Payment propensity model for operational efficiency
Ageing accounts receivables and credit risk management is an important factor towards cashflow for most companies, with some sectors impacted more than others.
According to a recent report by Dun & Bradstreet, over 90 days ageing for some of the sectors is over 50%.
Package software service: 67%
Misc. publishing: 59%
Commercial printing: 55%
Retail auto and home supplies: 44%
Wholesale medical/hospital equipment: 24%
An aggressive post billing follow-up process could be a natural tendency to solve challenges around payment collection and improve the payment ageing past due date.
However an aggressive strategy is also likely to have its side effect, which will impact:
Too many follow-ups are likely to annoy your clients and customers and, at the same time, increase the effort required by the accounts receivables team. This is definitely a case of striking a balance. Knowing where to concentrate your effort is an important operational decision which can contribute positively towards maintaining a healthy cashflow. A payment propensity model, when designed and integrated with your billing and post invoicing processes, can provide the necessary answers to where you should concentrate in order to maximise payment collection with minimal effort.
What is a propensity model?
In its simplest form, a payment propensity model is a machine learning classification algorithm, but, depending on the data used, it could also be a regression model.
Irrespective of the algorithm used, the primary objective is to use past data about customers to create segmentation and derive the likelihood of receiving payment in a given time period.
The propensity scores and the billing amounts can be visualised as follows. The high-high and high-medium are examples of key accounts to focus the collection efforts on.
An ongoing evaluation of propensity scores can be used for strategic reporting and planning purposes, providing key insight into important areas like cashflow, customer behaviour, pricing strategies and even toward product design.
Another important concept is to design and implement this as an operational process instead of treating a propensity model as a tool for data science only (reporting and collecting insights).
When implemented and integrated with an operational business process like post invoicing follow-up, a propensity model is used for deriving, prioritising and managing work lists (or task lists) for your accounts receivable team.
Here the pending accounts receivables data is merged with the payment propensity scores created at the time of billing to derive a prioritised order. This ensures that payments which could be received are received on time and also the ones at most risk are given the necessary importance.
Prebuilt solutions and packaged software definitely have their place, however, payment propensity solutions are not one of them. It is crucial to ensure that the solution works for a given organisation and that the tendency to deploy a one-size-fits-all solution is avoided.
Like most AI models, knowing which data features to use and where to source them from is the key challenges. We recommend exploring external (public) data sources along with internal data points to improve propensity scores.
Avoid "one-size-fits-all" models. A prebuilt propensity model which works for one company may not necessarily deliver same results for another. It is necessary to undertake the required data science for your products and customers and then design a bespoke model using features unique for your organisation.
Wondering how this might be of use to your organisation? Schedule a call with one of our consultants to discuss how such a strategy could help your business improve operational efficiency and cashflow.
Want to know more about possible AI use-cases for you business?, request a case study here.