Scot Wingo observed consumers’ shopping journeys while running ChannelAdvisor, the marketplace management firm he started in 2001. He says consumers approach the process in three stages: researching the market, finding suitable products, and buying the right item.
The acronym — ReFiBuy — is the name of his latest company. It’s a generative AI optimization platform for retailers and brands.
By any measure, Scot is an ecommerce pioneer. We first interviewed him in 2006, when he introduced us to marketplace selling.
Last week, I asked him about ReFiBuy. The entire audio of our conversation is embedded below. The transcript is edited for length and clarity.
Kerry Murdock: Tell us about your ecommerce journey.
Scot Wingo: It began in 1999 when I launched Auction Rover, an auction search engine. We sold it to GoTo.com, which became Overture, the company that invented paid search. The auction search engine wasn’t great after the dot-com bubble burst. But we had built the selling tools, which became ChannelAdvisor, which I launched in 2001.
Murdock: Tell us about ReFiBuy, your new venture.
Wingo: The idea started with my experience at ChannelAdvisor. The company went public in 2013, and I was still CEO and founder. By 2015, running a public company had become a drag.
I left the CEO role but stayed on the board. ChannelAdvisor was ultimately taken private by a private equity firm, which merged it with Commerce Hub. It’s now called Rithum.
So I left ChannelAdvisor in 2015 and launched an on-demand car care company called Spiffy. Then, in August of 2024, I decided to start what is now ReFiBuy. I wanted to do something in the AI world. I have a technical degree, and as a technologist, I thought AI would create much disruption, which creates opportunity.
So I was poking around, learning more about it. And then, in December 2024, Anthropic, the makers of Claude, published a paper on “agentic” AI that can perform tasks. Prior to that, large language models were read-only. The agentic component meant they could do things.
And that reminded me of a problem we had at ChannelAdvisor. Our clients were retailers and brands with large product catalogs. The issue for us was the absence of an industry standard for electronically storing and sharing the product info, such as specs, colors, dimensions, and weight.
Clients would send us a file of their product catalog in a disorganized mess. Yet we had a hundred marketplaces that wanted to receive beautiful, clean catalog files. So our job became catalog cleaners, to convert clients’ inventory files into a format acceptable to those external channels. Again, there was no industry standard.
We came up with algorithms for cleaning the catalog that worked only half the time. The other half required humans. Eventually, when we had 300 people in Bulgaria working on it, serving our 3,000 customers and 15 billion annual transactions.
That memory was my light bulb moment for agentic AI. Could we solve the product catalog problem for LLMs? We started working on it late last year.
Simultaneously, Perplexity introduced what we now call agentic commerce, or agentic shopping, where you can not only research products but also buy them.
That’s the inspiration for our name. ReFiBuy is “research, find, buy.” It’s the shopper’s journey.
We launched our Commerce Intelligence Engine last week. It ensures that the LLMs — Perplexity, Claude, ChatGPT, and others — have accurate, current, and comprehensive product catalog data for our clients, which are retailers and brands.
Murdock: How do you do that — organize the data and then ensure the LLMs digest it?
Wingo: We start with the product catalog. We take a traditional Google Shopping feed or even data from a merchant’s ecommerce site. We analyze it through the lens of an LLM, which helps us identify missing or incorrect components. We then recommend changes, fixes, and additions. LLMs want every piece of content that ties products to the context of prompts. That includes Schema.org markup, Reddit discussions, prompt history — much more than product data alone.
That’s our evaluation phase. Then we help our clients whitelist the right bots to crawl their sites. Most retailers and brands block all bots except for Google. Certainly there are good reasons to do that, as many bots are malicious or from competitors.
So we help merchants know which LLM bots to allow.
Murdock: How do you know that an LLM receives and stores your optimized data?
Wingo: We monitor product cards, the visual representations by LLMs of recommended products. We run thousands of prompts daily across all the LLM engines to ensure our clients’ products appear in those cards and that the data is accurate.
Our AI agents evaluate the cards and classify into buckets. If our client owns the product card, our job is done. We have achieved Nirvana for that SKU. If our client’s item appears in a card of another merchant, there are 20 to 30 things that have likely gone wrong. Our AI agents detect it. Sometimes it’s as simple as a missing slash or an extra space in the file.
The agent also detects missing SKUs — when our clients’ goods don’t appear in the cards at all. That’s usually caused by an infrastructure problem with the crawler, or something is broken on the merchant’s site.
We keep cranking the process until we’ve optimized our clients’ entire catalog.
Murdock: What is the cost of ReFiBuy?
Wingo: It depends on the number of SKUs. We start at roughly $2,000 per month — $20,000 to $25,000 per year.
Murdock: Where can merchants learn more?
Wingo: We’re at ReFiBuy.ai. My Substack newsletter is “Retailgentic.”