Why 73% of Your Declined Orders Are Actually Good Customers

Steve Hutt: Welcome back to eCommerce Fastlane. I’m your host, Steve Hutt. Today’s conversation is all about fraud and the good customers we lose trying to stop it. If you’ve ever canceled a risky-looking order in your Shopify admin and never found out whether that customer was real, this episode is for you. According to Signifyd data, around 73% of declined orders are actually good customers. That’s revenue you’ve already paid to acquire, and it’s walking out the door. My guest today is Nicole Jass, Senior Vice President of Enterprise Strategy at Signifyd. Before Signifyd, Nicole was Chief Product Officer at the identity company Prove. Before that, she spent several years on the product side at Worldpay. She’s seen fraud from the payments side and the identity side, and now she’s on the merchant side. We cover a lot of ground in this episode: why fraud is growing faster than ecommerce itself, how AI is putting the playbooks of organized fraud rings into the hands of amateur fraudsters, what a card testing attack looks like from the inside, and how fraud has moved beyond checkout into account takeover, loyalty points, and returns. We also discuss where Shopify Protect stops covering you, especially if you ship outside the U.S., and what AI shopping agents change about who or what is placing your orders. With holiday traffic starting to ramp up and a wave of first-time customers hitting your store, the timing couldn’t be better. Hi, Nicole. Great to have you on the show.

Nicole Jass: Hey, Steve. Thanks a lot for having me.

Steve Hutt: My pleasure. You’ve worked on the payments side and the identity side. What did those experiences teach you about fraud that you believe most merchants are missing right now?

Nicole Jass: Great question. I was at Worldpay on the product side for quite a long time and always got to work on what I called the fun products: identity, fraud, and loyalty. Anything involving data, including the data that comes out of payments, what we could collect, and what we could do with it for our merchants. What I found fascinating about fraud, and you touched on it early, is that it’s a two-sided problem. If you lean too far in one direction, the other side kicks in. If you focus too heavily on stopping fraud and reducing chargebacks, chances are you’re turning away good customers. On the flip side, if you’re focused on getting good customers in and approving everything, you open the floodgates for fraudsters. That’s what I found so fascinating about the problem: you really have to tackle both sides together.

Steve Hutt: Some of the terms that come to mind are false positives and false negatives. When you look at the risk profile inside the Shopify admin, for example, and a merchant decides to decline an order because it looks risky, how often do you believe that’s actually a good customer?

Nicole Jass: It’s a great question. I remember driving people nuts at Worldpay, asking, “What is the truth value? How do we find it?” It’s elusive. We’ve done some things here at Signifyd that we call forced approvals, and we can talk more about that. Based on studies we’ve done, we believe around 73% of declines are good people. That’s a broad average, and it depends on the rules, platforms, or AI you’re using. The real goal should be reducing that number while also reducing chargebacks.

Steve Hutt: Very interesting. From what I’ve read, fraud seems to be growing faster than ecommerce sales, even though ecommerce itself is growing. Harley Finkelstein at Shopify has commented on that growth, and I think Adobe has also talked about how large the market is going to be in 2026. But alongside that growth, there’s this risk problem with fraudsters. Can you talk about what’s driving fraud to grow faster than ecommerce sales?

Nicole Jass: In our latest State of Fraud Report, the numbers we used showed ecommerce growing just over 8% year over year on a same-store basis.

Steve Hutt: Yeah.

Nicole Jass: Fraud pressure is growing by 38%. As you mentioned, fraud is really outpacing the organic growth of ecommerce. It’s probably no surprise that we strongly believe AI is at the core of that. It’s not that AI has a mind of its own and is out there asking, “How can I defraud this company?” It’s fraudsters using AI to scale themselves. AI puts what used to be sophisticated, organized fraud-ring methods at the fingertips of amateurs. They can use it to spoof a website or carry out card testing attacks quickly and at scale. When it’s so easy to use, even if they get one out of 1,000 attempts right, there’s still a payoff.

Steve Hutt: That’s interesting, especially the card testing attack. I understand the term, but I’m not familiar with the technical side of what people are using to do this. What’s going on? What do you see from the inside when these attacks happen to a brand?

Nicole Jass: It often starts with a breach, and a fraudster gets their hands on a bunch of cards. They might be the person who finds the cards, or they might buy them on the dark web. Then they need to figure out which of those million card numbers are still active and have money they can steal. They go into a checkout and try a low-dollar transaction. One large customer of ours had a $1 donation option on their page. A fraudster could cycle through cards using that donation, hoping the transactions would fly below the radar. If a transaction goes through, they know it’s a good card. We also see this through account takeover. Someone takes over a dormant or inactive account and tries to load a card. When a card is added to a wallet or a logged-in account, the merchant often checks with the bank to make sure it’s valid. That tells the fraudster, “I was able to load this card, so it must be good. This other card wouldn’t load, so it must be bad.” Card testing can come in different forms. It can happen at checkout, when adding a card, or during an account modification.

Steve Hutt: That’s interesting because fraud used to be thought of mainly as a checkout problem. You’re suggesting it’s showing up elsewhere in the customer journey.

Nicole Jass: Yes. We’re seeing it with account takeover, as I mentioned.

Steve Hutt: Yeah.

Nicole Jass: It can happen further up the funnel. It could involve creating false accounts or logging in as someone else to test cards or use their loyalty points. We all know loyalty points are a valuable currency and drive a lot of behavior. We’re also seeing it at the other end of the journey, with returns. If someone can’t manipulate checkout, they may manipulate returns and take advantage of generous return policies.

Steve Hutt: Right.

Nicole Jass: That could mean shipping rocks in a box and getting a refund, shipping back old jeans, or asking for some kind of appeasement. We’re definitely seeing it across the funnel.

Steve Hutt: Most people listening will have a Shopify-powered business. I’d say most are probably on Shopify Payments, unless they couldn’t get approved because they’re in a regulated market, such as cannabis, CBD, or alcohol. Those businesses may use third-party gateways, which is fine. We’ll talk about that in a few minutes. Most probably have Shop Pay turned on as their accelerated checkout, which seems to work well. Correct me if I’m wrong, but American consumers buying from American companies can use the built-in Shopify Protect feature. I know that’s different from the risk profile Shopify provides when you’re deciding whether to ship an order. I’m trying to contrast the native Shopify solutions with what your company does. Can you talk about the relationship between Shopify and Stripe for approving orders, along with Shopify’s internal screening? That seems like a good starting point for some merchants, but I’d like to contrast it with Signifyd.

Nicole Jass: Sure. You’re starting to draw the distinction between when someone is a good candidate for a Signifyd solution and when they can use the built-in Shopify products. When you’re getting started, your concerns are getting customers, growing the business, and managing inventory. To put a number on it, if you’re doing less than roughly $5 million in annual sales, your fraud needs are probably best served by Shopify Protect and the built-in rules, without bringing on anything else. As you get bigger and scale, manual transaction reviews or chargebacks can become a significant enough problem that you need more sophisticated solutions. You need to battle AI with AI, so to speak. That’s when you start growing into Signifyd. The other problem is much harder to identify: declining too many good customers.

Steve Hutt: Yeah.

Nicole Jass: That can show up in approval rates or customer calls asking, “Why am I getting declined? Why is my transaction getting declined?”

Steve Hutt: Right.

Nicole Jass: Those are symptoms of the sneakier, but sometimes bigger, problem of turning away too many good customers.

Steve Hutt: I recently learned more about why I prefaced Shopify Protect by talking about American businesses and American consumers. That’s where its coverage applies. The challenge is that many American brands ship to Canada, Europe, South America, and elsewhere internationally. Those shipments aren’t covered by that service. Having worked inside Shopify, I’ve seen the fraud and chargebacks that can happen on international shipments. Sometimes products are going to PO boxes or addresses that have a name but no number. Some delivery addresses are unusual, and it can be hard to verify whether the customer received the product. Then someone says, “I never got it,” even though they probably did, and it turns into a chargeback. It can go down an ugly path, and Shopify Protect doesn’t cover those international shipments.

Nicole Jass: It’s almost as if the bad guys know where the boundaries are, right?

Steve Hutt: They know. They definitely know.

Nicole Jass: They know exactly where those boundaries are. You bring up a great point. There are limitations, so it’s not just about how much business you’re doing. The type of business you’re doing may mean you’ve outgrown or moved beyond what Shopify Protect can cover. Cross-border commerce is definitely part of that.

Steve Hutt: You brought up AI, and we obviously want to talk about it. It’s the elephant in the room right now. I think a lot of AI-assisted conversions are happening, but an AI agent actually making a purchase is still rare. Many people don’t trust it yet, although some transactions are happening. When an AI agent places an order on behalf of a shopper, how does a merchant tell a good bot from a bad bot? I don’t know what I’d look for in the admin or how Signifyd identifies the origin of that transaction.

Nicole Jass: That’s a really good question, and you’ve identified the core problem. A few years ago, a bot was bad. If you saw something crawling your system that didn’t feel human, you’d stop it, decline it, and get it out of your system. Now, when you see a bot, you have to ask, “Is this Steve’s agent coming here to buy something, or is it a bad bot?” That’s how the problem is changing. We’ve categorized what we call agentic commerce into three broad buckets. The most common use of AI in the shopping journey right now is discovery. I recently set up a fish tank and was having problems with either not enough algae or too much algae. I could go to Gemini to troubleshoot, and it would suggest a product that might help. Then I’d ask, “Great, where can I buy it?”

Steve Hutt: Right.

Nicole Jass: In that scenario, the agent is helping me with discovery. The purchase itself still happens on the ecommerce website. The referral and discovery are different. I might land on a site, go straight to a product, and put it in my cart because discovery happened through my agent rather than on the website. But it’s not that different. I’m still a human on a site. The next step is saying, “Great, I need that algae treatment. Go ahead and buy it and have it shipped to my house.” When I send an agent to do that, there are really two approaches. It depends on the merchant’s setup, and we’re still early in this process. One approach is that the agent scrapes the site as a bot and puts items in the cart. We’re getting certain header information and other signals to help us understand whether it’s a good agent, whether it came from a trusted place, and whether it’s authenticated. The direction I’m more excited about is when I ask Gemini to buy something for me and it goes through the Universal Commerce Protocol, which is an API. Then you have APIs talking to APIs, and machines know how to do that. That’s their language. What’s great is that Gemini can send the merchant more information: “This is Nicole. I trust her. Here’s her IP address and her device signals.” It could also pass a risk score to help make it a more trusted transaction. When people hear about UCP, ACP, or these other commerce protocols, they’re essentially APIs that let machines talk to each other and pass more information. When we send machines to scrape a website and act like humans, things get more complicated.

Steve Hutt: It’s interesting. I recently read about Meta and Shopify starting to work together, with Meta having a dedicated bot that I think is called Muse. OpenAI and Google are doing similar things, and there are lots of acronyms for how these machines communicate. I don’t think it’s completely fleshed out yet because four or five major billion-dollar or trillion-dollar companies are trying to work this out. It’s a big opportunity, and the question is which protocol people will accept as the standard. There might be two. Amazon has been pushing back hard, keeping things internal and pushing back against scraping. We’re living in an interesting time. Bots are scraping and doing all kinds of things, but they’re also making purchases, including fraudulent ones. We don’t know where the dust will settle.

Nicole Jass: That’s a great point. Right now, the volume is minimal. It’s less than a percentage point, or at least in the very low percentages of overall traffic.

Steve Hutt: Right.

Nicole Jass: It’s coming, though. You can tell from your own shopping experiences where it would be helpful to have something transact on your behalf and what it would take to make that happen. But you’re right: it’s the Wild West. Fortunately for everyone involved, the volumes are still low. As those volumes grow, it’s a chicken-and-egg situation. Some of this needs to get sorted out, including which commerce protocols will win, before we can have broader adoption. It’s definitely an exciting time in the space.

Steve Hutt: I want to go back to false declines because I think there’s an interesting opportunity there. Traffic is increasing as we move into the holiday period, and the Super Bowl of commerce is on its way. I saw some figures, possibly in your fraud report or elsewhere, about consumers’ holiday spending and when they spend. Many people have already started purchasing. A lot of shoppers want to buy early because the selection is good, and they’re happy with 20% or 30% off. They know they might get another 10% off during Black Friday or Cyber Monday, but selection could be an issue by then. With traffic increasing, Signifyd seems to be in a unique position. We’re always trying to drive new customers and improve blended ROAS, our return on ad spend. Brands are working on their ad creative and acquisition. But from your perspective, if we can approve and convert more orders without taking on that risk, the net ROI seems significant just from turning on Signifyd.

Nicole Jass: You bring up some good points. Around the holiday season, we see a huge increase in spending across the board, and it becomes an important time. You nailed the point about new customers. Now that you mention it, I made my first Christmas purchase yesterday. It’s a time when I’m trying new sites, picking things up, and exploring places I haven’t shopped before. It becomes a moment of truth: is this a good experience that I’d repeat and tell my friends about, or is it a bad experience where I’m one and done? You want to say yes to as many good customers as possible. Signifyd’s superpower in that scenario is that they’re new to you, but they’re not new to us. In 99% of transactions, we’ve seen some element before in our commerce network. We’ve seen the credit card, device ID, or email address, and we can start linking it to good or bad behavior. It’s not as scary to us because we’ve seen it before. A merchant might be thinking, “This is the first time I’ve seen this customer. It’s a really big purchase. Do I trust it?” We’re bringing that commerce network together at that moment of trust.

Steve Hutt: What about the chargebacks that can happen? Every year around this time, I worry about January, when credit card statements start coming in. A lot of orders get approved and shipped, but then the returns start. I used to work in retail, and we called it National Return Day.

Nicole Jass: Yeah.

Steve Hutt: I worked for the Best Buy organization as a commissioned salesperson on the floor. There was a return deadline, perhaps January 7, and a particular day we called National Return Day. You’d end up with negative commission because a large portion of the revenue you’d generated in November and December came back. You’d spend the day trying to sell enough new products to offset the returns and get back to zero commission. How do you think about the online version of National Return Day in January, and how does Signifyd protect merchants?

Nicole Jass: Part of it is distinguishing between a good return and a bad return. I consider myself a good shopper, and I definitely make returns. Things don’t fit, or you change your mind. I find myself shopping more with companies that offer great return policies: extended returns, no questions asked, and no charge to return something. That hassle-free policy can create a big increase in shopping. I think we’ve seen studies showing a 30% increase in loyalty and lifetime value when you offer things like instant refunds or make returns easier for good returners. But it’s okay to introduce friction for the not-so-good returns. There’s fraud in the system, and there’s abuse. We draw a distinction between them. Fraud might be sending rocks in a box. Abuse is taking advantage of a policy. One thing we do with merchants is identify good returns and where they want to drive that increase. If I have a $50 item to return and you make the process friction-free, maybe I spend $300 that day in the store or online. We help extend those good policies while identifying where to add friction because there’s fraud or abuse.

Steve Hutt: A lot of brands use returns tools. I don’t know exactly how Signifyd works with those partners, but many mid-market and enterprise brands have a returns solution connected to Shopify, such as Loop, Narvar, Happy Returns, or Returnly. There are several doing great things. What’s your approach to connecting with those partners? Do the systems work together? I think you have a returns assurance option within the platform.

Nicole Jass: It’s important to know what we’re good at and what we’re not. We’re not a returns logistics company. Some of the companies you mentioned are fantastic and very necessary for tracking shipments, letting customers know when something will arrive, and providing a platform to process returns. That’s not us. What we can help with, using our network, is deciding whether you should offer an instant refund or an appeasement to a customer we’ve seen across the network, or whether you shouldn’t. We might say, “For this customer, you should make sure the item gets back to your warehouse before issuing a refund so you can verify that it’s legitimate.” We connect with some of those platforms to provide that decisioning. Our job is to help make those tough calls and provide decisions. We’re not the user experience or the logistics platform.

Steve Hutt: That’s perfectly clear. You also have a Shopify app. Can you talk about that? People may want to explore it themselves now that they understand what you do. What’s the onboarding process when they go to the Shopify App Store, download the app, and install it in their admin? Is there an ingestion period? What’s happening in the background, and what gets turned on or off that a brand should know about?

Nicole Jass: That’s a fun part of the conversation when we’re talking to a merchant on Shopify. They’re asking, “How many developers do I need?”

Steve Hutt: I know, right?

Nicole Jass: “How long is this going to take to integrate?” And we’re saying, “You’re on Shopify? Just download the app.” It really is that simple. We’ve had merchants go live in days. We can also run proofs of value that way if someone wants to install us, turn us on, and send some traffic so we can evaluate it. That’s what’s great about our partnership with Shopify. The data we need starts being collected, and we can turn our services on fairly quickly.

Steve Hutt: Perfect. It’s nice to have a public app, and you have lots of five-star reviews, so people are clearly seeing value in adding it to their systems. My next question is about case studies. There are plenty on the website, and perhaps some internal examples where you can’t name the brand for confidentiality reasons. Are there any notable Shopify businesses you can talk about, named or unnamed? What was life like before Signifyd, and how did it change after they activated it?

Nicole Jass: Sure. I have a couple I can name.

Steve Hutt: Okay.

Nicole Jass: One of the core ways we help is by improving approval rates and decreasing false declines, as we discussed. The other part of our business model that I love is that we guarantee our decisions. If you get a fraud chargeback, we’re on the hook for it. We take full liability for those fraud chargebacks. For one apparel brand, Carbon38, we helped them reach a 99% approval rate. That’s saying yes to 99% of customers coming through, which translated into a 3% revenue boost. Those are real dollars in their pocket, along with happier customers and better customer experiences. That doesn’t even account for lifetime value. Across the board, we often see opportunities to increase top-line revenue by 3% to 5%. On the other side, I spoke with a Shopify customer last year who had a full-time person doing manual reviews. They were overwhelmed and had a long queue, which meant customers were waiting for decisions on their orders.

Steve Hutt: Right.

Nicole Jass: We were able to help them. That person now gets to do something more enjoyable than manually reviewing transactions under a tight service-level agreement. We’ve helped them improve approval rates by multiple percentage points and taken away their fraud risk.

Steve Hutt: That’s amazing. How is the product priced? My thought is that getting an order approved when it’s in that middle-ground or higher-risk category, and you otherwise wouldn’t approve it, could be valuable. There are two sides to this. Sometimes you’re desperate for sales and accept an order you know is risky. You have a gut feeling, and unfortunately, it turns into a chargeback. On the other hand, you might decline a legitimate order because it looks marginal or high-risk, even though your system could approve it. Paying a small amount to get those medium- to high-risk orders approved and protected seems like a no-brainer. I don’t know what the financial obligation looks like for a brand working with Signifyd. Is it related to GMV, order value, or basket size?

Nicole Jass: We charge in basis points, so it’s a fraction of a percent of the order value. What’s great is that it applies only to orders that are fully approved and authorized. If you don’t ship the product and it isn’t going to turn into revenue, we don’t charge for it.

Steve Hutt: I see.

Nicole Jass: Part of what we’re doing is covering that transaction. If you have a $50 order in that gray zone, we approve it, the bank authorizes it, and you ship it, then we charge a percentage of the order value for coverage.

Steve Hutt: Right.

Nicole Jass: If it comes back as a chargeback, you’d otherwise have to repay that $50, usually pay a $15 to $25 chargeback fee, and have someone spend time working on it. That’s covered by us through the fee you paid to Signifyd.

Steve Hutt: Let’s do a high-level overview of the 2026 State of Fraud Report. I’ll include the link in the show notes. I downloaded it before recording, and there are some interesting trends. What are the high-level takeaways you’ve found in the report?

Nicole Jass: I found it fascinating. We’ve already talked about one of the key themes, which is AI.

Steve Hutt: Yeah.

Nicole Jass: I think I read the report and then went and changed some passwords.

Steve Hutt: Yeah, right.

Nicole Jass: There’s real urgency around what we need to do to protect ourselves as shoppers and to protect merchants. That’s what you’ll find in the report, along with real stories and links to videos where customers talk about their experiences. That includes things like using AI to manipulate photos to make it look as though an item arrived broken. It’s not just a collection of statistics. You’ll see real people talking about the problems these activities are causing for their businesses and what they’re doing about them.

Steve Hutt: That’s great timing with traffic increasing. There are buyers out there right now, and merchants need to understand what’s happening. Unfortunately, some of it is AI-fueled. There are also great things artificial intelligence is doing for businesses and people. You mentioned agentic commerce and how AI changes the discovery journey to a brand. That’s why listicles are so popular right now. These tools are pulling information from top-ten review sites, citing them in their models, and bringing traffic over. That seems to be where we are right now, even if we’re not yet at the point of widespread purchasing between two computers. What do you think the next steps should be? There’s clearly a sweet spot for merchants who can get the most value from the platform. It’s not always about complexity, maturity, or GMV, although there probably is a revenue threshold where the value becomes measurable. People listening are at different points in their journeys. They might be marketing directors or operations professionals. Who is the best fit for the platform?

Nicole Jass: If you’re a growth company doing at least $5 million in annual revenue, and probably measuring your business in the tens of millions, you may be seeing manual reviews or chargeback recovery become a problem. Your approval rates might look low, and you may wonder whether you’re declining good customers. I’d suggest visiting our website. Check out the State of Fraud Report, look at what we’re seeing in the market, and see whether those trends match your experience. It’s also easy to get in touch with us. One of the first things we do with customers is an ROI analysis. We’ll collect some metrics and tell you directly, “We think we can help you generate this much additional revenue and save this much money.” Then you can make an educated decision about whether it’s worth giving us a try.

Steve Hutt: Do you find that the main user in an organization is an operations manager, or are there other titles and roles that get the most impact from the platform?

Nicole Jass: The users are typically on the fraud team, if there is one, the payments team, or the operations team.

Steve Hutt: Yeah.

Nicole Jass: Often, the person who benefits most doesn’t realize this is an important decision for them. That’s whoever is responsible for conversion and the funnel. The ecommerce or marketing person benefits because they’re working hard to bring people to the site, get them to add items to the cart, and hit the submit button. If you’re falsely declining good customers, that can be a hidden reason why conversion isn’t leading to revenue. The marketing or ecommerce person may care a lot about that, even though the user or buyer is usually on the operations side.

Steve Hutt: That’s amazing. I want to give you some credit because a lot of world-class brands have chosen Signifyd. Vuori is one of my favorites. I’m wearing their pants right now, and I love their Meta Pants. You also have brands such as Walmart, Lenovo, Mango, and Lacoste. There are other notable companies trying to solve a similar problem, but these large enterprises have chosen your platform. That’s a big checkmark for me because you’ve matured enough to serve companies on that scale. You also have a dedicated Shopify app. Getting an app approved and connecting to Shopify through a third party involves a lot of requirements around data and privacy, and you’ve worked through those hurdles. I commend the Signifyd team, and thank you for coming on the show today. We’ll include a link to the State of Fraud Report in the show notes. For anyone listening who’s north of that $5 million range, I think the best next step is to talk to an expert. They can click the button on your website, and I’ll also include a link in the show notes. Would you agree that’s the best next step?

Nicole Jass: Yes, that sounds great.

Steve Hutt: Good stuff. Nicole, thank you so much for recording today.

Nicole Jass: Thank you.

Steve Hutt: Bye-bye.