This bank wondered if they could use their data to understand their ideal customer and identify opportunities. Here's how we helped them use machine learning to achieve their goals.
A growing bank was relying on word-of-mouth and referrals to acquire new commercial customers. Decisions on the best target customers were based primarily on a banker’s personal experience, and prospecting was somewhat random rather than based on strategy, making the entire sales pipeline unpredictable and difficult to scale.
In addition, the utilization of the CRM was inconsistent and incomplete across the organization, and data was siloed and decentralized. All of this limited the effectiveness of using data insights to support sales and marketing.
The bank’s leadership was seeking answers to mission-critical questions that could allow them to grow market share in the region. These questions included:
- Who is our ideal customer?
- What characteristics, behavior, and experience with the bank make a customer ideal?
- How can we use what we know about our ideal customers to identify the best prospects?
- Can we improve the number of prospects that become qualified sales opportunities?
- How many ideal target customers are in our region?
- Can we develop an outbound marketing program for our prospects and effectively use CRM to manage it?
- Is there a way to prioritize which prospects to target?
- How can we improve the quality of data used for sales and marketing?
- How can we create a true segmented marketing strategy?
Bank leaders wanted to change the paradigm used by the bankers to influence the focus of sales and prospecting efforts, so they brought Fusion in to assess the situation.
ON-DEMAND BANKING WEBINAR: Learn how to turn data into insights that drive cross-sell revenue
Fusion presented a strategy and machine learning approach to build a solution that would allow them to identify the ideal customer using data, target prospects based on the understanding of an ideal customer, and identify new opportunities in the market.
Machine learning is a data science technique that uses the breadth and depth of data, especially historical, to rapidly predict or forecast future outcomes, behaviors, and trends. The technique lets the data learn from itself without human bias, preconceived notions, or the need for explicit instructions — and because the quantity of data can be massive, machine learning can identify patterns that cannot be uncovered or recognized by humans, especially when there is a need for real-time decision making.
There’s much more to this story than just defining a problem and throwing machine learning at it. The real story is about the process you take to find your solution. It is an iterative process where you learn from the insights gained and use them to make continuous improvements.
Our approach to identifying the ideal customer and improving prospecting for this client consisted of five key steps:
1. Defining the characteristics that make up an ideal customer.
To gather different perspectives within the organization (finance, marketing, sales, etc.), we spoke individually with key business stakeholders and asked what they believed were the characteristics of the ideal customer. Then we brought the stakeholders together in a workshop to create alignment and clarity about which attributes they collectively would identify with the ideal customer. Next, we brainstormed and selected high-value uses cases for the machine learning models.
2. Understanding their data, its viability, and its readiness to support the “ask.”
Machine learning initiatives are only as successful as the quality of the data, so profiling the bank’s existing data was a necessary step. We identified all available data and performed an analysis to assess data quality and completeness to support the defined objectives. Identifying where to find the best source of the data was an essential part of this exercise. Analysis at this stage can reveal where remediation must occur to either improve any data deemed to be of poor quality or fulfill gaps for essential data to best support the objectives. After the remediation of the data elements has been performed, there is a solid foundation upon which to leverage machine learning.
3. Developing a machine learning model that leverages the significant characteristics for use against prospects or existing customers.
In this step, we allowed the customer data to speak for itself and identify characteristics of an ideal customer. This involved:
- identifying data elements that should be input for the machine learning model based on the data profiling
- provisioning a cloud environment and developed data ingestion
- defining and developing machine learning predictive models that supported the defined use cases
- executing the model against real data and assimilating the output to graphically show the customer segmentation
4. Finalizing the ideal customer definition and refining the model.
We used stakeholder inputs, data profiling outputs, and machine learning to let data and actual outcomes influence the definition of the ideal customer. We explained to stakeholders what the model said was the ideal customer compared to what they said. This knowledge enabled a discussion that led to alignment about a final definition. Then we agreed on the final criteria/attributes and how they would be used to align with specific prospecting initiatives. Ultimately, the definition of an ideal customer is based on the context of your objective and, therefore, will result in multiple profiles that align with those objectives. Next, we built the machine learning models that would score (or rank) customer and prospect lists against the ideal customer model. Then we operationalized the model for use in marketing campaign processes.
5. Running the model against lists of potential or existing customers for the purpose of acquiring new customers and business.
In this stage, we developed the prospect target list and scored the prospects. When ready, the company can execute the marketing plan based on the prospects. Over time, there is an expectation that new models, based on different features, are developed to align with different sales and marketing objectives.