On March 17th, 2021, Zuora announced the launch of Zuora Collect AI as an addition to its suite of subscription management applications. As a research firm focused on the practical use of AI in the enterprise and a firm that provides guidance on subscription finance, expense management, invoice processing, and business model management, we explored this announcement to consider what aspects of artificial intelligence have been brought into this product and the business value that can be expected from this launch.
Zuora Collect AI is focused on the value of payment processing and collections. This application analyzes a variety of transaction characteristics to discover top drivers that affect payment and revenue recognition. These characteristics include payment methods such as the bank, credit card, or payment gateway involved as well as client characteristics such as the region, payment value, and history of payment success.
Rather than simply treat all customers as a single payment cohort under the same calendar for all payment processing activities, Zuora Collect AI allows for greater granularity of payment pursuit, transaction channels, and recovery tactics. As a result, Zuora customers can be expected to reduce the dunning rates and delinquent payment status while providing a more efficient dunning process for late payers. In addition, Zuora Collect AI collects data daily from a variety of payment gateways and payment methods to continue updating parameters associated with payment success over time.
This use case is a strong value proposition for machine learning in that it combines the volume of billions of transactions across hundreds of currencies and regions and uses this information to solve a million dollar problem for enterprises. Zuora’s use case fits into the framework for practical, high-value AI that Amalgam Insights has advocated for years.
To elaborate on how this is a scalable issue, consider that Amalgam Insights estimates that the average subscription revenue company faces revenue leakage of 2% when using a billing system and will only recover 40 – 50% of that revenue over time. With better tuned and scheduled payment processes, organizations could gain an extra 10 – 20% in revenue recovery.
This difference would have equated to a 5% – 10% percent increase in net profit in 2020 for the average non-financial services firm that registered a 4% net margin in 2020.
In Zuora’s public-facing documentation, they have documented three organizations (Whitepages, RankingCoach, and Motortrend Group) with results ranging from 10 – 18% increases in payment recovery through the use of Zuora Collect AI compared to manual payment retries and an umbrella policy for retrying payments within collections dunning cohorts.
Although this type of payment analysis could theoretically be done by an on-site data scientist, Amalgam Insights notes that one of the key challenges with expense and payment optimization is the changing nature of the data as new payment types, schedules, quantities, and patterns emerge. Because of these changes, transactional optimization is not served well by creating a one-time model to find optimal payment processing times and patterns for each customer and payment type. At volume, it requires ongoing data intake and monitoring to maintain the efficacy that maximizes revenue optimization which is better served through treating revenue optimization as an ongoing process and service rather than a one-time audit and model creation.
Based on this announcement, Amalgam Insights makes the following recommendation for subscription and usage revenue-based organizations:
First, pursue the use of machine learning to support revenue recovery.
This one activity has the potential to increase your net profit margin by 10% or more even if your organization already has a mature dunning process for subscription customers.
Second, invest in the ongoing maintenance and updating of these models.
To maintain these gains over time, your organization will need to invest in daily data processing and ongoing model optimization to ensure that your payment collection schedules and processes keep up with ongoing trends. Otherwise, Amalgam Insights estimates that the models created will lose their value within 18-24 months and leave the organization back at a point where significant revenue is being unnecessarily lost due to poor dunning and payment processing schedules.
The big takeaway here is that machine learning continues to be brought into the business world to solve the highly transactional analyses that are too time-consuming to be solved through manual analysis or even through traditional data analysis tools. Take advantage of the increasing availability of productized AI and the data associated with digital payments to solve operational issues with million-dollar payback potential.