Back to Blog

A/B Testing & Machine Learning in Digital Debt Collection

A/B Testing & Machine Learning in Digital Debt Collection

All companies—no matter their industry, size, or location—have one ultimate goal in mind at all times: improvement. It’s only natural for companies to want to get better at what they do. Perhaps you want to improve your productivity, the quality of your products, or your customer satisfaction.

These are all valid goals. So how can you make these desired outcomes a reality? How can you systematically improve what you do by relying upon cold, hard, objective data instead of subjective hypotheses? By conducting ongoing, machine-learning-driven A/B tests.

We’re currently in the Golden Age of A/B testing. Tests can be devised, automatically sent to huge numbers of consumers in a matter of seconds, and the results analysed instantaneously. With machine learning, drawing actionable insights from the wealth of corresponding data is as easy as 1, 2, 3.

In this piece, we’re going to run through the basics of A/B testing, examine why it’s gaining popularity, and look at A/B testing in debt collection.

The concept of A/B testing

The core principles of A/B testing have likely been around for millennia, though it first came to prominence back in the 1920s. Since then, A/B tests have been used across a wide range of industries: medicine, agriculture, marketing, and more. A/B testing is a fairly straightforward concept—you essentially test out two versions of something (Version A and Version B) to see which performs better.

Anything can be A/B tested: lactose-free versus dairy-filled ice cream, two-wheel-drive versus four-wheel-drive cars, or landing page copy A versus landing page copy B. The first step is to define precisely what it is that you want to test.

Let’s imagine your landing page copy has been created to drive repayments. In this case, you’d obviously want to test out the repayment rate for each version of the copy. Each visitor would be randomly assigned Version A or Version B and after a predetermined amount of time (or amount of visitors), you’d see which copy was the most successful. That’s all there is to it.

Machine learning and A/B testing

A/B testing helps companies improve what they do and how they do it. However, the emergence of AI and machine learning in recent years has reinvented A/B testing itself: making it easier, more efficient, less expensive, and ultimately more effective.

Once manually devised, tests can automatically be sent out to large swathes of customers and the results immediately analysed in real time. You’ll gain up-to-date metrics regarding the performance of both Version A and Version B. Say goodbye to manually crunching through large datasets—with machine learning, you’re automatically presented with the results on an ongoing basis.

A/B testing should be an ongoing, iterative process within your organisation. You need to constantly test out new strategies to see which work and which miss the mark. As consumer preferences change, and new technologies appear on the scene, what worked in the past may no longer have the same effect.

Machine learning provides constant data-driven insights at the click of a button. Internal decision-makers have easy access to quantitative data, helping them create the most optimal strategies at all times. Test results are immediately synthesised and must-know information is presented in a clear, easily-digestible format. In other words, machine learning and automation do all the heavy lifting on your behalf. All you need to do is sit back, watch the results come in, and tweak your strategy accordingly.

A/B testing in collections

So how can A/B testing be used to improve your collections strategy?

Let’s first think about one industry where A/B testing has proven to be so successful: marketing. Marketers love A/B testing, and rightly so. They might analyse the size of the subscribe button on their website, their new logo’s design, or promotional outreach emails (to see which result in higher click-through rate).

Marketers spend so much time A/B testing so that they can drive consumers towards their desired outcomes—whether this is purchasing a product, signing up for a newsletter, or registering their interest for an event.

When you think about it, collections is no different—you need to drive debtors towards a desired outcome. In this case, it’s to get as many of them to repay as quickly as possible. Debt collection used to be fairly simple. You might send a letter or even get a collector to visit the debtor in person. Nowadays, however, the options are endless. For example, you can choose to automatically send a series of repayment reminders across a variety of channels: SMS, email, push notifications, and more.

But which channel will be most effective, to which consumers, and when? You shouldn’t simply try a bit of everything in the hope that something eventually works. In order to make your collections efforts as successful as possible for each individual customer, you’ll need to conduct a series of A/B tests.

Have you ever wondered: Which channel works best? Which copy results in a higher repayment rate How soon should we get in touch with customers after their purchase? How regularly should we checkin after that first initial follow-up?

All of these questions can be answered with A/B testing. Automated collections management software like receeve allows you to A/B test your collections outreach on an ongoing basis. Machine learning algorithms will analyse the results, segment your customers accordingly, and allow you to devise the most effective communications strategy possible for each individual or classes of debtor(s).

What your A/B testing might look like

Imagine you’re sending out an email reminding debtors that their repayment is overdue. You obviously want to test out which copy works best—so you create version A and version B.

  • Version A

{First_Name},
Your outstanding payment of {Amount_Due} is OVERDUE. Your dunning fees have therefore been INCREASED by 100%, leading to a new total of {New Amount_Due}.
PAY NOW to avoid further late-payment fees and the risk of legal action.
{Company_Name}

  • Version B

Hello {first name},

Despite multiple requests, we still haven’t received any feedback from you. Your outstanding payment of {Amount_Due} is still pending.

Ignoring it only makes the problem worse. Delaying payment can increase an initially small amount by 200% and more due to dunning fees that continuously pile up. You can easily prevent this by being proactive!

Just conveniently pay the open amount here online or select options to help you go through financial hardships:

Alternatively, the classic bank transfer to the following account is

also possible: {Transfer_Details}

Many Greetings,

{Company_Name}

You might hypothesise that Version A, with its sense of urgency, will be more effective—but your colleague thinks that Version B’s softer tone will strike a chord with modern consumers. Simply set up these two emails, use your automated collects management software to send one version out to each relevant debtor, and wait for the results to come back in.

Make A/B testing your secret weapon

Machine-learning-driven A/B testing has the power to transform your collections strategy going forward. But it can be hard to know where—or how—to begin. And believe us, it’s no walk in the park.

Do you want to turn A/B testing into your competitive advantage?. If so, keep an eye open for our digital workshops on best practices in modern debt collection. Feel free to subscribe below to our blog to get the latest updates. Make ongoing improvement a reality, not a pie in the sky.

Author
Jan Frommann
Linkedin Icon
Whitepapers

next-arrow

Ready to get started?

If so, head over to our demo page and learn more about receeve’s leading collections management software.

Book a Demo
Debt Sale
Gradient Hexagons

Looking for some inspiration?

Sign up to receeve's newsletter and never miss a beat.

Gradient Hexagons