Machine Learning Based Liquidation Looks Different – and That’s a Good Thing

By on June 26th, 2017 in Industry Insights, Machine Learning, Product and Technology
TrueAccord Blog

TrueAccord serves major issuers, debt buyers and lenders across the US. We compete with traditional collection agencies and beat them: TrueAccord collects more than 1.5 times the competition in a typical 90 day placement period. We use a machine learning based system, HeartBeat, that replaces the traditional call-heavy model with digital first communications that compliment consumer behavior and mode of communication, but that makes our liquidation curves look different from traditional agencies.

Traditional liquidation curves

Traditional liquidation curves typically shoot up in the first 30-45 days, followed by a plateau around 60-80 day, with a possible bump towards the end of the placement window. This pattern is driven by several factors.

Routine: agents receive fresh accounts and are eager to call them. They fire up dialers and quickly reach consumers who can either pay or be lightly pressured to pay. After a few weeks of calls, agents are tired of calling the same consumers. They heard what they think are excuses, have driven all the easy payments they could drive, and are ready for new accounts. “Old” accounts, as old as 30 days, get a worse treatment. Collection managers know this, and try to trick collectors into thinking they got fresh accounts by pulling the accounts out from the system and re-entering them. This rarely works. Collectors lose focus and with it, performance.

Net present value: settlements are better than payment plans for collectors – they mean more money now, versus a payment plan that may fail, and require reminders and additional work by the collector. Collectors opt for more settlements earlier, if they can get the consumer on the line. Under the pressure of a call, the consumer may commit to a payment plan. In this case the collector prefers as high a monthly payment as possible, since they assume the payment plan will fail early. The consumers, struggling with irregular cash flow and a large payment they shouldn’t have committed to, fail payment plans at a staggering rate: as much as 50% of payment plans fail.

Remorse: consumers who agree to settlements or plans often feel remorse after getting off the call, and tend to charge back on payments they made. Chargeback rates in the debt collection industry are so high (rates as high as 2% are not rare), that most payment providers won’t work with collection companies.

The initial bump in liquidation is often enough to beat other phone based agencies. Since all agencies use the same methods, a slight advantage in selecting the right accounts to call first can get an agency ahead of its unsophisticated peers.

The TrueAccord liquidation curve

In contrast to the traditional curve, TrueAccord’s liquidation curve is somewhat linear. It often starts lower than the traditional agency, but continues to rise through the placement period until it crosses and exceeds its competitors. That inflection point can happen as late as day 80 (before the algorithms have been tuned, early in a pilot) and as early as day 15 (once the algorithms have learned how to handle a new product). The difference is driven by several factors.

Data driven treatment at scale: TrueAccord’s system is machine learning based and digital first. Since it starts with an email, it can initiate contact with all consumers easily, without having to call them often – and consumers are much more likely to respond to digital communications than to a phone call: while Right Party Contact rates often hover around 4-5%, email open rates on TrueAccord’s platform reach 65-70% and click through rates reach 30-35%. Once it sends its first email, it uses real time tracking of consumer responses to tailor its next steps. The system relies on hundreds of millions of historic contact attempts to optimize its contact strategy. If the consumer doesn’t reply, the system can automatically switch between channels (from email to text, call, letter, and so on) to reach the consumer. It also uses data to figure out what time of the day to contact the consumer that will yield the best response rates, call centers are limited to just making phone calls, which often consumers ignore because they are busy or just don’t pick up calls from unknown numbers. Since a machine doesn’t get bored, it continues contact attempts (3 a week on average) until it is told to stop. Targeted, consistent communications at scale mean that more consumers will interact with our system compared to a call center.

Optimizing for liquidation: a data driven system can use historical data to understand what best fits consumer needs and leads to better liquidation. It doesn’t need to push for early settlements because its automation lets it serve each consumer according to their needs – making custom tailored plans viable. Consumers get easier payment terms that fit their needs, and end up paying more. We convinced several of our clients to move from a default payment plan length of 6 months to 12 months. Contrary to call center based intuition, these longer plans get more consumers to sign up and don’t cannibalize settlements, in turn leading to an increase in liquidation. The machine learning system can service these plans at scale and reduce failure rates: TrueAccord payment plans complete as much as 89% of the time (as low as 11% breakage). By the time payment plans for traditional collectors fail their second payment, TrueAccord’s liquidation rates start soaring.

Best in class user experience: consumers don’t like phone calls or letters. They prefer 24/7, personalized, easy to use services – and collections aren’t any different. Using our system they can customize and sign up for settlements, payment plans, or ask for debt verification. Having access to their account information, and a sense of control over payment options, consumers don’t feel pressured or remorseful after paying. TrueAccord’s chargeback rates are next to not existent.

Bottom line

Machine learning based debt collection is different in many ways that benefit creditors and consumers. Our liquidation curve tells the story of how our system behaves differently than call center based collections – serving consumers at scale, using their preferred communication channel, and while tailoring payment solutions that work for them.

Hear our CEO talk about AI in Fintech at LendIt

By on February 28th, 2017 in Industry Insights, Machine Learning
TrueAccord Blog

Our CEO , Ohad Samet, will be part of a panel discussing Artificial Intelligence Uses in Fintech. The panel will be held at 2:15pm Eastern on Tuesday, 3/7.

Continue reading “Hear our CEO talk about AI in Fintech at LendIt”

Why batch-sending emails aren’t all there is for debt collection

By on December 13th, 2016 in Industry Insights, Machine Learning
TrueAccord Blog

Though historically resistant to innovation, the collection industry feels pressured to make changes. Consumer preference, requirements from clients and mounting costs dictate increased use of technology – a welcome trend. Among those new tools, we are starting to see increasing adoption of emails for collections. Agencies have a small selection of vendors to blast out an email. Agencies with large call centers view this as a cost reduction exercise, and another way to get consumers to call in and talk to their agents.

Continue reading “Why batch-sending emails aren’t all there is for debt collection”

Self service portal vs. machine learning-based collections

By on December 6th, 2016 in Industry Insights, Machine Learning
TrueAccord Blog

Consumer behavior is changing. As more of us are glued to our mobile phones, emails, and social media accounts, it’s clear that the old ways of collecting debt are quickly becoming irrelevant. Still, the market doesn’t offer a multitude of collection solutions aimed at responding to the digital consumer. When we present our machine learning-based solution to prospective customers, we’re often asked about the difference between our solution and a self service portal. Although both solutions are digital, they cannot be less alike.

Continue reading “Self service portal vs. machine learning-based collections”

Augmenting your strategy with automation: part three of three

By on August 2nd, 2016 in Industry Insights, Machine Learning
Augmenting your strategy with automation: part two of three

Automation and digitization offer new tools for the collection strategist, augmenting the traditional building blocks. These new tools, introducing flexibility and sophistication that are usually attributed to other parts of the business, can mitigate common pitfalls.

Continue reading “Augmenting your strategy with automation: part three of three”

Augmenting your strategy with automation: part two of three

By on July 19th, 2016 in Industry Insights, Machine Learning
Augmenting your strategy with automation: part two of three

Automation and digitization offer new tools for your collection strategy, augmenting the traditional building blocks. These new tools, introducing flexibility and sophistication that are usually attributed to other parts of the business, can mitigate common pitfalls.

In this series, adapted from our free eBook Automating Debt Collection 101, we’ll review the three major areas where automation and digitization can boost a collection strategy:

  • Early contacts and improved segmentation
  • Persistent communication
  • Improved customer satisfaction

In this second part, we’ll focus on improving performance with persistent communication.

Customers in debt are in a dire situation, cannot pay the balance in full, and many times even a payment plan isn’t feasible. A call center is limited in its flexibility – beyond a certain number of payments or customizations, a human agent is just too expensive. These accounts risk being mishandled, and end up paying less than they could with some “hand holding”.

Automated collections have a tremendous advantage in handling complex cases. The platform consistently follows up with customers using multiple channels, offering various solutions according to an optimized offer strategy, and administers changes in those solutions (split payments, rescheduling and more) over time as needs change. These tools can accept and administer a monthly $5 payment that increases over time, even if the customer misses a few payments and needs consistent follow-ups. When the vast majority of contacts are automated, even small amounts are profitable – and add up. The system doesn’t get tired, doesn’t get angry, and doesn’t need to go home by the end of the day. It’s there to service the customer.

TrueAccord sees more than 35% of customers in an average placement click on a link and negotiate with an automated system, thanks to diligent and relevant follow ups. In tests, working on the long tail of underserved accounts yields 4-8% of additional recovery – dollars that would otherwise be considered lost.

Augmenting your debt collection strategy with automation: part one

By on July 6th, 2016 in Industry Insights, Machine Learning
Augmenting your strategy with automation: part one of three

Automation and digitization offer new tools for the collection strategist, augmenting the traditional building blocks for your debt collection strategy. These new tools, introducing flexibility and sophistication that are usually attributed to other parts of the business, can mitigate common pitfalls.

In this series, adapted from our free eBook Automating Debt Collection 101, we’ll review the three major areas where automation and digitization can boost a collection strategy:

  • Early contacts and improved segmentation
  • Persistent communication
  • Improved customer satisfaction

In this first part, we’ll focus on using automation to facilitate early contacts and improved segmentation.

Automated collections are scalable. This means communicating with all customers as early as possible in the collection cycle, quickly working to resolution with those who can pay, and a more robust debt collection strategy. In traditional call-center collections, up to 50% of meaningful interactions are made within the first 30 days of communication. With an automated strategy, most of that value can be captured in a much more cost-effective manner, in a much shorter time span. No more guessing who to call first because everyone can be contacted at scale.

Further, automated and digital collections create a wealth of data that cannot be gleaned form calls. User clicks and browsing, time and day of activity and more. The data can be used to segment accounts to those who are engaged, those who’ll respond better to phones, and those who should be sold or handled in other ways. It allows much more flexible recall criteria than placing for a set number of months, no matter what happens with the account. This means giving accounts the treatment they need at the right time, improving liquidation as well as cost to collect thanks to the scale of operations.

Want to use our tools to optimize your strategy? Visit our website to learn more.

The three technology keys to automating debt collection

By on June 14th, 2016 in Industry Insights, Machine Learning
The three technology keys to automating debt collection

You may have already downloaded our free eBook, Automating Debt Collection 101. This is an excerpt.

Flipping the traditional butts-on-seats model on its head and teaching a machine how to do a human’s job is not an easy process. We’re talking about domain-expert based automation. This is a grueling, operational process of understanding why some people pay and others won’t, and translating it into algorithms that grow with the data they accumulate. To realize the benefits of automation, you’ll need to pay attention to three elements:

Data Infrastructure

The key in this process is defining our key performance indicators. One can’t start this task if data are unavailable, corrupt or fragmented. Most collection teams use a tapestry of systems – for scrubbing bankruptcies, for calling, a mail processing system, a payment processing interface and so on. That leads to a fragmented data store, which makes it impossible to know which actions were taken on a debt and attribute success to any of them. You can’t improve what you can’t measure.

Your first step is creating a unified data store for all your data.

Feedback Loop

Extracting knowledge from domain experts can be frustrating. Often they decide intuitively and cannot explain their reasoning. It takes training, ongoing conversation, and an iterative process to structure their knowledge. The feedback loop includes three steps:

  1. Interviewing your experts: presenting several cases that were successfully converted and those that weren’t, and asking what they have in common.
  2. Implementation: the resulting model is validated against data trends.
  3. Deployment: the model is deployed to your system, and agents can comment on its performance in real time and compare it to the way they would act under similar conditions.

Creating a feedback loop between your agents and data scientists is incredibly important. Without it, your data scientists are guessing, and your agents work without guidance, their knowledge untapped.

Increasing Relevance

The human brain is an incredible machine, and it offers intuitive connections that computers can’t make. Whenever faced with new information, even the slightest addition, the brain recalculates its route and makes new assumption about the person they are interacting with. A machine can’t replicate the brain’s ability but it can mimic it – with some help.

Use your experts’ understanding of a customer’s response to inform the way you send your initial communication, as well as using responses you get from them to inform your next communication. While deploying follow up flows based on browsing patterns, we realized some flows converted up to 7 times better than a regular message.

Find pockets of customers who don’t get personalized treatments and create those responses.

Bottom Line

Consumers are increasingly reliant on credit to fund their consumption – whether short or long term. This leads to defaults, and to debt collection being a part of any business’ tool box. As you grow, using automation or an automated solution like TrueAccord is the right way to minimize your costs while increasing your performance, scalability and customer satisfaction.

Interested to learn more? Pick up our free eBook: Automating Debt Collection 101

Two approaches (and a third) to automating the debt collection process

By on June 7th, 2016 in Industry Insights, Machine Learning
Two approaches (and a third) to automating the debt collection process

Debt collection and account receivable departments often start with one person contacting late customers and evolve from there. Even third party collection agencies grow this way as they get more business. As a result, most collection departments are comprised of large teams of operators trying to negotiate with customers. Data science teams that are tasked with improving performance and profitability usually approach the task in one of two ways: process automation or agent-independent decision automation.

Process Automation is the effort to automate manual tasks done by collection agents, replacing them with an automated process or a self-service portal. This may mean skip tracing, logging payments, or queuing up phone numbers to call. The data science team acquires data sources or builds a process that replaces manual work with automated one, reducing the amount of time an agent spends per case. It’s about optimizing agent time on the phone, making sure that every action an agent takes is a high yield one, while busy work is replaced by some level of automation.

Decision automation means trying to teach a machine how to make the same quality of decision an agent makes in the collection process. For example: how to talk to debtors, what to tell them, how to respond to their issues. Because most agents have a hard time explaining in detail why they made one decision and not the other (they “just know”), often data science teams treat agents as an unreliable source of information. The team determines what they are trying to optimize – for example, right-party contact or the number of calls ending with a payment. They then build models that optimize these metrics, but without asking agents for feedback – only looking at long-term liquidation results.

While both approaches are important and are often used at TrueAccord as well, there’s a third one that often gets overlooked because data scientists and agents don’t interact often: Agent Dependent Decision Automation, or Expert Based Automation.

Interested to learn more? Pick up our free eBook: Automating Debt Collection 101

How can computers collect better than humans?

By on May 3rd, 2016 in Industry Insights, Machine Learning
TrueAccord Blog

When we started working on our patented collection engine, Heartbeat, the industry told us: you’ll fail. Computers can’t collect. Humans do. The best you can do with automated communications is to drive inbound calls, so human collectors can “seal the deal”. Fast forward 18 months since our launch, and Heartbeat beats call-center based agencies in a growing number of segments.  It turns out that computers collect debt pretty well. How come?

Debt collection is a numbers’ game. Consumers are ready and able to pay at different times, react to different stimuli, and need varying levels of support in the process. Teaching a machine to respond to these needs was historically more expensive than hiring humans, but as technology improves and compliance requirements grow, this is changing rapidly.

Humans are great at acting on intuition and responding to a changing situation. We act well based on partial information, guesses, slight changes in tone of voice and intonation. Good sales people do so without thinking. Humans are great at identifying and understanding corner cases and responding to complex inquiries. Machines can’t learn these things unless explicitly taught, and many of these skills are nuanced and complicated. Machines are “robotic”, for better and worse, and can’t have empathy.

Humans do have downsides, too. We are susceptible to biases. We make decisions based on the few past examples we remember and ones that fit what we believe. Collectors fixate on high balance accounts, worry about missing their goals, fight with their significant other and lose focus. Machines do not. Machines don’t forget a thing, and they always take as much data as available into consideration. Machines don’t talk back or get angry.

Historical attempts failed because they either tried to replace humans with even lower-paid humans, or tried to automate and get rid of humans altogether. We realized that a hybrid approach was the best one: machines make accurate decisions based on historical data when available, and learn from humans when not. Humans understand corner cases. We had to create a combination of a strong engine, and a team of experts to continuously improve it.

How does that work? When Hearbeat doesn’t “know’ what to do with a customer, it defers to our team of experts in San Francisco. They resolve the issue for the customer, and also give enough input so Heartbeat will know how to deal with the same situation in the future. The combination allows us to hit incredible productivity rates, while beating other “robotic” and passive “payment gateway” solutions.

Can machines collect? They can, and apparently many who are in debt prefer their targeted approach. When you think about the user experience, the ease of use and the automation, it’s actually not that surprising.