Improving collections strategy by 30% with machine learning
Contacting customers digitally to recover overdue accounts is one way to improve your customer experience.
But predicting how each customer prefers to engage, and adapting each step of their journey accordingly takes your collections to entirely new levels of personalisation.
Let’s break down how our proprietary machine learning does exactly that, to drive Collect’s strategy and deliver a 30% uplift in collections performance.
What are machine learning models?
Machine learning (ML) models are a form of artificial intelligence. They use large amounts of data to generate insights and predict specific outcomes. But the real beauty behind this technology is that the models automatically improve in real-time, as they receive more data.
So how is it used in debt collection? Well, our ML models remove the guesswork.
Using machine learning, Collect can:
- Predict how a customer is most likely to engage with their debt
- Personalise each collections journey according to unique preferences and behaviours
- Adapt communications intuitively in real-time - how a customer interacts with one communication informs what they receive next
- Protect customers with our hard-coded compliance firewall, ensuring all collections actively stays within local regulations
- Scale recoveries with no limit and equal work effort across all customer accounts
Unlike traditional methods that rely on human intuition or selective metrics, ML applies real-time insights at a customer level to create the ultimate collections strategy.
How do Collect’s machine learning models work?
Every time a customer engages with us, it gets recorded as a unique event in our data warehouse, contributing to our 850 million events (and counting).
Think of events as interactions - such as a customer opening an email, the time they opened it, how long they spent reading it, whether they clicked on the payment link or contacted our Customer Experience team.
These events are then analysed by machine learning models to determine what communication should be sent next, driving strategy at each individual touchpoint.
At a high level, our models are split across three areas: when to contact, what to send, and how to contact.
Let’s look at how these models work in practice. In these examples, we’ll follow David who has an outstanding debt with MyBank.
What to send
Analysis
To get his journey right from the get go, Collect analyses David’s profile against our data warehouse, to understand his behaviours and preferences. This includes factors such as the amount of debt he has and the origin of that debt.
This analysis feeds into our ML models, to determine what message David should receive.
Messages
Using a combination of human copywriters and our custom-built
AI copywriter, our models have thousands of communications to pick from.
Each message undergoes a thorough Compliance review and performance testing, before being deployed. This ensures that every communication Collect sends is high-performing and compliant. Take our AI copywriter-written messages for example, which have increased conversion rates by 32%.
Selection
When it comes to selection, our ML models meticulously choose the message that’s most likely to encourage David to take action on his account. They factor in every detail, such as the subject line, tone, call to action and more into their decision making - maximising conversion at every opportunity.
And it doesn’t stop there. With our
Customer Journey model, David’s interaction with one message informs what he’s sent next. The result? A personalised and responsive customer experience - throughout the entire collections journey.
When to contact
The right message at the wrong time is all too typical in traditional collections.
To contact David at the perfect moment, every message is sent at the time he’s most likely to engage - down to the specific day and time. This shift alone increases email payment conversions by up to 20%.
Frequency is equally important here, to prevent overcommunication or customers left feeling harassed. Our ML models strike the perfect balance between protecting both deliverability and breathing space, resulting in InDebted maintaining some of the lowest spam report rates in the industry.
How to send
David’s frictionless collections experience is driven by Collect’s omnichannel engagement.
By recognising which channels David is most responsive to, our ML models can tailor his experience to fit his engagement style. Simply being available on the channels our customers prefer removes the roadblocks and drives performance further, delivering up to 7x higher customer engagement than traditional methods.
The next phase: Introducing the AI Collector
Reliance on human agents was one of the biggest hurdles holding debt collectors back from growth – until now.
Launched in 2024, our AI Collector uses bespoke conversational AI technology to handle inbound customer inquiries.
Using our data warehouse and hard-coded compliance firewall, the AI Collector operates with unmatched efficiency. Thanks to two machine learning models - one to classify the purpose of the inquiry and another to determine what action to take, 80% of all inbound email enquiries are resolved by the AI Collector.
For example, it can provide customers with account information, update existing details or send a unique payment link within seconds. This means simple queries are addressed instantly and accurately, giving our customers the autonomy they need, to get back on track faster.
The knock-on effect is significant, as it frees up our Customer Experience team to focus on more complex cases. Currently live in Beta with one of our largest Australian clients, the AI Collector is already improving efficiency and customer experience.
Collect: The new generation collections solution
Collect’s machine learning models are changing how customers manage their debt, and how businesses maintain a gold standard customer experience.
This technology is already driving 30% higher performance - and we’re continuing to break that glass ceiling. With continual investment in model development, data security and compliance, our clients can recover debt faster, safer and better than ever before.
Step into the new generation of collections, you won’t look back.
See how it works
Únase a nuestro boletín para conocer las últimas novedades sobre colecciones
Thank you for subscribing!
Únase a nuestro boletín para conocer las últimas novedades sobre colecciones
Thank you for subscribing!
Compartir
Otros recursos
AI debt collection vs. traditional methods: what's different?
While traditional debt collection is increasingly seen as out of touch, a new generation solution is disrupting the landscape — AI-powered debt collection. It puts the power back into customer’s hands, to manage their debt their way.Improving collections strategy by 30% with machine learning
Contacting customers digitally to recover overdue accounts is one way to improve your customer experience. But predicting how each customer prefers to engage, and adapting each step of their journey accordingly takes your collections to entirely new levels of personalisation.Best practices for chatbots in consumer finance
The Consumer Financial Protection Bureau has outlined their key concerns around using chatbots. Find out how to implement these new tools safely and efficiently with this guide from Tim Collins, Chief Compliance Officer.Choosing a debt collection partner that understands today's consumers
In such a competitive environment, fintechs and buy now, pay later providers are laser focussed on innovation and growth, leaving little time to develop a collections strategy. Outsourcing can be an effective solution, as long as you know what to look for in a debt collections agency.