For years, Scot Wingo observed consumers when buying channels, market management, which began in 2001.
APT shortcut – refibes – is the name of its latest society. It is a generative platform to optimize the engine for retailers and brands.
In any case, Scot is a pioneer of electronic trading. First we talked to him in 2006, when he introduced us on the sales market.
Last week I asked him about the refib. The whole sound of our conversation is inserted below. The transcription is modified for length and clarity.
Kerry Murdock: Tell us about your electronic trading journey.
Scott Wingo: It started in 1999, when I launched a auction search engine Rover. We sold it to Goto.com, which became a prelude, a company that invented paid search. Auction search engine was not great after bursting bubble dot-com. However, we built the sales tools that became Channeladvisor, which I launched in 2001.
Murdock: What is refib, your new business?
Wingo: The idea began with my experience in Channeladvisor. The company was published in 2013 and I was still the CEO and founder. Until 2015, the operation of a public company has become a drag.
I resigned to the role of the CEO, but I stayed on the board. Channeladvisor was eventually a private private company that merged it with the mushroom Commerce. Now it is called rithum.
In 2015, I founded a car care company on request called Spiffy. Then, in August 2024, I decided to start what was now refib. I wanted to do something in the AI world. I have a technical title and as a technologist I thought that AI would create a great disruption, creating an opportunity.
So I quarreled and learned more about it. And then, in December 2024, Anthropic, Claude creators, published a contribution on the “agent” artificial intelligence that can perform tasks. Previously, the large language models were only reading. The agency meant they could do things.
And that reminded me of the problem we had at Channeladvisor. Our clients were retailers and brands with large product catalogs. The problem for us was the absence of industrial standard for electronic storage and sharing of product information, such as specifications, colors, dimensions and weight.
Clients would send us a set of their product catalog in a disorganized mess. Nevertheless, we had a hundred markets that wanted to receive beautiful and clean catalog files. So our work has become the catalog cleaners and converted clients’ stocks to a format acceptable to these external channels. Again, there was no industry standard.
We came up with algorithms for cleaning the catalog, which only worked half the time. The other half required people. In the end, we had 300 people in Bulgaria who worked on it, served our 3,000 customers and 15 billion annual transactions.
The memory was my moment of bulb for AI. Could we solve the problem of the product catalog for LLMS? We started working on it at the end of last year.
At the same time, confusion has introduced what we are now calling an agent shop or agent, where you can not only buy products but also buy them.
This is an inspiration for our name. Refiby is “search, find, shop”. It’s a shopper’s journey.
Last week we launched our business news module. It ensures that LLMS – confusion, Claude, Chatgpt and others – have accurate, current and comprehensive product catalog data from our clients, which are retailers and brands.
Murdock: How do you do it – organize data and then ensure that LLMS spends them?
Wingo: We start the product catalog. We will take a traditional Google Source shopping or even data from the trading e -business website. We analyze this through the LLM lens, which helps us identify missing or incorrect components. Then we recommend changes, repairs and addition. LLMS wants every piece of content that combines products with the challenge context. This includes the marking of Schema.org, Reddit discussion, fast history – much more than the product data itself.
This is our assessment phase. Then we help our clients whitelist the right robots to go through their pages. Most retailers and brands block all robots except Google. Certainly there are good reasons to do because many robots are harmful or from competitors.
So we help traders know which LLM robots allow.
Murdock: How do you know that LLM receives and stores your optimized data?
Wingo: We monitor product cards, visual representation of LLMS recommended products. We run thousands of challenges daily in all LLM engines to ensure that our clients’ products appear on these cards and that the data is accurate.
Our AI agents evaluate cards and classify them in buckets. If our client owns the product card, our work is done. We have reached Nirvana for this SC. If the item of our client appears on the card of another merchant, there are 20 to 30 things that have probably gone wrong. Our AI agents detect it. Sometimes it’s as simple as a missing slash or other space in the file.
The agent also detects the missing fact – when the goods of our clients do not appear on the cards at all. This is usually due to the problem with the search with a looper or something broken is on the site of the merchant.
We continue the process until we optimized the entire catalog of our client.
Murdock: What are the cost of refibes?
Wingo: It depends on the number of SKU. We start about $ 2,000 per month – $ 20,000 to $ 25,000 per year.
Murdock: Where can traders learn more?
Wingo: We are in refibuy.ai. My Restack newsletter is “retail”.