Home > Columns > Executive Interviews

Anthrolytics Executive Interview

Click image below to view the ebook

Sheri Greenhaus, Managing Partner of CrmXchange, discussed Anthrolytics’ SaaS solution with Jonathan Hawkins, co-founder, Anthrolytics Ltd. Anthrolytics combines behavioral and data science to predict what a customer will do next, and why.

Sheri Greenhaus: Please give us an overview of your solution.

Jonathan Hawkins: The genesis of our solution stemmed from a major bank's inquiry into the likelihood of their customers leaving if they did something unpopular, such as increase prices. This is a question many organizations have pondered at some point. The answer however, partly depends on the customers' feelings towards the brand. If customers love the company, a minor misstep will probably be forgiven. On the other hand, if they already dislike the brand, it could be the final straw prompting them to leave.

So, the real question became: How can we gauge our customers' feelings towards our brand? Traditionally, people have relied on customer surveys, speech analytics, and other similar methods to try and gauge the mood of customers. However, these methods have limitations. They often rely on direct interactions, and for a major bank with millions of customers, only a small percentage responds or contacts the customer service center annually—usually less than 5%. Moreover, these methods offer lagging indicators, meaning they might capture a customer's positive experiences from over a year ago, despite recent negative experiences.

Employee surveys in contact centers face similar issues. People tend to be honest about their feelings only when they are about to leave. My co-founder developed algorithms that focus on "moments that matter," understanding that behaviors result from the cumulative impact of many experiences over time. It's usually not a single big event but a series of small incidents that affect loyalty or lead to resignations, unplanned absences, or customer churn.

To address this, we use existing operational data, particularly in contact centers. We create emotional profiles for agents daily, determining their moods and categorizing them into empagraphic segments (cohorts of employees that have similar sets of feelings), which are then linked to behaviors. Some agents may be at risk of resigning in the next 90 days (with varying degrees of probability), while others might be likely to to take unplanned absences or see a fall in performance. Others might be highly motivated and engaged. We are currently focusing on contact centers due to their high attrition rates. For instance, a center with 60% attrition among 10,000 agents or 40% out of 50,000 agents – staff attrition and absence in the U.S. costs companies $100 millions each year.

In reality, almost every agent will eventually leave; the question is when. Our solution empowers team leaders to understand their agents' daily emotions, who's at risk of unplanned absences or resignation and why. The goal isn't to eliminate attrition but to keep those predicted to resign for a few more months. The ROI is remarkable. In a 1,000-seat contact center with 20% attrition, it typically costs $6 million annually in management time, lost productivity, recruitment, training, and onboarding. By retaining employees for just two more months, companies can save around $1 million annually, contributing directly to the bottom line.

When people consider contact centers costs, they often overlook the hidden costs, like recruitment and training expenses. We can help mitigate part of that issue. The harder-to-quantify aspect is employee wellness. Research shows that mental health issues, primarily stress and burnout, account for 62% of long-term sickness. While we're not offering a diagnostic or therapeutic model, if we can observe an employee's mood declining, we want to intervene for their well-being. A recent study reported that 87% of employees appreciated being treating empathetically and that in turn drives improved loyalty and productivity.

Today, people struggle because, until we introduced Predictive Behavioral Analytics, they had no way to understand how employees truly felt at scale. Empathy requires this understanding, and remote and hybrid work environments have made it even more challenging. The days of walking around a call center looking for actionable insights are gone. We utilize data companies already possess, such as WFM and ACD data, and leverage it in a unique to predict resignations, unplanned absences, and motivation levels.

Sheri Greenhaus: Most people don't grow up aspiring to be contact center agents. What we have found through our webinars, is that those who are agents often look for a path to advance within the organization. Providing a path for agents within the organization may also help to retain them. Are there particular industries that you work with?

Jonathan Hawkins: We work with any organization that operates a contact center, typically those with multiple hundred agents or more. So, it encompasses a wide range of organizations, industries, and territories.

We find that many BPOs need to demonstrate to their clients that they are effectively managing their programs. Employee wellness may be a genuine concern for some, while others see it as a necessary requirement. We enjoy working with companies that recognize the significance of employee well-being, because that is where we can have the greatest impact.

Sheri Greenhaus: You mentioned core factors that people care about. Could you categorize the top 5 most important aspects overall?

Jonathan Hawkins: Certainly, while there may be variations depending on the company and industry, broadly speaking, you can group them into three buckets: 1) shift time, 2) type of work, and 3) time off.

Sheri Greenhaus: Are there specific requirements, such as speech analytics, needed to use your solution?

Jonathan Hawkins: There are no specific requirements for speech analytics. Our focus is on predicting likely behavior based on individuals' experiences with the company. While some outliers, like personal issues at home, may exist, our goal is to provide valuable data within the systems that organizations already use.

Unlike other software companies that create proprietary dashboards and struggle to get users to adopt them, we take a different approach. Our belief is that users want valuable data integrated into their existing systems. For example, we integrate with performance management systems, providing an employee wellness section alongside traditional contact center performance metrics. Team leaders can now have a holistic view within their existing engagement systems and identify employees at risk of unplanned absences or other issues. This approach enables more effective team management, especially in remote work scenarios where identifying potential problems becomes crucial.

As an operational level, you can consider the center as a whole. For example, if 15% of individuals appear to be at risk of taking unplanned absences, how does this affect workforce management and productivity? Now, consider if it's 10% or 20% of people who seem likely to resign within the next 90 days. How does this impact our recruiting efforts? One particularly valuable aspect is evaluating the overall motivation and wellness score of the center as a whole, and I'm currently reporting this as a single number within the range of -100 to +100.

Do we need to implement management changes? If I'm looking at this from a historical perspective, and I have one team, but everyone's performance seems to be declining, is it the manager or the work environment? So, now we can start examining factors like work type, call type, shift type, and manager type for various users. The key point we were discussing is integrating all of this into existing platforms.

Sheri: Let me ask you this: if I've just purchased Verint performance management and I'm using your solution, will it appear that I'm fully utilizing Verint's solution?

Jonathan: Yes, the way we're currently integrated, we are one of their marketplace partners. We have the capability to effectively integrate with any system due to our flexible architecture. We can take data in flat file format, process it within our platform, and return it in the desired format for their use. There's no one else offering a solution like ours – simple to use and understand, powerful in its impact.

Sheri Greenhaus: People will inevitably leave regardless of our efforts; the goal is to extend their stay. With your solution, when you identify a group likely to leave within the next two months, does it enable better recruiting planning?  

Jonathan Hawkins: Absolutely. If we can predict resignations in the next 30, 60, or 90 days, and let's say our training program takes an average of 8 weeks, we can plan ahead. For instance, if we anticipate 200 resignations within 90 days, factoring in 60 days for training and 30 days for recruitment, we can begin recruiting immediately. By intervening with team leaders, we might retain some employees for an additional month or two. This way, we can adjust our recruiting strategy accordingly, and it aids in our forward-looking planning.

Traditional employee surveys can't provide this level of insight; they offer lagging indicators, not leading ones. My focus with AI is on providing organizations with leading indicators for effective workforce management and employee well-being while maintaining operational efficiency. Without effective management, we won't have the resources to sustain our business in the long run.

The reality for most organizations is that they will inevitably have to do some things that upset employees from time to time, such as asking them to work over a public holiday, or take a late night shift. What we give those employers is the ability to see which employees are likely to take that badly so that they can take proactive steps to secure the employees willing cooperation.

Sheri Greenhaus: Could it also involve identifying the 50 out of those 200 employees that the company wishes to retain? Then, the company could determine their specific needs, such as career progression or flexible schedules, and focus on retaining them.

Jonathan Hawkins: I fully agree. We had a similar situation a few months ago with a BPO company that lost a contract and needed to reduce its workforce. They viewed it as natural attrition to save costs, avoiding layoffs. So, absolutely, we can help with that. Identifying individuals for fast-tracking and investing in their development is crucial. If we understand what motivates them, we can allocate resources accordingly, nurturing the next generation of leaders who have experience on the front lines.

Moreover, understanding the factors driving positive behavior, such as work type, shift type, and their combinations, can be integrated into the recruitment process. If we know that certain roles require specific traits, we can begin screening for those early on. However, it's important to note that we don't have the capacity to diagnose mental health issues, and we don't handle the recruitment side of things.

But perhaps the biggest difference in our approach is that it considers each employee as an individual – all of whom have had different experiences of work, even if they do share some common characteristics. This is fundamental to delivering an empathetic experience – understanding the wants and needs of each employee.

Sheri Greenhaus: What about on the customer side? We've been talking about the employee, right? What is the ability to predict what customers are going to do?

Jonathan Hawkins: The system was originally built for customer attrition. The initial use case we had was for one of the top three banks in the U.S. with tens of millions of customers, and these are the people who asked us, "If we do something our customers don't like, can you do a better job of predicting how likely are they to leave?" That was the genesis of the idea on the customer side.

We said to the bank, "Okay, here's a cohort of people. Forget about your demographics. Forget about your typographic segmentation. Here's a group of people who have a high propensity to leave because they just don’t like you. And here's a practical idea: rather than trying to sell to everybody in the same demographic (you and I may be in the same financial demographic, but I may hate the brand, and you may love them), why not target people like Sheri and sell to them because your conversion rates will be much higher?" What they saw by doing that very simple thing was an increasing conversion rate of 18%. But also, interestingly, a reduction in marketing costs of 38% because they weren't selling to people who didn't want to buy.

Sheri Greenhaus: Is this for new or existing customers?

Jonathan Hawkins: It was actually cross-selling to existing customers. Our two areas of focus are the contact center and existing customers.

If you think about this concept of behavior being the result of a cumulative impact of experiences, what that means is you can now start tracking people on a graph effectively. So, as an existing customer, I've had this experience, then that one, and the other one, and the next. And I start tracking the impact of all of those experiences. What's interesting is if you now start segmenting those customers before they hit the threshold where they're likely to churn, you can put loyalty campaigns in place before a customer even thinks of leaving. You can track this path and build a cohort of people who are potentially at risk before they even realize it themselves. That case study with the bank was fascinating. They also saw a reduction in churn in the segment we were looking at, over 11%, because they were able to intervene very early. Everything we do allows people to proactively intervene to change behavior and make a difference whilst it is still relatively cheap and easy to do so.

In another case, we worked with a relatively small telco, with around 4 million customers. We conducted a proof of concept with them, implementing A/B testing. They operated in an emerging market characterized by a pay-as-you-go environment where people often had multiple SIM cards and bought data from various providers based on their preferences.

They ran consistent outbound campaigns to upsell their services. Initially, their approach was to target customers based on classic demographic segmentation - who had the money to buy, rather than who was most likely to buy. We provided predictions that pinpointed a cohort of individuals we believed were not only most likely to purchase but also able to spend the highest amounts.

What's particularly intriguing is that their average revenue per user had been stagnant, fluctuating slightly above and below the baseline. On a scale where 0 was the baseline, they occasionally achieved a modest +8% increase, but often dipped below it.

Once we implemented the campaign, focusing on the targeted group within the A/B test, they experienced an impressive 89% increase. There were two main factors behind this success: Firstly, they were now selling to people who genuinely wanted to buy from them because the customers liked them – a seemingly simple but powerful concept. Secondly, these individuals were economically active and willing to spend more.

Additionally, beyond the financial gains, the most valuable insight they uncovered was that they had been investing substantial amounts in improving their network infrastructure. This investment was based on their belief that network quality was the top priority for their customers, as indicated by customer surveys.

However, during our initial calibration study at the project's onset, we discovered that this specific cohort of customers didn't prioritize network quality. Surprisingly, their primary concern was their ability to purchase data for friends and family. The company had implemented a data usage limitation within their app, and by removing this restriction, they witnessed a significant increase in spending. This success was the result of selling to the right audience and continuously adapting to their needs.

Furthermore, it highlighted that much of their previous survey data might have been biased, as respondents tended to be individuals who were dissatisfied with their services. People often ask us if we utilize social media data, but we generally avoid it due to its extreme polarization. Social media feedback tends to come from those who either absolutely love or loathe a brand, making it less valuable for gaining nuanced insights.

We've got a proof of concept running at the moment with some financial institutions, which I find really interesting. We're predicting likely customer behavior, which is being integrated into another platform, for customer journey orchestration. So, as you hit the customer journey orchestration, instead of pushing everyone down a predetermined path, like, "This is the best way for people to buy," you can now say, "Okay, well, Jonathan is a high-value customer from a customer lifetime value perspective, and Jonathan's really unhappy" — that's the prediction. So, actually, what I want to do is, rather than push Jonathan down an automated route, I want to connect Jonathan straight to live agents in the contact center.

I think as these things evolve, when we start having intelligent routing in contact centers, you can start thinking about this type of technology as a front-end. So, when somebody hits the contact center, how do I route them most appropriately? If we think Jonathan's happy and going to buy, well, why don't I just push Jonathan through self-service? If Jonathan might have an issue with loyalty, well, actually, I probably want him to speak to an experience employee with high EQ. So, you can start to get very intelligent to maximize customer service through hyper-personalization, to a degree. Its hyper-personalization based on how I feel about you and how I treat you empathetically. Even with a chat, if I know how somebody feels when they interact with the chatbot, with the advent of Generative AI, I can dynamically change the content to be empathetic.

Sheri Greenhaus: Is there anything that we didn't touch on that you would like our audience to know?

Jonathan Hawkins: I think the key things for me really are empathy is a message. You've got to understand how people feel, but it's such a significant untapped resource. Whether you're thinking about the customer side or the employee side, the ROI is untapped and spectacular. That's probably the biggest message — that by doing this, even if you're only improving by 5%, you're still going to make money on it. It's a no-brainer.