Time to Face the Facts on Failed Payments and Subscriber Churn and the AI/ML Remedies That Can Fix Them

Whoever said “what you don’t know can’t hurt you” clearly never operated an eCommerce subscription business.

Take failed payments. They’re endemic to subscriptions and cause scary levels of involuntary churn, yet they are usually so murky that teams often don’t even know there’s a problem until it’s too late.

That’s the opinion of FlexPay Founder and CEO Darryl Hicks, who identified the main reasons that recurring payments fail, the high cost of resulting churn and the technology solutions that are turning this problem around.

“Most of the companies we see out there are using digital engagement or a people-powered strategy, which is fine and works really, really well,” he told Karen Webster. “The challenge comes in that some of these strategies can actually further exacerbate churn, causing even more churn.”

Hicks said that every customer engagement carries risk, and none more so than those dealing with declines.

“Maybe it’s embarrassing to the customer,” he said. “They feel like there’s nothing wrong with their card because in [most] cases there is nothing wrong with their card. They feel there must be something wrong with you, the merchant, just causing issues.”

That shadow looming over delicate subscriber relations calls for an equally stealthy response, “which is why we talk a lot about invisible recovery,” Hicks told Webster. “An invisible recovery solution is always the best recovery solution. Whatever you can do to fix the friction, to fix the churn without engaging with your customers at all, is always going to be a better experience.”

Research supports this view. According to PYMNTS’ new Optimizing Subscription Payments: How Providers Can Take The Sting Out Of Payment Declines report, done in collaboration with FlexPay, “Of subscribers who suffered from payment declines and payment-related issues within the past 18 months who say these issues improved, 73% report being highly satisfied with their subscriptions. This indicates that subscription providers still have a critical chance to retain customers’ loyalty by offering ways for them to overcome payment-related snags.”

See the study: Optimizing Subscription Payments

When ‘Suspected Fraudulent’ Isn’t Fraud

Failed subscription payments are a complex issue with many moving parts, Hicks explained, adding that artificial intelligence (AI) and machine learning (ML) are offering vital solutions.

Referencing a Mastercard white paper estimating that fraud losses for issuing banks were running at $35 billion to $40 billion annually pre-pandemic, Hicks said it’s a figure banks and issuers are doing everything they can to reduce, and he noted that the only thing they really can do is decline transactions that are suspected to be fraudulent.

Webster emphasized the “suspected fraudulent” part as where the breakdown happens, and Hicks agreed.

“They don’t know 100% for sure, they’re trying to stay ahead of the fraudsters and figure out what they think is fraud,” he said, adding this represents 87% of all failed payments that exist in card-not-present (CNP) payments today.

Data is both the problem and the solution, he said, noting that many banks and issuers don’t have the best quality data to make educated decisions about which transaction to approve and which to decline.

“They don’t receive data like your email address or your IP address, or the device ID fingerprint, geolocation, shopping cart contents, and all of these things that are directly relevant to the fraud profile of a transaction,” he said.

That’s where AI/ML-powered “invisible recovery” solutions do their thing.

“Why AI and ML?” he said. “Well, once you understand that the issuers have a checklist of all the things that they’re looking at that decide whether or not they want to approve or decline that transaction, you can use AI and machine learning to custom tailor all the different data elements to better match what it is that the issuers are looking for.”

Hicks recited what he called his favorite low-hanging fruit to go after in failed payments scenarios, from timing batches to East Coast business hours, to the risk model of a given merchant processor, use of correct merchant category codes, gateway flags, data fields and many other variables.

“There’s a lot you can do with best practices upfront to reduce the volume of the problem,” he said. “Beyond that, you just need to have a really sophisticated technology solution that’s leveraging AI/ML using giga-data sets with billions of transaction records, the quality of the data, as well as the creative understanding of how to actually train the machine learning models in order to fit into the problem and better ameliorate it.”

A key component is creating trust and transparency in payment rails “by bringing all the stakeholders into the conversation, which is one of the things we’ve been working on,” he said. “Bringing in the networks, the issuers, the acquirers, the merchants. Having everyone sit down and say, ‘How are these systems working? Where are they broken? How can we make a better overall system?’ AI/ML is a perfect tool that sits in the middle of that and makes it work better.”

Read also: Failed Subscription Payments Hurt but Don’t Have to End in Cancellation

Payments Rails Need Work

Hicks said he is a proponent of combining data from issuers, acquirers and merchants.

“It’s the juxtaposition of those three different data sources that really [improve] your transactions with a whole bunch of additional metadata that’s very rich and allows you to create a more holistic picture of what’s going on with the transaction,” he told Webster. “Then we use AI and ML on these particular aspects and attributes to get the best outcome.”

However, FinTechs trying to sort out card declines for the subscription sector run up against legacy constructs that were built for opacity in a competitive financial marketplace.

“I think that Visa and Mastercard deliberately created a culture of separating merchants from issuers and never letting them speak because they never wanted to be disintermediated,” Hicks said. “The problem is that now you have these two sides operating in silos, black boxes completely sealed off from one another who ultimately start acting in a self-interested manner. The networks would love it if there was more trust and transparency in the rails, but this runs counter-intuitive to their own business model.”

The answer, he said, is a “significant upgrade and update to the payment rails, and it can’t happen quickly enough. The good news is there are great things that merchants can do today. I think that they’re not doing enough because first of all, they’re not aware of the problem.”

Hicks noted that “there is a finite pool of consumers out there that are willing to engage with your brand and who want your product or service. The best you can do to [is] retain the people you have to increase overall profitability and create a better brand experience.”

To achieve this, Hicks told Webster, “I would really love it if CEOs would sit down with their executive team and drive answers to these four questions: what’s our current failed payments rate; what method are we using today to recover when a failed payment occurs; how much recovery are we earning off of that solution we’re using; and what’s going to be our technology strategy to increase the total recovery rates while avoiding recovery methods that can actually create churn? Have that; have the conversation. It will be well worth your time.”

See also: Almost Half of Subscription Churn Caused by Preventable Friction in Failed Payments