Running Up the Score: How Predictive Analytics Gives You an Advantage Over Your Competitors

“Football is like life. It requires perseverance, self-denial, hard work, sacrifice, dedication, and respect for authority.” 

— Vince Lombardi, former head coach of the Green Bay Packers

There’s an old saying in American football, “Offense wins games, but defense wins championships.” 

Every team in the National Football League (NFL) has an assistant coach called the defensive coordinator who is responsible for developing the defensive strategy, teaching it to his players and managing them during games. While the defense may not seem as sexy as the offense, which gets to score most of the touchdowns, it’s still a vitally important part of the team. 

The same holds true for technology such as artificial intelligence-powered Predictive analytics, which can function like the defensive coordinator of an advisor’s planning process. These tools can deliver insights such as identifying clients at the highest risk of leaving, surfacing events that will trigger lifestyle changes, and advice on how topics in the news can alter client behavior. Predictive analytics won’t sign new clients on its own, but it can play an invaluable role in advisors’ technology and help them to win their game, according to Henry Zelikovsky, founder and CEO of AI outsourcer, Softlab360.

Artificial intelligence and machine learning (AI/ML) are without a doubt some of the strongest tools any company has access to. Studies have shown that the most AI-mature companies achieve 50% higher revenue growth than their peers, but wealth management continues to lag behind with the lowest maturity of all industries surveyed. Use of customer behavior analytics in particular is becoming more widespread, and is estimated to be in use by 69% of marketing firms. Zelikovsky expects to see  the use of AI/ML expand across the industry, and continue to deliver innovative solutions as the technology gains wider acceptance. 

Most advisor technology applies a formulaic and linear approach to predicting client behavior, Zelikovsky noted. While this behavior may be quantifiable in a client’s trades, purchases and other choices, those actions have a root cause that is not mathematical but psychological. Focusing on the psychological basis can nudge predictive capabilities to the next level and enable more accurate predictions of everything from upcoming events in an individual investor’s lifestyle to rates of client churn, he said. By centering humans alongside numbers, a behavioral economics revolution can support financial planning. 

A History Book vs. A Crystal Ball

Regular analytics look at prior events and assume that patterns of the past will continue into the future. They chart along the X- and Y-axis to determine a future trendline, whereas predictive analytics provide the essential Z-axis, delving into the behavior of clients and the psychology behind that behavior to generate probabilities of how they might react in similar situations. One of the most powerful ways that this behavioral analysis can be harnessed is for predicting customer choices and identifying potential churn, Zelikovsky stated. 

Customer churn is a key metric for advisor firm profitability, with even small rates of churn quickly compounding over time. This is due to the heavy financial burden — it costs up to five times more to attract new customers than to keep the ones you already have. 

Even for those advisors who believe their clients are happy comes sobering news., Research from the University of Cambridge has found that the most widely used measures to determine client satisfaction fail to accurately predict customer sentiment and can even mask serious issues. Zelikovsky proposed that by studying unstructured data such as a client’s email, text, and conversational responses, AI/ML systems can determine client reaction patterns and assess which customers are most likely to leave. 

Well designed AI/ML technology breaks clients up into psychological denominations, or categories of psychological profiles centering on how people relate to money and the ways they want to use it. While many investors are saving for retirement, everyone is planning to use that money in different ways — buying a new house, supporting grandchildren, philanthropy or travel. Zelikovsky suggested that how clients intend to use their money impacts how they save it, and understanding the psychology behind those choices and the details of their lifestyle can help predict their behavior in the future and how they will react to different suggestions by their advisor. 

AI-powered analytics can also generate peer comparisons across groups of advisors by monitoring their behavioral patterns and offering adjustments to improve performance. The software can examine how an advisor is managing individual accounts and then predict the feedback that those households will most likely send back in response, as well as the feedback the advisor is most likely to receive from their firm based on their performance. 

“We are constantly learning and observing,” stated Zelikovsky, whose firm is developing monitoring tools to determine if clients are taking their advisor’s recommendations and quantifying the results. 

Another innovative use case for predictive analytics is determining the probability of negative life events occurring for a client such as diseases, cancer, Alzheimers or others. This data provides advisors the opportunity to offer additional insurance and other financial options.  While some advisors may not want to bring up these events with clients due to their uncomfortable nature, Zelikovsky explained, it is the responsibility of a financial planner to nudge their clients towards building a Plan B and even a Plan C. 

The Current State of the Market

Many CRMs offer basic estimates of client conversion and churn, but generally use simple mathematical formulas. These previous attempts at prediction materialized as regular analytics, Zelikovsy pointed out. “If I have my numbers, I can chart them. If I look at the history, I can chart my risk. But is that prediction, or is that just looking at the history and assuming that the way things have been is the way that things will be in the future?” 

Of course, predictive analytics is not perfect since the algorithms that produce the analytics rely on the availability of a large enough volume of high-quality data to ensure accurate predictions, according to digital analytics platform provider, Amplitude. Most predictive analytics tools generate a score based on characteristics including data quality and quantity. A score above 70% is considered to be a usable model.

Companies must also plot out customer interactions for users to trigger throughout their customer journey. These triggers can be touchpoints like clicks, signups, video views, or reaching certain milestones. This is the behavioral data the predictive analytics algorithms will crunch.

In a previous project  for one of the Big 4 custodians, Zelikovsky’s firm analyzed over 7 million anonymized trades to identify patterns in trading behavior. “Why do people do what they do– and when do they do it?” Zelikovsky asked. 

His team was then able to extract groups of customers with shared trading behaviors. The software looked at those groups and predicted how they would react to a new product line, or who would be the most likely to leave them for a competitor. Zelikovsky now plans to take this intensive research into client behavior a step further by looking beyond numbers and quantifiable data.

Reading Between the Lines

Most standard forms of pattern analysis rely solely on structured data  such as gains and losses or quantifiable strategies, which are more easily entered into equations. AI strategies work best when bold, Zelikovsky advised, such as combining traditionally used structured data with unstructured data such as conversations, media, opinions, and notes. The incorporation of unstructured data is vital to forming a complete predictive picture, he insisted. Think of structured data as the skeleton that gives the analysis body and clarity, with the unstructured data as the muscles that make it move and bring it to life. 

Zelikovsky refers to a client’s words as their ‘semantical behavior’.  AI/ML software can process and recognize patterns by interpreting their semantical behavior to translate feedback into a behavioral profile and then predict their intentions. Through natural language processing (NLP), AI software can pick up on clients’ emotions and their subtle reactions to an advisor’s suggestions. 

Taking the Wider World Into Account

It is important to take into account the various media that investors are consuming, such as CNN or the Wall Street Journal, and how they impact their decisions, Zelikovsky noted. These sources often provide information on clients’ feelings about their finances, and highlight the types of events they would be most likely to react to. A software system can read through popular news articles across a range of sources and group the topics into categories. 

A predictive analytics team can then collect client feedback and mix and match the client feedback received on a certain day to the topics in the news. By evaluating the interactions between media topics and client reactions, they can determine which articles were the most useful in predicting changes in client sentiment. 

Unlocking the Map to Your Future

By combining cutting-edge semantic analysis of unstructured data with more traditional projections from structured data and looking into the “why, how and why” behind the “what” of client behavior, expert AI/ML outsourcing vendors like Softlab360 are on the bleeding edge in   predicting customer behavior. 

For more information about how a partnership Softlab360 can expand your predictive potential and give you a competitive advantage, visit



The Wealth Tech Today blog is published by Craig Iskowitz, founder and CEO of Ezra Group, a boutique consulting firm that caters to banks, broker-dealers, RIA’s, asset managers and the leading vendors in the surrounding #fintech space. He can be reached at