How Recurly Uses Machine Learning to Reduce Transaction Declines
One of the significant benefits of subscription commerce is the amount of data and related insights this model generates compared to one-time purchases. New data related to marketing, payments, and customer lifecycle events is generated regularly, at each new billing cycle. This data is invaluable for gaining insights and making decisions that help you to optimize your business.
For example, one strategic way that subscription businesses can use this data is in terms of payments. When a recurring payment fails and is not repaired, the subscriber is lost to involuntary churn. Minimizing this kind of churn is critical for healthy business growth. But payments is a complex process with many players, varying approaches, and unpredictable results. What is predictable is the fact that recurring credit card transactions fail for a variety of reasons—and these reasons can and do change over time.
Machine learning for pattern analysis
Increasingly, Artificial Intelligence (AI) and machine learning are being utilized to efficiently and intelligently analyze large sets of data to identify patterns, which can then be used to improve decision-making. For example, online fraud is an area that uses AI to improve the identification of potentially fraudulent transactions and the patterns that may be behind different types of fraud.
The payments process is another area that can benefit from the power of AI and machine learning. In fact, Recurly began incorporating these powerful analytic capabilities several years ago, with the launch of our Revenue Optimization Engine. When a recurring transaction fails, this technology creates a customized retry schedule for each transaction, so subsequent retries have a higher chance of succeeding.
Significantly improved transaction success rates
Through this sophisticated retry technology, Recurly has been able to significantly improve transaction success rates for our merchants. In fact, Recurly is able to recover, on average, 61% of initially failed transactions through our decline management and retry technology.
Machine learning models are most effective when they are constantly updated and based on new data. We’ve updated the model that powers the Revenue Optimization Engine twice since it was launched, resulting in significant improvements to the accuracy of the model’s predictions.
Iteration for the win
One of Recurly’s core values is to iterate everything to continually improve our product, features, and processes. While retraining our retry model is an excellent first step, there are other ways to improve the model’s performance. For example, we recently developed a newer version of the model that includes new training data, additional model features, and a different machine learning algorithm.
For more information about decline management and how Recurly uses sophisticated tools, including AI and machine learning to minimize and repair declined transactions, be sure to read our blog.
And, check out our Subscription Commerce Predictions for 2020 and Beyond and see how the world of subscription commerce may change and evolve in the year to come and beyond,