Stop Over Declining Good Orders, Chargebacks and Fraud Prevention with Ellie Ronning of Signifyd
Watch the Episode
This episode is sponsored by the B2B Commerce Trends and Opportunities Report, presented by Shopware and written and researched by Content Cucumber. Grab your free copy at contentcucumber.com/b2b-commerce-trends-report.
Ellie Ronning works on the enterprise strategy team at Signifyd, which she calls a fancy way of saying she gets to partner with and advise some of the largest direct-to-consumer retailers and merchants in the country. She spent fifteen plus years on the merchant side in New York before she crossed over to the vendor side, and I sat down with her at Shoptalk Fall in Nashville, where I opened by suggesting we rename the show Talk Shop. My old Magento agency was a Signifyd partner, so I knew the company, and I asked her for the elevator pitch anyway.
A Decision With Money Behind It
“Signifyd is a commerce protection platform. We’re effectively a decisioning engine, a decisioning layer that spans the customer journey,” she says.
That layer runs from account login and account modification through checkout and returns, sitting in the background while the merchant decides whether to approve or decline. The part that sets the model apart is what comes attached to each decision.
“With that comes what we call a liability shift. So we actually indemnify our retailers and we say, hey, we’re gonna give you a decision on whether or not to approve or decline this transaction and we’ll back it for you,” she says.
The Merchant With No Chargeback Problem
I asked about false positives, the good order a merchant turns away because the order looks risky, and Ellie described a conversation she has with merchants often.
“Oftentimes merchants will come to us and say, you know, we don’t have a chargeback problem, right?” she says.
When Signifyd pulls the numbers, the low chargeback rate turns out to have a cost. The merchant is declining orders to keep chargebacks down, and the orders being declined belong to people who were going to pay.
“They’re leaving a lot of money on the table by over declining,” she says.
“These are decent shoppers, they’re probably not fraudsters, they’re probably not, you know, abusers,” she says.
The goal she described is to open up that funnel, let the good orders through, and drive up the approval and authorization rate, with Signifyd carrying the chargeback when a Signifyd approval turns out to be fraud.
Seeing Around Corners
I gave her a scenario. A bad actor buys something at Ellie’s cupcakes in downtown Saint Paul, and somebody with the same profile is buying in San Diego. Ellie pointed out that Signifyd plays in the card not present space, which means ecommerce, and then explained what the network does with a shopper who is new to the merchant.
“We can truly see around corners for retailers,” she says.
A first-time customer is a blank for the merchant. Signifyd draws on transactional data across the merchants it works with, so the same shopper has often turned up somewhere else.
“When you work with a Signifyd, we’ve probably seen some attribute of that customer, whether it’s your email address or your phone number, elsewhere in our network because you’re shopping other merchants that we work with, right?” she says.
Machine Learning Before It Was Cool
I asked how Signifyd uses AI, since a lot of fraud work looks like if-then rules with a lot of gray in between. Ellie said the company has been around for about fifteen years and has worked this way from the start.
“We’ve been a machine learning company since before it was cool,” she says.
The new work is agentic. A transaction executed by an agent with a real person behind it looks different from a person checking out on a website, and Signifyd is defining what an agentic transaction is so its models can tell the two apart.
“Making sure that our models know when it’s an agent versus when it’s a human, right?” she says.
Agentic Commerce Starts Upstream
Ellie thinks the hype runs ahead of adoption, and she has not made an agentic purchase herself yet.
“Okay, you had one an agentic transaction last year and you have three, so my gosh, it’s two hundred percent growth,” she says.
Where she sees merchants leaning in is search and discovery, because a shopper asking an agent for marathon shoes in a certain city in a certain month is a far more contextual query than anything from the old SEO days. Her example landed close to home, since I had been searching for shoes myself and was switching from Hoka to Adidas.
“They’ve gotta start to refine their item data, the attributes, like the the brand sentiment, the brand voice, right?” she says.
She expects categories to split, with subscription and utility products adopting agentic buying faster than apparel, where touch and feel and the emotion of shopping still matter.
AI Slop and a Campaign Pulled Back
On AI slop, Ellie told me about an old colleague who leads brand marketing at a footwear company in New York. The brand ran a campaign with a real person augmented by AI, consumers called it out right away, and the team pulled it back.
“It was such a learning for us because it felt like we were putting AI slop out into the world,” is how her colleague put it.
The lesson Ellie took is that consumers can tell, and brands that swing the pendulum too far will hear about it.
Unified Commerce Becomes the Forcing Function
Her hot take tied the conversation together. Merchants are cleaning up item data for search and discovery first, and the transaction side will follow on the platforms already lined up to power it. The bigger shift is what agentic commerce does to the data underneath.
“I do think moving into this agentic space will sort of be a forcing function for merchants to actually roll out a unified commerce strategy, right? You really can’t have siloed data in an agentic world,” she says.
We left it with plans for coffee in Minneapolis. Find Signifyd at signifyd.com and Ellie on LinkedIn at linkedin.com/in/ellieronning.