Amazon Business hits sixty billion dollars in annualized gross sales
Brian was back this week after a trip to Ireland, and the ranking news segment picked up on Amazon’s biggest B2B disclosure in three years. Amazon Business announced it has reached a sixty billion dollar annualized gross sales run rate, up from the thirty five billion it reported in 2023. The company now serves more than eleven million organizations worldwide, with 1.8 million new organizations added in just the first half of 2026. Ninety seven of the Fortune 100 buy through the platform, and Amazon reports more than 500 million deliveries with a large share arriving same or next day through Prime Business.
Hoar said two figures stood out to him. “Ninety seven of the top one hundred companies essentially in the world have Amazon accounts that they’re actively using, probably for long tail spend, and that long tail can creep into the short tail at a moment’s notice,” he said. He also pointed to the delivery scale: Amazon is running its own trucks at a volume that dwarfs most B2B distributors, “five and a half Graingers,” as Brian put it, comparing it to Grainger’s own reported revenue.
Both hosts were careful to frame this as a floor, not a ceiling. Brian noted the sixty billion figure likely captures only fifty to seventy percent of Amazon’s actual B2B volume, since many organizations buy through Amazon without a formal Amazon Business account. Hoar added that distributors have historically been able to compete on services and domain expertise that Amazon does not offer, but that edge is not permanent. “Domain expertise might have a crosshair on it,” he said, “because over time AI is going to get better at this.” With Amazon investing roughly two hundred billion dollars in AI this year, Hoar’s advice to distributors was to treat this as a moment to double down on the expertise that still requires a human relationship, before AI narrows that gap further.
Why platform investment keeps climbing even as ROI scrutiny rises
The hosts used Master B2B’s own research to set up the main topic. Fifty percent of practitioners surveyed said they are increasing B2B ecommerce tech spending this year, versus just six percent decreasing it. But thirty seven percent named demonstrating ROI on digital investment as their single biggest challenge in making commerce technology purchasing decisions, and thirty one percent identified the ecommerce platform itself as a top investment priority for 2026.
Hoar pushed back gently on how that platform number gets read. “It isn’t like the AI tools are separate from the ecommerce platforms,” he said. “A lot of the ecommerce platforms are using AI, so when you say we’re spending money on AI with air quotes and we’re not spending money on ecommerce platforms, that’s a contradiction in terms.” In his view, the real spending on platforms is understated because AI investment and platform investment increasingly overlap. He also flagged a smaller but real trend of teams building point solutions themselves with AI coding tools rather than buying software, though he stopped well short of calling it a wholesale shift away from SaaS.
What the Paradigm B2B Combine actually is
That backdrop set up the core of the episode: Andy’s eighth annual Paradigm B2B Combine, a from-scratch evaluation of B2B ecommerce platforms. Hoar was explicit that research is one input to a platform decision, not a substitute for the decision itself. “Research isn’t going to make the decision for you,” he said. “Any more than Dr. Google can cure your diseases. But who would go into a doctor’s office knowing you have a disease and not do research online, especially with AI?”
He described research as useful at multiple points in a selection process: building a long list, narrowing a shortlist, validating a choice between finalists, or breaking a tie at the very end. But it sits inside a broader process that starts with defining the differentiated experience a company wants three years out, then works backward to requirements, platform and partner selection, and implementation. Brian, drawing on twenty years running ecommerce and roughly three dozen consulting engagements on platform selection, agreed that the Combine adds something most research does not: “It goes to a double click level down on functionalities in a whole bunch of different areas that more traditional research doesn’t.”
Two editions, because a fifty million dollar company and a six billion dollar company need different partners
The Combine comes in a midmarket edition, for technology buyers at companies between fifty and five hundred million dollars in total revenue, and an enterprise edition for companies above that threshold. Hoar’s reasoning is organizational as much as technical. “If you’re going to be at a six billion dollar company, you need different partners, you need different support infrastructure, versus somebody at the fifty million dollar level might just have one guy on his team implement it,” he said. “You’re not going to have one guy on your team implement a new platform at, say, Grainger.”
This year’s midmarket edition includes twelve platforms, among them Virto Commerce, new to the assessment after qualifying for the first time. The enterprise edition largely overlaps but is not identical. Znode, historically evaluated only in the midmarket edition, asked to be assessed for enterprise this year as well, so it now appears in both editions with separate results for each.
How to read the Combine: the one-pager, the medals, and pricing
Hoar builds each vendor entry in three layers. A one-page summary distills the vendor down to bullet points for a busy reader. A second layer awards gold, silver, or bronze medals, or no medal, across twelve categories, with what Hoar says is 16.7 million possible medal combinations rather than a single composite score.
I don’t like composite scorers. I think that’s like the decathlon approach. The fastest guy in the world is not the guy who runs the hundred meter dash in the decathlon. It’s the guy that runs the hundred meter dash, period.
Andy Hoar
A third layer breaks out pricing, an area Hoar says is unusually hard to get vendors to disclose cleanly. “Good luck trying to figure out how much something costs,” he said. “Many of them are like, well, we’ll price it however the client wants to pay. I’m like, no, is it GMV? Is it number of order transactions?” Beck pointed out that this level of specificity, on categories like content and data management, marketplace capabilities, promotion management, native versus third party site search, and partner ecosystem strength, is what distinguishes the Combine from broader analyst research: “This gets into specific aspects of the solution that are more important to me as an ecom executive.”
Andy also stressed that the Combine evaluates a specific product, not a company as a whole. “If somebody was going to evaluate cars, they wouldn’t just evaluate General Motors. They would evaluate a specific car,” he said. He also noted that not every household-name vendor appears in the Combine, and not always by his choice. Some vendors declined to participate this year because they could not meet new AI-centric evaluation scenarios he introduced. “I had some scenarios this year that are required that were AI centric scenarios, and some of those companies contacted me and said, hey, we can’t really do this,” he said. “I’m not going to lower my standards.” His advice to buyers: if a familiar vendor is missing, ask them why, and read multiple research sources side by side rather than relying on any single one.
What changed in the 2026 edition
Andy flagged four shifts from last year’s Combine. First, “bring your own LLM,” where buyers ask vendors to plug in the buyer’s own model, such as ChatGPT or Claude, rather than the vendor’s built-in AI. Second, development cycles have compressed sharply, with vendor roadmaps that once took a year now landing in about three months, so a missing feature today is a much smaller obstacle than it used to be. Third, and Hoar was most emphatic on this point, data quality remains the gating factor regardless of how strong the platform itself is.
You can build this great plumbing system, but if the water going through it is sludge, you’re going to deliver sludge to every room of the house.
Andy Hoar
Fourth, the role of systems integrators and partners is in flux, as more buyers use AI to do configuration work themselves and push back on heavy, hard-to-upgrade customization in favor of configurable platforms.
On the broader “SaaS apocalypse” narrative, Andy was skeptical. “I’m not a believer in the SaaS apocalypse. I don’t think these SaaS companies are going to go away,” he said, pointing to how difficult complex, edge-case customization remains to replicate with point solutions. His caveat: buyers are far less patient than they used to be, and are not willing to be handed a toolkit and told to build a platform themselves.
Where to get the Combine
The Paradigm B2B Combine is available free at ParadigmB2B.com. Andy also previewed an interactive tool called the eCombine, which lets a buyer weight up to 28 criteria across the same 12 categories against their own requirements and get a matched shortlist by email. An update to the eCombine is expected next month, and Brian floated a future episode once it launches to walk through how to use it against the Combine’s platform list.