Our client is a major financial services organization with millions of European customers and a presence in nearly every community. With a strong stable of brands, their Retail division serves millions of customers through one of the largest branch networks, in conjunction with digital, telephony and mobile services. They provide a comprehensive product range that includes current accounts, savings, loans, mortgages, and insurance.
The Vision
The effectiveness of their customer advisors is key to customer satisfaction and the bank’s success. To help understand what influences this, the client was keen to investigate how analytics could help explain what contributed to customer advisor effectiveness and if including additional sources of data could provide further insights. Would it be possible to combine the disparate sets of data underlying one of their Key Performance Indicators and consequently identify the drivers around customer advisor effectiveness? Could analytics help identify any consistencies or differences across brands that could further explain what was important? To this end, a Proof of Value was commissioned.
Fulfilling the Vision
Our approach to analytical projects is based on the Cross Industry Standard Process for Data Mining (CRISP-DM). We use it because it:
- focuses on understanding of the business value and data (available & potential)
- acknowledges the need for strong data preparation – mapping & merging of multiple sources
- addresses the need to deliver capability to end users for repeatable analysis as scenarios change
A Senior Manager in the Business Insight Team had this to say:
Thorogood were great at understanding a real business issue we were facing at the bank and showed us the power of ‘big’ analytics tools to highlight correlations that we could otherwise have missed.
Thorogood worked closely with the Bank’s Business Insight Team to understand their business questions and the data they had available – this enabled us to identify some initial scenarios to investigate. The next step was to combine data, including branch attributes, sales activities and new business volumes. We needed to understand its granularity, address data quality issues and agree business logic on how to merge the information until we had a consolidated set of data that we could start to analyze. We then used a variety of data visualization and statistical techniques to get a better understanding of the data and relationships, and identify the attributes of the most productive branches:
- Data visualizations to spot patterns
- Correlations to understand relationships and possible causations
- Decision Trees to segment the data into similarly performing groups
- Box Plots to understand the spread of data and identify outliers
Using this approach we quickly ascertained that one of the brands behaved very differently from the others. Analyzing this in isolation provided further insights, and we were able to identify the main drivers of best and worst performing teams for this brand, and thus gain insight into how learnings could be spread across branches to increase effectiveness. We used open source R to complete our analysis – a freely available statistical tool with 5000+ pre-built packages of common statistical techniques. To add extra insights, outputs were presented to the Business Insight Team using Microsoft Excel and Tableau visualizations. The company has found real value in this exercise, confirming some hunches and introducing a new understanding of branch and customer advisor effectiveness.