Like a financial fingerprint, the analysis of customer transactions yield unique patterns. These financial fingerprints help us discover commonalties and differences between customers and ergo, better anticipate and service their unique financial needs.
In our last article we introduced and discussed a very powerful group of exploratory statistical procedures called Factor Analysis. In an effort to begin the data conversation, we outlined a step-by-step tutorial of conducting a Factor Analysis in order to uncover a data’s unique story. As we described previously, Factor Analysis is analogous to finding the structure of a data set. Often viewed as an extension of mathematical correlation, Factor Analysis gives us the ability to simultaneously see the correlations and the interactions between all variables. With this ability, we can dive deep into the data and begin discovering the unique story it has to tell.
Quick: Get the camera
We define life stages as pivotal moments when a customer’s financial service needs change. Determining the customer’s life stage would be obvious if we had the ability to look into the living rooms of each one of our customers. We could simply peek inside, see where they are and give them custom offers to fix all of their financial needs. Examples of life stages might be the sale or purchase of a home, the start of a new career, the decision to have children or the planning of one’s retirement. All these examples provide valuable images into the lives of our customers. Unfortunately, with the online shifting paradigm, we are losing the benefit of the personal banking relationship of yore.
Often, we make generalizations about a customer’s potential needs by gut feeling, an approach fraught with uncertainty. We may group customers based on age, the products they currently have and, if we are lucky, we may base it on some personal information we have about them. But what happens when our customers no longer come into the bank, when we don’t know them or when we haven’t seen them in years? With every touch point lost, we limit our ability to anticipate needs.
In the current environment, the most reliable connection we have to our customers is through data. By employing predictive analytics, we have the ability to create a mathematical model of the data in order to painter-create the picture that we have lost.
Following the clues
Life stages are those discernible moments that are determined by fluctuations in transactional data. In our last article we assembled a small data set of 100 customers with 3-4 years of transactions in order to begin the data conversation. Each transaction is a possible clue to the future needs of a particular customer. Transactions, or more importantly, deviations in transactions, provides insights into the nature of purchases a customer is currently making, and spending habits. On the other side of the equation, savings habits are also indicative of an upcoming need or may be used as an indicator of an upcoming event or purchase. The results are even more powerful when patterns of spending and saving habits are used in tandem.
Intuitively, we know that a transaction could mean different things depending upon the context of the transaction. Deviations in purchases are based upon individual decisions, however, it is possible to map customers to significant groups, that we call customer segments for simplicity.
Segmentation for life stages
Customer segments (or clusters) are defined as groups of customers whose actions define their similarities. The goal of customer segmentation is to assemble groups of customers who act most like one another. This simple structure has proven to be very effective for increasing the reliability of predictive analytics. To reach our specific goal of determining life stages, we define the clusters slightly differently, we look at clusters as ‘paths’ that customers take for product acquisition over the relationship with the bank. We want to define these paths so that we can learn more about them. We will discuss a similar but different approach when customers do not follow a distinct path, as used in wallet share maximization. Lookout for this in our next article.
The determination of life stages gives rise to a series of questions:
- How many different paths (segments) are there?
- What transactions are leading indicators of a potential customer need along a particular path?
- What paths are most profitable to my business? And why?
- How do my current customers align along these paths?
A step-by-step guide
Imagine for a moment assembling a very large group of customers with diverse portfolios comprising a varying mix of financial products. We use sampling to assemble a group that has customers who closely represent the majority of the customers within the institution. For a detailed description of sampling methods see our previous article, “Smart Banking”. You already have a detailed record of the products each customer has purchased and when they purchased each product, along with the recorded transactions leading up to the purchase. Each person in the group will have a slightly different history. Identifying individual paths and the life stages along each path becomes an iterative process of factor analysis, segmentation and classification and is performed in two (2) distinct stages, as described below.
Step #1: Perform a factor analysis on the complete data set.
- Determine number of required factors (explaining at least 85% of variability)
- Identify the most significant variables from factor loadings
- Eliminate variables that do not provide significant explanation
Step #2: Perform a cluster analysis of the remaining variables.
- Determine the number of unique paths
- Classify each customer to a path
Step #3: Perform a factor analysis within each cluster.
- Determine number of required factors
- Explain the inter-correlations between the variables within the cluster
- Define significant variables (transactions) that signify life stages
Step #4: Classify each customer to a path
The Big Picture
Life stages define common paths of customer financial need. There is often a large percentage of customers that have common needs. These needs, once identified, can be managed for increasing wallet share. By the analysis of transactional data these common trends can be identified. The predictable and predictive identification of need, allows us to better service our customers and build customer loyalty while increasing customer retention. By defining the life stages of our customers, we gain additional insights that can define and shape our business and marketing strategy by pushing beyond current offerings to identify potential sources of future need.
Determining life stages is the first step to wallet share maximization and realizing customer lifetime value. Life stages are the mileposts along the way. Stay tuned for our next article in which we begin digging deeper into customer life stages as we outline and develop the maximization of wallet share.