Close this search box.
Leveraging Salesforce to Identify Growth Opportunities In Wealth Management Using Lookalike Audiences and Machine Learning

Leveraging Salesforce to Identify Growth Opportunities in Wealth Management Using Lookalike Audiences and Machine Learning

Contributor: Gene Zeyger

From acquisition to growth and retention, one of the focus areas for many wealth management firms in this turbulent market is building better relationships with their clients. One way of doing that is to understand what the similar clients are doing, not only from an investment perspective, but across the board. Who are they? How are they spending? What do they care about? Top advisors really get to know their clients in order to provide personalized services, and tailor investment advice and services to their needs.

Ultimately, your goal is to find out what kinds of products and services you should position and recommend to individuals that they will most likely buy, based on historic data in their portfolios, at the right time. To make this happen, top firms are beginning to leverage machine learning and statistical analysis.

What Are Lookalike Audiences/Clients?

Lookalike audiences/clients share common characteristics that can be used to help drive marketing campaigns, client interaction, coaching, or other personalized actions. Identification of lookalike clients can be done in various ways, ranging from methods that segment clients based on business rules to methods that leverage machine learning to “cluster” similar clients together.

Key client data that help drive client segmentation might include: 

  • Age
  • Demographics
  • Asset Types
  • Purchase Behavior & Transaction Patterns
  • Balance/Account Information
  • Engagement

Understanding metrics related to these categories can help to drive a better understanding of where shortfalls exist in how client relationships are handled, and inform how they can be improved.

For example, let’s say you are looking to prioritize actions to prevent attrition. Like many firms, you have a diverse set of clients with different goals and desired outcomes. However, let’s say you identify a client that has a declining trend in banking transactions, which is an early indication of attrition, we’ll call them Client A. You can then leverage the data you have on Client A, such as their net worth and demographic characteristics, to identify other clients that are similar. This information allows you to identify Client B.

Of course, you would immediately engage with Client A, typically with a specific set of actions that are optimized for that of Client A’s type to prevent attrition. But more importantly, understanding that Client A and B are similar clients can help you proactively identify and work to keep Client B happy, preventing a possible attrition event altogether.

Platforms Needed to Successfully Leverage Lookalike Audiences 

Not all problems need to be solved with complex statistical models. In some cases, determining criteria for client segmentation can be done with a set of thoughtful, well-defined business rules. Even in this case, we recommend having a good plan to identify key insights related to specific types of clients. Typically, that is done by using analytics dashboards to track key metrics for each client type, set specific sales targets for specific client segments, and drive personalized insights for clients from each group.

Data-Driven Clusters: Specific Clustering Methods

If you need to leverage tools that are more dynamic than using a set of static business rules, it may be worth considering various statistical methods that can be used to perform client segmentation, and identify look-alike clients. Most methods to do this involve clustering clients based on a suite of KPIs and metrics. Many tools for clustering data exist, each with different pros and cons.

K-means clustering is a useful way to group similar clusters if you have numeric features to use as input. Model-based clustering, on the other hand, can assign a probability of cluster membership to each client, allowing you to quantify how well-classified a particular client is in a given group. Either way, the key idea is to determine a viable clustering method and use the right data to create a set of client clusters that can be clearly interpreted. Clusters should be interpretable, easy to define and differentiate, and drive specific, differentiated business action.

Most clustering tools need to be built on a set of historical data, and would need to be built using “off platform” tools outside of Salesforce. Some tools, like Snowflake, support dynamic cluster analysis using compute environments like Snowpark, that can be built in Python and operationalized in an orchestrated flow. Like all ML models, the efficacy of cluster analysis is limited primarily by the quality of the data used – so it makes sense to build clusters using a mix of statistical tools with the context of industry knowledge.

Like any statistical model, clustering models need to be monitored and occasionally updated as changes in data, business processes, and downstream uses change.

Using Lookalike Audiences to Attract and Retain Clients

Identifying lookalike clients is useful for understanding which clients are similar to each other, but how does this knowledge truly move the needle? Good client segmentation models can be useful in a variety of use cases.

The Most Efficient Approach to Client Acquisition

Identifying lookalike clients and understanding client segments is a critical part of improving client acquisition and driving sales efficiency. By better understanding the profiles of leads and opportunities that are more likely to convert into sales, you can direct your sales efforts at clients who are similar to that profile.

Additionally, some sales techniques, products, and strategies may be more effective with certain types of clients than with others. Identify the most efficient strategies for converting clients of a different type, and use those strategies as a directed playbook to make sales with different types of clients.

Additionally, you can drive smarter revenue operations by taking advantage of client segment information (such as conversion rates, time to buy, etc.) when setting sales quotas, defining goals, and measuring sales performance. Trying to focus on key client segments may also help an organization to grow efficiently (e.g., target white space in client profiles or “hone in” on a specific type of client) to gain market share.

Develop A Game Plan for Client Retention

Understanding common client dynamics is important for managing relationships with current clients, as not all clients should be treated the same. Driving differentiated relationship management based on client characteristics, or client segments, allows for a more personalized experience (for both in a B2B or B2C scenario) and can help bankers and managers be more efficient in getting the right products or promotions in front of the clients. This can be done in a handful of ways:

  1. Identify client segments leveraging business rules or a clustering model
  2. Determine baseline client behaviors in each segment
  3. Identify “early warning signs” of potential attrition – these may be different for different client segments. (This could be a machine learning model in and of itself.)
  4. Develop and test directed strategies for increasing client retention in each client segment

By developing a specific game plan for how to identify potential churn risks, and then proactively intervene to avoid it, organizations can be more efficient in keeping their clients.

Product Recommendations

Typically, product recommendations are driven by a good understanding of client segments. They are generally built off the idea that if Client A is more likely to buy a product, then a similar client, say Client B, is also more likely to buy that product. Less-similar clients, say Client C, might be less likely to prefer that same product (or mix of products).

Building a good client segmentation model and being able to identify lookalike clients is just the first step of building product recommendations, but it is a useful tool that can provide meaningful insights even as the other components of a recommendation engine are being developed.

Bringing It All Together

Atrium is uniquely positioned to help financial services firms become data-driven companies. We specialize in driving definable business outcomes and a roadmap for scale at the intersection of data science, analytics, and CRM strategy. We combine the power of Salesforce with actionable insights to help maximize value in your technology stack.

We are data science and machine learning experts hyper-focused on helping wealth management firms identify growth opportunities. Contact us to see how we can help you discover lookalike clients.