Signifyd says yes.

§ 01 — Who they are nowThe quiet referee at checkout.

Open any major retailer's cart page today and you may never see Signifyd's name. That is precisely the point. The company has spent fourteen years embedding itself into the part of the internet that nobody enjoys thinking about - the seconds between when a customer hits "place order" and when a warehouse begins to pick a box. In those seconds, somebody has to decide whether the order is real.

Signifyd makes that decision for a roster that now stretches across the Fortune 1000 and the Digital Commerce 360 Top 500: Samsung, Walmart, Lenovo, Lacoste, Rite Aid, Quiksilver, Mango, Ferguson. The company calls what it does "commerce protection," which sounds like a job for a bouncer. It is closer to a job for an actuary - one with very strong opinions and a checkbook.

§ 02 — The problem they sawFraud was the polite name.

The dirty secret of online retail, circa 2011, was that retailers were saying no to a startling amount of perfectly good business. False declines - legitimate customers turned away on suspicion of fraud - were quietly costing the industry billions of dollars a year, often several times what actual fraud cost. The math, as the industry liked to say, was complicated. The customer service was worse.

Signifyd's co-founders, Raj Ramanand and Mike Liberty, had spent their early careers staring at this problem from inside it. Ramanand led PayPal's Emerging Markets Risk division and, before that, spent eight years at FedEx running risk for payments and shipping. Liberty managed New Ventures Risk at PayPal. Between them they had a granular, slightly exhausted view of how the existing system worked: humans reviewing flagged orders one at a time, retailers absorbing chargebacks, fraudsters iterating faster than analysts.

§ 03 — The founders' betWhat if the vendor wrote the check?

Ramanand and Liberty quit PayPal, rented two desks at a Palo Alto coworking space, and started with what sounded, at the time, like a stunt: they would not just score risk for a retailer. They would underwrite it. If Signifyd approved a transaction and it turned out to be fraudulent, Signifyd - not the merchant - would eat the chargeback. The vendor would have skin in the game, in dollars.

It is the kind of idea that makes lawyers nervous and product people excited. It also creates an unusual incentive structure: the company is forced to be right. Not "directionally helpful." Right. Each false call has a price, paid by Signifyd, and that price gets tallied at the end of the quarter. Few B2B SaaS contracts include the words "we'll pay you if we're wrong."

The original pitch was even quirkier - the founders thought they could track fraudsters across social media, watching for the patterns that betrayed organized rings. The social-graph idea became part of the model, but the bigger product-market fit was the guarantee itself. Retailers, it turned out, did not want a probability score. They wanted permission, with a backstop.

§ 04 — The productFive layers, one promise.

The platform Signifyd sells today is broader than the original guarantee. It still ships under the marketing name "Commerce Protection," but it has spread, with the cheerful logic of a SaaS company, into adjacent corners of the buyer-seller relationship. There is Revenue Protection, which is the original chargeback guarantee, dressed up. There is Abuse Prevention, which polices the gray zone where shoppers exploit return policies, promo codes, and loyalty programs. There is Account Protection, which monitors login behavior for takeovers. And there is Payment Optimization, which exists mostly because European regulators decided that PSD2 and Strong Customer Authentication should make checkout harder.

Underneath all of it sits the Commerce Network - a shared identity graph built from billions of historical orders across Signifyd's customer base. When a stranger lands on a new merchant's checkout, Signifyd has, more often than not, seen them before. Maybe under a different name. Maybe from a different device. The Network knows.

§ 05 — The proofThe case for taking them seriously.

Companies that promise to pay the bill tend to be unusually transparent about the bill. Signifyd's customer roster is not boutique: multiple Walmart divisions, Samsung, Lenovo, Mango, Rite Aid, Quiksilver, Build with Ferguson, and Lacoste have publicly worked with them. The company processes orders for retailers in more than 100 countries.

The 2021 Series E - $205 million, led by Owl Rock Capital, with FIS, CPP Investments and Neuberger Berman alongside - carried Signifyd into unicorn territory at a $1.34 billion valuation. The round was less about ambition and more about timing: pandemic ecommerce had become permanent, and so had the fraud that came with it.

§ 06 — The missionSaying yes, professionally.

If you ask Signifyd executives what they actually do for a living, the language flips. They will not talk first about fraud. They will talk about commerce - the long, sloppy, occasionally beautiful exchange between people who want to buy things and people who want to sell them. Their stated mission is to create an open and trusted commerce experience for everyone. Their internal mission, you suspect, is more pragmatic: stop the false declines.

The fraud industry, traditionally, optimizes for caution. Signifyd optimizes for approval. It is a small philosophical shift that has large consequences for how the company recruits, trains models, and writes its product roadmap. The default question is not "is this dangerous?" It is "is there any reason to say no?"

That orientation is also why the company has spread into abuse and payment optimization. Every additional layer is another way to remove a "no" from the system. Every removed "no" is a transaction that would not have happened otherwise. It is, in a quiet way, a growth strategy that aligns the vendor's incentives with the retailer's revenue line - an alignment uncommon enough in B2B software to be worth pausing over.

§ 07 — Why it matters tomorrowThe AI part. The part everyone wants to talk about.

Generative AI is making fraud cheaper. Synthetic identities, deepfake KYC selfies, bot networks armed with stolen credentials at industrial scale - the cost curve for attackers has fallen and is still falling. Retailers will not solve this with bigger review teams. There aren't enough analysts in the world.

Signifyd's bet, fourteen years in, is that the answer is more data, faster decisions, and a vendor who is willing to be financially responsible for being wrong. The company's tech stack reflects that: Python, PySpark, Cassandra, Elasticsearch, Kubernetes, a heavy investment in machine learning infrastructure, and - more recently - the same large-language-model tooling that everyone else is also reaching for. The difference is the chassis. Signifyd has been doing machine-learning-driven fraud decisioning since well before "AI" became a marketing word, and it has fourteen years of labeled outcomes to train on.

§ 08 — Back to the scene2:47 a.m., revisited.

The stranger on the new phone, on the new Wi-Fi, buying the $1,400 sneakers - she's a real customer. Her old phone died yesterday. She is buying the shoes for her partner's birthday and she is, frankly, annoyed that the page is taking this long. The order goes through. Somewhere, a Signifyd model has scored her in 90 milliseconds against a graph that includes her shipping address, her email's behavior across other merchants, the device fingerprint of the new phone, and the historical pattern of late-night gift purchases in her zip code. The model says yes. Signifyd backs the call.

She gets the shoes. The merchant gets the revenue. The fraudster, in a basement somewhere, gets nothing. That is the entire business, and it is more interesting than it sounds.