Can AI Reverse The Subscription Churn Curse?

“Recurring payments” is a bit of a misnomer, when approximately 13 percent of them don’t recur at all. Instead, the card on file gets declined, despite the fact that it was working fine last month, last week or whenever the customer subscribed to the product or content.

This is, after all, the onboard phase fraud checks. The merchant knows the card is good and the customer is in good standing but, for whatever reason, the payment isn’t going through. And, unless it gets resolved, that customer is about to churn out of the business whether he or she wants to or not.

With churn, it doesn’t matter if it’s voluntary or involuntary. Lost customers represent lost revenue, damaged subscriber relationships and declining reputation.

According to Dan Burkhart, CEO and co-founder at subscription and recurring billing platform Recurly, this is a problem across every type of subscription business, from physical “box-of-the-month” clubs to purely digital subscriptions to things like music, TV shows or games.

Although reasons vary between business-to-business (B2B) and business-to-consumer (B2C) settings, Burkhart said recurring payment declines are often due to one of these factors: The card is past its date of expiration, it has insufficient funds, there’s a temporary hold on the card or its credit limit has been reached.

When a merchant tries running the payment a second time, each of these scenarios calls for a different response, he explained. An automated static response across all scenarios may solve for some declines, but it will not return the desired result in every instance.

Without looking at each situation individually, responding in the proper fashion to save that transaction and that subscriber is a shot in the dark. Yet 13 percent of all recurring payments is far too high a number for anyone to be reviewing these issues manually, or to rely upon a one-size-fits-all static model.

This is a job for artificial intelligence (AI) and machine learning, Burkhart said. In a recent interview with Karen Webster, he explained how Recurly approaches the problem with its new Revenue Optimization Engine and why it’s important to optimize the response depending on the individual merchant’s needs.

Try Smarter, Not Harder

If at first you don’t succeed, don’t “try, try again.” Get it right the second time by taking a logical approach to transactions that need to be retried.

Burkhart said Recurly has trained its machine learning models to keep track of hundreds of dynamic variables, just as search engine algorithms track and weigh hundreds of variables — and send businesses scrambling to optimize for the new algorithms anytime something changes.

In the same way, Recurly is keeping an eye on what changes throughout the day, week and month in conjunction with issuing banks, countries of origin, expiration dates and prior attempts to use cards that may have been declined.

According to Burkhart, it’s not just about optimizing in a default logic way based on error codes and gateway combinations. It takes a whole array of combinations to provide insight at the transaction-specific level, which is required to differentiate oneself in this space.

Merchant Priorities

Before offering a solution to recapture declined transactions, it’s critical to understand the merchant’s priorities. Is it concerned about the cost of goods, which are expensive to keep delivering during the retry process? If so, then optimization should be for days outstanding to limit exposure to costs.

Burkhart said this would be true for box-of-the-month clubs, which don’t want to ship goods until revenue has been captured. Time is also of the essence for media companies with a high cost of goods, such as licensing rights for sporting events.

Or, say an ad network is delivering media, such as banner ads. If a recurring payment fails because the card was declined, the network will want to stop its negative outlay and halt or pause the banner ad from appearing until payment is received, he added.

In these cases, the Revenue Optimization Engine could help merchants decide whether to provide access to their goods and services before incurring costs they can’t recover, Burkhart explained. They’re essentially going negative on the customer in hopes it will make good on the payment — just as they do by extending free trials to prospective customers in hopes of making a conversion.

On the other hand, if the merchant’s priority is revenue recovery, merchants aren’t worried about it taking a little longer — often from 15 to 30 days. In categories with high-gross margins like digital goods and services, there’s nothing wrong with taking 30 days to resolve a payment issue if the merchant’s goal is to recover the highest percentage of revenue possible.

“Each merchant is a unique snowflake — that has unique characteristics in terms of card type, currencies, transaction amounts, gateways, et cetera — that we can point this model toward so [the] engine can deliver optimal results for each merchant,” said Burkhart.

Take the Right Risk Upfront

Recurly fed eight years’ worth of transaction data into its machine learning algorithms to create the Revenue Optimization Engine. That data can also be used to help merchants determine which acquisition channels are most attractive on a relative basis, he noted.

Most companies do testing to determine which of their channels are delivering lower churn, higher lifetime value customers and better economics. Subscription companies should do the same, said Burkhart, but the really critical piece is the speed with which they’re able to prune low-performing partners and channels.

Companies must understand their customer acquisition costs and the relationship between that and the channel sources that bring in those customers. They must then focus their dollars on higher-performing acquisition channels. That, Burkhart said, is the difference between success and failure.

The interplay between payments processing and marketing automation technology is not just about billing automation, he added. The end goal should always be to optimize revenue through better decisions and identifying the highest quality sources of customers.

In short, it’s not just about recovery. It’s about acquiring better customers in the first place.