Advanced analytics may not be able to put a big dent in a cost base driven by expensive branch networks, but it can be a source of competitive advantage that will help banks secure revenue streams even under adverse conditions. Use cases exist in many segments of the banking industry. In this article, we look at the real-life application of advanced analytics in a retail banking use case before examining potential applications in other segments.
Proven impact across the entire banking value chain
Advanced analytics holds the potential to grow revenues, control risks, and increase efficiency across the entire banking value chain. Impact has been proven in real cases – examples include value-based client segmentation, eligibility assessment for cards or credit, and customer retention. For details and typical impact ranges, see Exhibit 1. It is time for banks to determine the most promising areas of application and reap the benefits before others beat them to it. But how does this play out in practice? In what follows, we will leverage an actual use case to illustrate the challenges and benefits of implementing advanced analytics at scale.
Deep-dive: Churn reduction analytics use case
A global bank was seeking to prevent churn in its consumer retail division. Prior to the effort, the checking account cancellation rate was in the magnitude of 5 percent per year. While the bank had win-back offers for selected cancelling customers in place, there was no pro-active retention program. The goal of the bank was to identify potential churners and reach out to them before they cancel their account. In the given case, the proactive churn prevention program helped the bank sustain ten times the revenue it had managed to reclaim with its previous win-back scheme. At the same time, better prioritization helped double the efficiency of the bank’s churn reduction efforts.
Step 1: Data cleaning
Working with messy data is dangerous. If the data you feed into a predictive algorithm is inconsistent or incorrect, you can easily arrive at inaccurate or downright erroneous conclusions. Before you know it, you will unleash churn-prevention measures on customers who are loyal, or miss potential churners and lose their business. Data cleaning and alignment are key prerequisites of any successful application of advanced analytics. In our case, the dependent variable of the model was defined as “Is the customer likely to churn”: "Yes" (1) or "No" (0)? Examples of predictors derived from historical data included:
- Products (credit card, insurance, custody account, mortgage lending, etc.)
- Demographics (age, gender, profession, number of persons in household, ZIP code, etc.)
- Customer information (segment, “customer since”, etc.)
- Volume (sum of balances of various products, etc.)
- Income (deposits, loans, securities, etc.)
- Interactions (adviser contacts, online banking log-ins, cash withdrawals, etc.)
- Rating (credit standing, credit, etc.)
Step 2: Modelling
All approaches have their limitations. Some algorithms will produce clear results, but tend to oversimplify the relation between predictors and target variables. Others are more accurate, but tend to be overly dependent on historical data (“overfitted”). This is why we recommend combining multiple algorithms to overcome the shortfalls, and leverage the individual strengths, of each one (“ensemble learner”). In the given case, the bank used three different types of algorithms, including decision tree and random forest modeling, to predict churn.
Step 3: Intepreting
Once the bank in our example had identified high-risk customers, it set out to determine the drivers of churn. As a first step, it used statistical clustering to create segments of customers with comparable behavioral patterns. In a second step, these segments were analyzed to get to the root cause of churn. In the given case, life events turned out to play a big role. Examples include:
- Young customers switching banks when they start their first job
- Married couples setting up a joint account and cancelling individual accounts
- Empty nesters consolidating their pension accounts
Based on these segment profiles, the bank devised a set of actions to prevent cancellation.
Implementing advanced analytics at scale
To create sustainable competitive differentiation, banks will want to apply advanced analytics not only in selected areas, but across their entire business. The prerequisite for advanced analytics at scale is an agile organization. Agility is about balancing stability with flexibility in a way that is conducive to sustainable success in a dynamic environment; compare our article on the “5 Trademarks of Agile Organizations.”
To implement advanced analytics at scale, banks will want to take a step-by-step approach:
- Define the bank’s advanced analytics ambition and strategy
- Build and execute a pipeline of use cases to generate buy-in of key stakeholders
- Build the foundations to scale; see Exhibit 7 for a glimpse of the future.
Key take-aways
- Because of high cost pressure and disruptive market conditions, banks should take advantage of advanced analytics to generate competitive differentiation and protect their market share.
- Advanced analytics is more than algorithms. Thorough data preparation, diligent interpretation of the outcome, and effective actions to monetize the insights are equally important.
To take full advantage of advanced analytics, a bank should treat data as an asset, implement agile principles in its organization, and make analytical capability a core competence across all its functions.