis helps operators hone their understanding of why customers discard offers or abandon a network. Through a four-step framework, telcos can access customer transactional data to perform feature generation, customer segmentation, inactivity identification, and comparative analysis. Insights from these analytics will make marketing campaigns more effective by powering targeted offers that encourage customer activity on mobile wallets, cumulating in higher revenue for operators.

Author: Puneeth Kumar M U

The article is originally posted on Virtual-Strategy Magazine.

Views: 31

Tags: Analytics, Data, dsc_analytics, dsc_fintech

Comment

You need to be a member of Data Science Central to add comments!

Join Data Science Central

© 2021   TechTarget, Inc.   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service