# Statistical Segmentation in the Wild

Some may recall from an introductory statistics course that to calculate the average of an average is a statistical error.  Yet, it’s done all the time

Let's look to nature for some examples to illustrate this issue.

While the duck-billed platypus is a mammal, it is one of very few remaining primitive mammals, termed monotremes, that lay eggs. The baby platypus arrives 10 days after the egg is laid.  Another water loving mammal, the hippopotamus has a 243 day gestation period, while the elephant’s can stretch to 660 days.  Obviously, it would be an error to say, the average time for a mammal to be born is 335 days (10 days for platypus and 660 days for elephant).

What if used a weighted average: there are approximately 100,000 platypus (I promise that platypus is both singular and plural!) in the world and 50,000 elephants.  Therefore, the weighted average given the population of 150,000 combined mammals is 226 days.  Perhaps one is emboldened because 226 days is pretty close to the 243 day hippopotamus average.  But it's nonsensical, right?!

We can all agree that neither mathematical result provides accurate information from which to make an informed decision, but I see this type of analysis done all the time.

Like the elephant, hippopotamus, and platypus, credit union member segments vary significantly. Consider the behavioral pattern of CD-only vs. indirect auto vs. engaged-checking members.  Even within checking account member segments, a digital member vs. traditional member vs. executive member are all quite different.  The revelations of how member’s choose to engage via products, services, and transactions can be viewed in statistical clusters, yet they are distinctly different.

• Dozens of intuitive actions are revealed as one reviews member engagement segments and the variations within segments. Engagement varies even with different checking account segments.  To view the number of debit card, ACH, monthly deposits, branch transactions, fee income and average balances results in easily targeted actions.
• Attrition models should be constructed within each segment. Clearly attrition-alerts for members within the traditionalist segment (engages via branch and check transactions) versus a digitalist (mobile logins, debit card and ACH transactions) will be based on different components.
• Wouldn’t you agree that the Next Best Action, the Next Best Product, and the Ideal Product/Transaction/Service Mix would vary by segment? And how to measure success of Actions without a quantitatively proven Engagement Score?
• Finally, the ultimate goal is to have dozens of targeted prescriptive-actions (statistically proven from past data and trials resulting in prescribed actions via measured-likelihood percentages). The action results will further impact segmentation results or suggest clear sub-segments for future actions.

In conclusion, engagement, next action, and next product offer should vary by segment. To use the average of all members, or average of all members with checking likely produces an offer or action that doesn’t meet the needs of any particular segment. The ideal-bundle of products, transactions and services is quite different for no engagement, deposit, loan, digitalists, traditionalists, and executives. To have a single action plan, algorithm, or model for all members makes as much sense as saying platypus to elephant average is proven because it’s close to the average for hippos.

# Segmentation and Computation: Part 2

Significant time and resources are focused on data collection and cleansing steps.  While a single-source-of-the-truth in the form of a data warehouse are both foundational and necessary goals, often the biggest immediate ROI contributor lies within the computational layer.  These calculated fields, or synthetic data elements, are the most predictive, action-oriented elements of models and equations.  Specifically, what are the “most important” data elements that drive action and result in immediate alerts at your credit union?

When we ask financial institutions what specific outcomes they seek to use data to make better decisions on, the responses tend to include:

• More Members
• Increase Loans and Deposits
• Increase Member Engagement
• Better manage risk
• Improve bottom-line results

We’re all pretty capable of effectively measuring the first two; at least top-line measures by product, sub-product, region, branch, etc.  There is an art to measuring attrition and some science is involved when measuring within different member segments . . . but we’ll leave these discussions for a later post.

But how do you measure member engagement?  We’d likely all agree that strong new member on-boarding is key to engagement.  But, how do we measure on-boarding?  What is “success”?  Which branch does it best?  Which team members stand-out?

Even if we start improving on-boarding, do we know how it impacts our bottom-line results? Most credit unions don't or struggle to calculate member contribution (or profit) by member.  Most cannot tell us the average profit per product much less their “ideal product mix”.

Consider that the “most profitable bundle” of products has an average-line with some members above and below that average.  The challenge, of course, is how to add more members above the average-line to preserve, or ideally increase, the average and total contribution?  However, without these critical measures it’s likely that we can land more members into that “ideal product mix” while we drive down the average-line and reduce the average member contribution of said bundle.

Consider the importance of segmentation.  Members choose how to engage, and, therefore, member segments vary significantly.  Consider how differently members choose to engage within the following segments:

• Indirect vs. direct members
• Deposit-only versus loan-only members
• Digitally-focused members versus traditional banking members

Because of these differences, the next action (i.e. the next-best-offer), also varies.  The goal is to have a path for each member segment towards the ideal product, transaction and service bundle.  The contents of the message, channel, and timing will all vary based on the segment; and their maturity level within that segment.

In conclusion, executive management meetings, board room discussions and strategic plans often include goals related to improving member experience, member engagement, cross-sell success, better on-boarding, and improving bottom-line results.  Yet the vast majority of credit unions do not have quantified metrics to measure if things are improving, which branch does best, which team-members succeed most often, that generate the next best-action to move the meters that matter most.

I’m sure we’d all agree: What Gets Measured Gets Done.

# Segmentation and Computation: Part One

Every data analytics strategy contains the 4Cs: Collect, Cleanse, Compute, and Consume.  Collect and Cleanse are the focus of a data warehouse - aligning data and addressing disparate data silos.  Consume pertains to the user interface whether it be Oracle BI, Power BI, WebFOCUS, Tableau, etc.  Yet it can be argued that the Compute layer is the most important and is frequently misunderstood from three perspectives.

## #1: The Value of Calculated Variables or Data Elements.

As an example, it’s less important to know that an RDI occurred (return deposit item or charge back check), it’s more important to know that this was a first RDI on an account following a recent mobile deposit.  Such a scenario may trigger immediate action to place a hold and investigate potential fraud on the account. However, without an algorithm or model in place, this scenario often goes unseen.

Another example: consider the number of debit card transactions or ACH transactions on an account.  It's more critical, for analytic purposes, to know that, by mid-month, an account has a 50% decline in their transaction volume compared to prior months.  This is a potential indication of an account about to close.

## #2: What about Machine Learning and Artificial Intelligence?

The second perspective has been recently elevated by many credit unions wanting to define a strategy for Machine Learning (“ML”) and Artificial Intelligence (“AI”).  In this frequent conversation, I ask what is the dependent variable you’re trying to:

• Take action on
• Improve or increase
• Measure

If we dump large quantities of data into ML and AI models, be warned, you’re likely to get some statistically meaningful results.  But, you’ll waste significant resources running around and trying to build a model that does what?

ML and AI are best applied to data already aligned correctly that includes numerous calculated fields, and, ideally, used to improve upon an already proven model or equation.  In other words, data for the sake of data is meaningless.  What is your desired outcome?  Focus on that first and the data second.

## #3: The Right Messaging

The third perspective revolves around where we all really want to be!  We want to KNOW what message works best, to which member, using which channel, at what time, to help drive [insert your CUs goal here!].  That is, we want to move from predictive models to prescriptive action.  Doing this properly yields outcomes like: do this and our desired out come will occur 40% of the time.  Or, immediately do [action here] because 60% of the time this series of events is related to fraud.  These are the action lists the line-of-business leaders, branch and risk managers need to succeed.

Almost always, the calculated fields are the most predictive, that have the highest correlation coefficient and are the most valuable for predictive and prescriptive models.  It is the results of calculated fields, algorithms, and models, within specific segments, that direct the actions to generate the data analytics journey return on investment.

Interested in discussing these ideas further?  Drop us a note!

# Increasing Deposits through Advanced Analytics

Increasing core deposits has become a key priority for community financial institutions. While you could, metaphorically, fish with a net or throw darts blindfolded, using advanced analytics is an ideal way to grow your FIs deposits. In this post, we discuss three practical methods for using advanced analytics to efficiently grow deposits.

##### 1. Segmentation is everything: mathematically it may even be more

We’ve all heard about the 80/20 rule. But in many applications, the segment concentrations are even more extreme. Don't be surprised in looking at data from your core if critical deposit segments fit a 90/10 or 95/5 concentration!

There are two keys to consider when embarking on a member or customer segmentation path. First, consider how you should identify the segments in which the extreme-minority produces the extreme majority. In other words, where should you look to find the 90/10 or 95/5 concentrations?

The second key is in finding the best way to drive action within the groups you identify from the previous paragraph. It's one thing to identify groups to target and another to act on the information you uncover.

The solution is to have a quantifiable metric to identify and measure change. Specifically, a metric to measure engagement is strongly suggested. Consider asking the following questions to help understand your members' engagement with your organization:

• Identify which members clearly have a primary checking relationship with your FI. (In other words, which of your members treat your organization as their primary financial institution?)
• Which members have a checking account, and make a loan payment from that checking account, but have very few other transactions?
• For members who account for the vast majority of deposits, what products, transactions, and services do they most frequently utilize?
• Answers to these questions (and a dozen more!) will reveal many intuitive actions
• Developing an engagement metric helps to track these segments and measure changes in engagement for the critical segments identified
##### 2. What’s more important than onboarding success?

Repeated statistical analysis reveals that habits are formed in the first several weeks of a relationship with a new FI. That is, the product relationships formed during the onboarding process represent a significant percentage of the entire products and service relationship an individual will have with a FI.

Consider questions such as:

• How good is our onboarding and cross-selling process? Do we have well-defined goals? Do we achieve success?
• Do we succeed in deepening the member/customer relationship to include multiple deposit products and loan products, strong utilization of digital services, etc.?
• How do we measure success? Do we know which branch does the best job? Do we know which branch employee does the best job? What are the methods that lead to their success?

Measuring engagement at the two-day, two-week, and two-month mark should reveal the success of your onboarding process by region, branch, and employee. Generating more deposits is made drastically easier by achieving strong relationships during the onboarding process. Sharing information across the FI with other lines-of-business, coordinating follow-up, and measured actions are key.

Remember: what gets measured gets done!

##### 3. Characteristics of our best deposit generating members

Once we have an operational engagement metric, measured action results, measured onboarding success, and a completed segmentation analysis of our best deposit-generating members, several analytical methods can be applied:

• Consider a simple geographic cluster analysis to identify where the ideal depositors live. Through purchased or internally-mined lists, reach out through their preferred marketing channels.
• What are the most frequent product mixes (typically 2-3 different products in a cluster) that occur within your strongest depositors?
• How quickly can you recognize new members that could be “the best” and take action to help quickly deepen the relationship?

The key to success with growing your deposits lies within your data. Analytics should be empowering your journey and providing insights and key metrics. Consider the various bullet points discussed in this article and don't discount the power of measuring engagement!

# 4 Ways to Take Your Analytics Initiatives to the Next Level

How confident are you that your organization’s analytics initiative is truly delivering real value?

Did you know, more than half of all analytics projects fail because they simply did not deliver the features and benefits that were intended at the onset of the project?

Your analytics project should contribute significantly to loan, deposit, and customer/member growth while also improving risk, retention, and profit. Analytics and even advanced analytics (machine learning based models) are empowering larger financial institutions, while many community-based financial institutions struggle to enjoy significant bottom-line impact.

To accelerate your analytics journey, consider these four topics:

##### 1. Accelerate to Avoid Incremental Learning

Every journey begins with the first step. However, after every long journey or project, there are always things we wished we’d done differently. Advanced analytics, like machine learning, is often hampered by not knowing what we don’t know. Establish the 4 Cs of the process first: Collect, Cleanse, Compute and Consume.

Disparate data sources limit a complete picture of the members' relationship. Once Collected and Cleansed, low hanging fruit is revealed within these data sources that would empower immediate action for the lines-of-business. The Compute phase should result in meaningful metrics that provide leaders feedback and measure progress. The Consumption layer permits leaders to view trends and identify key segments for action.

Each of the 4Cs has multiple iterations. More granularity is needed for deeper insights. As an example, some FIs have all debit card transactions grouped together into a single transaction code versus having separate PIN vs. POS transaction codes to reveal important insights.

Analytics can be used in identifying several recurring debit card transactions such as gym memberships and utilities. Another example is the integration of internet banking and mobile logins: helpful data in the computation of engagement, retention, and segmentation of digital vs. traditional users. The compute phase is obviously limited when the “right” data elements are not present. Worse, experience shows that the most statistically-predictive components of an advanced analytics equation are synthetic-variables. That is, a calculated metric derived from other data elements. The process to identify these best-elements takes time. Avoiding incremental learning can accelerate your analytics journey 2-3 years!

##### 2. Accelerate to Avoid Incremental Learning

For most community FIs, the minority of customers provide the vast majority of many things. Dozens of sorts reveal immediate opportunities, such as:

• Pinpoint the members that provide the majority of deposits and loans.
• Determine how many customers have a checking product and use their checking account in a way that demonstrates your institution is clearly their primary FI.
• Answer key questions like, of those members with a debit card, what percent generate the most transactions. In most cases, less than 20% generate 80% of all transactions.
• Calculate profit and/or contribution. Many FIs can calculate that the top 1% of consumers typically generate around 100% – 150% of profit. Percentiles 2-15, generate another 150%. Percentiles 16-89 contribute a negative contribution of 80% – 100%.  The bottom 10% (familiar through bankruptcy notices and legal bills) contribute a negative contribution of 100%. The net result remaining is the net profit or contribution.
• Identify top performing consumers. If we open 100 new relationships today, how long does it take to identify those with the potential to be in the top 1% or top 15%? Can you measure on-boarding success? Can you measure customer experience revealed by an engagement metric?

What gets measured gets done!

##### 3.  The Role of Math and Machine Learning

So much talk of analytics and too little math is also a recurring obstacle to avoid. Often, math and machine learning are typically applied with little understanding of product and industry knowledge. Two entertaining illustrations:

• A client engaged a multi-month machine learning consultancy that informed the Credit Union that the primary share account was their #1 product; nearly everyone has one...obviously!
• A larger FI client, with Ph.D. mathematicians and statisticians on-staff challenged a complex model equation to minimize risk and loss. The equation did not include “customer account balance” as it had no predictive contribution. The bank staff disagreed as their findings showed that it was the #1 predictive component. The bank staff used the balance after the account was already negative. If the FI charges-off accounts on day 45 as a business rule, there’s a high probability that a negative account on day 44 will be charged-off tomorrow. However, that doesn’t enable a business leader to make an informed decision! Once they adjusted their challenge-equation to consider “customer account balance” 30 to 15 days prior to the account going negative, they agreed that balance was not helpful. This is another example of incremental learning.

The reality is, it takes industry and product knowledge coupled with statistical experience to reveal actionable insights. Dumping data into a neural network model or machine learning tools will have you chasing statistically significant, but likely-irrelevant, findings. Machine learning is best applied to improve or identify a specific segment on an existing equation or model. That is, Machine learning is an evolutionary, not revolutionary, tool.

##### 4.  Put Actionable Insights to Work: Measured Action

Being able to collect the right data is one thing, but making it extremely useful and action-oriented requires a different skill and mindset. Contributing an immediate ROI should be your top goal. Initial success should be easily achieved because many opportunities will be revealed through the right concentration and segmentation steps. You can measure success by comparing targeted action against historical baselines. Over time, actions will be targeted on the right individual, at the right time, via the right channel, using the right message. Expected probabilities will be assessed and tested against previous results.

Soon the FI will formulate best practices and metrics through the data and insights. Your segments will become more narrow and refined. The FI’s culture will evolve into a data-driven culture, welcoming team suggestions and recognizing the best performing region, branch and individuals. On-boarding and experience goals measured at two days, two weeks and two months will result in improved retention and member/ customer service. Top performers can be recognized and learning organization theory can lift the performance of others. Over time, the bank’s culture begins to shift, strategic initiatives are measured, resources are provided to team members in the form of information and action lists, and performance can be assessed.

Let Coastline Analytics accelerate your journey by two years with our unique approach to establishing your organization’s advanced analytics capabilities.

# Using Analytics to Increase Member Engagement

Who doesn’t want to improve their member’s engagement with your credit union?

What’s the best way to engage our current members? How will we measure what works? These are common questions our credit union clients ask as member engagement is a vital metric to calculate.

Understanding and engaging with your members does require a robust data management, data integration and data visualization environment. Through compelling, personalized member engagement, credit unions will have the opportunity to attract new members, boost current member loyalty, and expand the spectrum of products and services in an existing portfolio.

There are patterns and business insights within the vast quantities of data credit unions collect on each member. Data analytics can help the credit union identify and respond to what members need today and can help forecast what they will need in the future. Here are a few examples of how data analytics can provide knowledge and better insights into preferences and desires that ultimately boost member engagement and member satisfaction.

Know Member Preferences:

What members prefer a digital experience over visiting a branch? Knowing member preferences, expectations, likes, dislikes, motivations and inclinations will boost engagement. Using data analytics, you can determine how your members prefer to engage with the credit union—leading to happier, more satisfied members. Additionally, by analyzing information about how your members access and consume content and applications, you can build marketing strategies around the devices, locations and preferences they already prefer. Using these preferences, the credit union can create experiences that are most beneficial to each user, leading to new opportunities to further engage with members and target new ones within these segment preferences.

Know Impending Life Events:

Data analytics can provide powerful insights and help uncover customer behavior patterns that reveal life events such as, retirement, new home purchase, or a baby on the way.  By tracking and measuring important indicators in a member’s life – age, gender, marital status, income, a move, and more – you can segment members into easily targeted groups. From there, you can develop targeted marketing campaigns that create relevant, personalized experiences that improve engagement. Additionally, it will create the opportunity to cross sell and upsell the right services or products that are needed during specific life events.

Know Member Sentiment:

From Facebook to Instagram, these social networks can offer a wealth of information about what your members are saying and feeling about your brand. Your credit union can use data from social media to gain insights that would normally require surveys and questionnaires. Through data analytics you can better engage with members and react quickly to address and identify issues, complaints, and product or services flaws. Social media sentiment analysis and another tools can be used to determine member sentiment about the credit union through an analysis of their posts and social interactions.

Not every member is going to be profitable, but through segmentation, there are opportunities to maximize the number of profitable members. By using data to identify unprofitable members, the credit union can target marketing campaigns to further boost engagement and entice them with more products and services the credit union offers.  By knowing your audience better, you can create personalized, compelling marketing messages that are relevant and will promote engagement and lead to business outcomes.

Beyond the knowledge gained in each of these areas, it’s what you do with this knowledge and the data that will lead to big wins.  Not sure how to transform member data into deeper engagement?  Consider partnering your credit union with the data experts at Coastline Analytics that can help create a roadmap to leverage all the advantages of data analytics for more effective and personalized member engagement.

Want to know how you can further improve your members' engagement with the credit union? Let’s talk.  Contact me today at brewster@coastlineanalytics.com or call 860-593-7842.