Debt collection has existed for as long as consumers have been taking loans. For the past few decades, collectors have been building call center businesses - hundreds and thousands of calling agents, using automated dialers to contact indebted consumers, compensated with commission once they reach their collection goals. Consumers are often harassed by overzealous collectors looking to meet their goals, calling as much as 6 times per day. It’s a stressful environment focused on one thing - get the money or get out.
In the past few years, collectors have seen a shift in consumer behavior, and the old way of collecting is becoming less relevant and effective. Mobile and digital experiences have changed how we interact and communicate, and consumers are demanding the same from collections. Reply rates to letters have plummeted, as well as the number of consumers who pick up the phone. Some issuers report double-digit growth in the percent of consumers who are strictly digitally engaged. By 2018, according to Gartner, consumers will consume more than 50% of content on mobile devices, and expect communication by email and text. At the same time, more consumers are taking loans and end up in collections (several banks have reported a surge in default rates above 5% in the past two quarters), and communicating with all of them at scale, and while staying compliant is impossible. Burdened by thin margins and a legacy call center approach, collection agencies have failed to make the deep investment required to use modern, integrated technologies and adapt to consumer needs. This, along with ongoing changes in regulation, is why collectors’ success rates are decreasing while litigation numbers keep rising.
In contrast, machine learning and digital first systems are emerging to deliver great consumer experiences at scale and with better results. Developed with the most modern tools, these systems offer the best possible user experience, a personalized communication, and enable a strategic, data-driven approach to collections. Using their advantage in efficiency, scale, performance and compliance, machine learning based systems are delivering a better alternative to call center based collections. Whenever the human-intensive and the machine learning based approach square off, these modern tools provide a superior result.
Machine learning beats call centers in scale and efficiency
Call center agents are humans, and like the rest of us, they too lose focus if a task is too repetitive or they see a bigger reward elsewhere. That is why, when given a new set of accounts, they call those furiously over 30-45 days before reducing their call volume. Consumers who get caught in this early push are often driven to pay, sometimes by coercion - but quickly fall off the wagon once the calls stop. Collectors often get angry when facing a difficult debtor, distracted by the end of the day, or stressed when they fail to meet their goals. Finally, when collectors see that they can’t get consumers on the phone, they move to the next fresh batch of accounts.
The debt collection process is broken and neither side ultimately is set up for success. Consumers in debt need a steady, personalized, and consistent communication stream to put the right offer in front of them, and then nurture them through the payment process. They have limited attention spans, they work irregular hours, are often tired and distracted. They need a helping hand and a service that’s available when they are. They may not be paid the same amount in regular intervals, and need a payment option that accounts for that. Because consumers can engage when it’s convenient for them, a lot of engagement can happen when a call center would not legally be permitted to contact consumers.
Machine learning based systems deliver great consumer experiences tailored to individual needs. A machine doesn’t get emotional, instead it uses data from hundreds of millions of previous interactions to tailor its strategy to every individual. When a consumer needs a highly personalized payment plan, these sophisticated systems detect that and offer the customization the consumer needs, subject to approval by their creditor. Then, they nurture the consumer through the process with personalized communication, through whichever channel they prefer, and follow up flows that get more consumers than ever beyond the finish lines of their payment plans.
Machine learning scales much faster than call centers
Call centers are a people driven business model, thus they are susceptible to negative emotions, lack of consistency in how they communicate with each client, and challenges to scale. One of the biggest challenges in call center based collections is staffing. Collectors are paid low base rates and high commissions, and are expected to deliver compliant collection calls while getting consumers to pay, and making hundreds and thousands of calls per day. This high stress environment yields high churn, so call centers are constantly hiring, rehiring, retraining and firing collectors. Churn rates of 40% are considered good, and 100% churn rate in 12 months is not uncommon. Call centers often struggle with not having enough collectors or being overstaffed, leading to low performance and profit margins.
On the other hand, machine learning based systems are the most flexible constructs. While often supported by a small team of customer care agents, the bulk of the system’s operation is controlled and delivered by machines. Firing up a server or turning it off in response to changes in required capacity is as easy as a flip of a switch. A machine learning platform could handle thousands of accounts per customer care agents, compared with a call center’s 750 per agent, delivering unprecedented scale that a legacy collection agency simply cannot achieve, and one that is much easier to control for compliance purposes. Instead of training new hires, the operation manager spins up another server that has the same, code controlled, compliance policy and the same pre-written content. As a result, consumers get the same personalized treatment governed by the same code-driven compliance policy, no matter which server actually runs the computations that delivered their collection communication.
Machine learning performs better than call centers
Based on its scale, efficiency and uniformity, machine learning based systems prevail when tested head to head against call center based agencies. Drawing from its vast amount of data, a machine learning based system personalizes the contact strategy with the consumer (timing, channel, frequency, and tone) as well as the offer strategy (whether to offer payment-in-full, percent of settlement, or what length of payment plan - all with the creditor’s approval). Consumers end up paying more when they are sent personalized communication - and the system can then track their responses, in real-time, to feed its scalable decision engine and decide what its next action should be. This way, it can trigger a phone call to a non-responsive consumer, while offering a longer payment plan to someone that reviewed a 3 month plan but wouldn’t sign up, all the while modifying the latest payment for a consumer who can’t make it on time.
This personalized, behavior based treatment strategy is what makes a huge difference between machine learning based systems and call center based ones. Consumers don’t feel coerced into making a payment and do not trigger a charge back. Payment plan breakage is phenomenally lower than traditional agencies’ which leads to superior liquidation, just weeks into each placement. In addition, this improved performance is achieved while significantly reducing contact attempt frequency, contacting consumers an average of 3 times per week compared to 6-8 times per day with traditional agencies, which in turn, means a significant reduction in complaint rates.
Bottom line - the future is here
Call centers have reigned in the world of collections since their model was the only way to collect profitably. That is changing. Mobile devices, consumer preferences, creditor focus on user experience and repeat business, regulatory pressures and above all - the maturation of machine learning and marketing technologies - are enabling a new paradigm in debt collection. Machine learning based systems are growing more and more prevalent, competing head to head with traditional agencies and gaining ground with the most sophisticated creditors in financial services. It is a change that garners great benefits for everyone involved - creditors, debtors, and service providers.