Subscription Companies Use AI To Keep False Declines To Record Lows

You could call it the false decline fallout. That’s where a good and functioning eCommerce subscription relationship suddenly goes haywire due to the unexpected and unwanted intrusion of a false payment decline into an otherwise perfect business arrangement.

Whatever the cause, these mistaken friction points, while aimed at preventing fraud, can also inject dissatisfaction into the relationship where none existed before.

There will always be declines in the billing process and transactions that get blocked, Digital River Chief Payment Officer Eric Christensen told PYMNTS in a recent conversation. Accounts will have insufficient funds to cover a purchase the customer wants to make, or a card will have been reported lost or stolen shortly before a bill hits. Getting to a 100 percent authorization rate, he said, is not possible.

“But trying to get to a rate at about 95 percent, that number is really what we strive to get to. That means that any legitimate transaction that you want to get through, you can get through,” he said.

And it matters, he said, particularly for the subscription consumer. That shopper is there for a reason, he said — they wanted that product or service enough to put it on subscription to ensure it would be regular and easy for them to access. Offering that certainty, alongside whatever good or service on offer, is critical for merchants looking to maintain a positive relationship with their customers.

With its myriad points where a transaction can slip through and fail despite being legitimate, the payment system doesn’t always cooperate, he said, meaning merchants have to go the extra technological mile to make sure they are minimizing the impacts and pain points false declines can introduce.

Tapping The Power Of Machine Learning

The payments ecosystem is a very dynamic place with all kinds of movement happening that can make an otherwise good payment fail to get approved. Some of those things are easy to foresee, he said, like when Bank of America ships millions of new cards to consumers at once. They can know well in advance, he said, that as banks issue those new cards, consumers aren’t going to update them properly and a lot of transactions are going to fail — meaning a lot of work has to go on behind the scenes to tie those new cards to old cards to avoid a lot of failed transactions.

But some things are a lot harder for human eyes to spot — like optimizing transactions can involve accounting for seemingly random details like time of day. And this, he noted, is where machine learning and artificial intelligence (AI) become critically important because they can account for all these tiny details in banks’ security protocols to find the most optimized path through which to route payments for authorization.

“We’ve spent years kind of poring over data and trying to figure out the right time of day, the right day of week, the right data points to process on a transaction,” Christensen said. “And as we’ve built out some machine learning models, we’re learning much faster that way than we were with a human with human eyes overseeing the process.”

The High Cost Of Failure 

Digital River increasingly sees companies, and subscription firms, in particular, starting to invest in their billing optimization strategies, Christensen said. However, he believes the market is still in the “early days” and there’s still a lot of learning that needs to happen.

Most critically, he said, what merchants need to understand is that a busted billing process that rejects a good customer transaction is a single problem that often leads to bigger issues downstream.

“Depending on what your billing cycle is, if it’s monthly or annually losing that customer, it isn’t just an impact on this year’s bottom line, it’s next year, and the year after,” Christensen said. “Being able to retain them is a revenue stream for the merchant on a continuous basis.”

And one that merchants can only maximally retain by better optimizing their billing process to ensure that every transaction that should go through actually makes it through.