Burning Leaf of Spending - Tatiana Bokareva

Learn Data Science

Burning Leaf of Spending - Tatiana Bokareva

0 Kudos
Started ‎09-15-2017 by
Modified ‎09-15-2017 by


Banking Industry



Tatiana Bokareva


burning leaf of spending, data science, banking industry

About the InsightsAbout the AnalyticsAbout the Benefits

Our lives are constantly changing and marked by life events. This is part of life and is often what connects us. Some events are happy like graduation from college, a new addition to the family or traveling to a great destination. Some are difficult like moving home and others may be traumatic like illness, death or divorce. Each life event has associated emotions and often carries immediate and future financial needs.

Gone are the days when customers remain loyal to a single financial institution. Many financial institutions don’t even expect to meet their customers face to face. If they are to remain relevant, it is imperative that they detect, classify, understand and predict which events a customer is about to undergo or has recently undergone. That ability enhances the organisation’s view of the customer. It enables a deeper understanding of their needs, complements the product and advisory  capability of the institution, and improves the relevance of next-best offers and conversations. 

In fact it is increasingly the expectation of customers that the company they do business with will understand their journey, needs, and preferences.

One of the way to identify life’s events is to look at the customers behavioural pattern and infer events from them. This analysis was performed to identify significant variations in the customer’s spending patterns.

The visualisation shows the significant change in customer’s average weekly spending. Each line on the graph represents a customer’s spending time series. The spike or fall in the weekly spending can potentially signify a life event, such as possible school fees, likely new dependents, and significant purchases. Such variations can also be used as one of the features in the propensity model to predict possible cases of defaulting on payments.

Teradata Aster Analytics was used to integrate and process rich transactional accounts and credit card spending data for 2016. The Change Point Detection Function (CPD) was used to examine customers’ spending. CPD functions detect the change points in a stochastic process or time series. Most customers have 1 to 6 points of significant changes. The number of significant changes ranges from 1 to 10 and is represented left to right in the graph. Far fewer have more than 8 points, the tail of the graph.

For the purpose of the art visualization, the data was smoothed with the LOESS method and the visualisation was produced by the plot stream package in R studio.

Understanding life events gives a more complete picture of who the customers are, where they are in their life journey, and what their financial needs are. Any major life event can trigger a sales opportunity. A marriage, kids enrolling in a private school or going to college, buying a house, having a child, or traveling overseas are all points along the customer lifecycle that are ripe for a targeted sale.

To identify events for up-sale or cross-sale offers, companies must have visibility across the entire customer lifecycle; from the first touch point until the present day. Moreover, the events library, the ability to expand, analyse, and act upon it needs to be accessible to the business across an entire organisation.

An ability to act on these insights develops stronger business relationships with customers, encourages and rewards loyalty, improves NPS scores, and pinpoints which particular financial products are most appropriate at any given point in time. In the end, all of these will lead to larger customer lifetime value and increased profits.

The Art of Analytics
To learn more about the Art of Analytics business case visualization initiative, please visit us on teradata.com/artofanalytics
Aster, Teradata, and the Teradata logo are registered trademarks of Teradata Corporation and/or its affiliates in the U.S. and worldwide

Version history
Revision #:
1 of 1
Last update:
‎09-15-2017 08:15 AM
Updated by:
Labels (1)