Optimizing Debt Collection
Increasing the probability of payment claims in the debt collection industry for coeo Incasso
An all too common scenario for individuals or corporations: products or services are purchased or rented, but invoices go unpaid. Most of these cases are settled after a simple payment reminder, but a large portion remains unpaid. It is an uncomfortable situation for all parties but especially aggravating to the debtor, for whom uncertainty over future payments starts to build and leads to the last-ditch effort for many corporations: hiring a debt collection agency.
In coeo Incasso, we find one of the most modern and technologically innovative debt collection agencies. At the center of all their actions is one central aim: finding an amicable solution for everyone.
Due to significant growth in recent years, coeo Incasso faced a high increase in incoming claims. This presented an immense challenge for the organization, as validating and processing every single claim is time and resource-consuming. As so, the organization saw potential for the use of artificial intelligence and machine learning early on.
‘Debt collection optimization’ and the challenges for coeo
Historically, debt collection has been a rather “manual” and time-consuming process. In present times, validating and following up on every claim, even those with a low probability of success, is simply not efficient or realistic.
To free time and be able to handle all claims properly,coeo looked into a first AI use case: prioritizing claims and automating actions. Based on the ‘likelihood-to-pay’ in relation to days of outstanding payments, a machine learning model was developed.
However, as many companies experience, simply deciding on and building a model does not guarantee actual usage. A key business process would be (partly) replaced by automated decisions and, since machine learning still presents an unknown element, model transparency and trust among employees were essential.
In addition, to ensure legal accountability, every prediction, as well as the variables that led to the prediction, need to be reproducible and explainable at any time.
These challenges are a perfect fit with Deeploy’s mission to ensure explainability, accountability, and manageability for machine learning.
Use case ‘Debt collection optimization’
- Ambition: ensuring that the largest possible share of outstanding payments is settled
- Situation: evaluation of claims, estimation of likelihood of payments per customer and resource prioritization (e.g. people, time) conducted manually
- Models: likelihood-to-pay in relation to days of outstanding payments resulting in automatically triggered measures
- Solutions needed to integrate machine learning into business processes to level fast growth while simultaneously reducing high manual effort
- Limited link and transparency between business units and machine learning applications
- Desire to reproduce predictions at any time to evaluate the taken measures
The solution: enabling human-machine interaction and accountability through Deeploy
Deeploy’s platform allows Coeo to easily deploy and manage models in one place. Teams can check on model health through performance metrics such as accuracy and input value drift as well as set alerts for when these metrics deviate from established limits.
But, even more importantly, Deeploy offers features to implement explainability and human feedback loops.
Based on developed custom explainers, Deeploy is able to give Coeo’s business and legal experts detailed insights on each prediction made by the model.
Complete transparency and explainability allow experts to evaluate predictions and provide feedback on the model logic.
Placing high emphasis on accountability and reproducibility, Deeploy also allows Coeo to log and trace every change and prediction made by a model, creating clear audit trails. In that way, Coeo is able to maintain full documentation on the decisions made by the implemented models.
Realized benefits through Deeploy
Coeo trusts Deeploy as a tool to establish transparency and explainability of implemented models, while simultaneously ensuring accountability and manageability.
By implementing ML models in this manner, Coeo was able to free precious time and resources without losing employee, stakeholder, and consumer trust. This extra time can now be used to focus on claims that need the involvement to generate payouts.
All in all, AI and ML is fostering Coeo’s fast expansion and growth and Deeploy is helping do so responsibly, thus building a solid base for Coeo’s future growth!
- Version control of historic models & predictions
- Automated audit trails
- Customized explainers for every prediction
- Human-feedback loop for business experts
“Deeploy has enabled us to respond to the customer’s situation at the debtor level and to take targeted personal actions so that we can put empathy in the way we communicate with them. This is an important aspect for our clients in the context of customer retention.” – coeo Incasso
Want to explore the full potential of Deeploy?