Digitally native outdoor gear brand Cotopaxi has expanded its reach since its founding 12 years ago, now operating 22 physical stores, selling wholesale and through online marketplaces — and it’s now preparing its data to be able to sell through agentic AI platforms as well.
Part of that data preparation has already begun. To sell on multiple marketplaces, Cotopaxi has to optimize its product details for each channel. That means adding both image- and text-based attributes as needed to meet each marketplace’s requirements. And that same approach can apply to large language models (LLMs) such as Google’s Gemini and OpenAI’s ChatGPT.
Integrating with those agentic AI platforms requires Cotopaxi to make its product catalog machine-readable, focusing on whichever attributes the LLM requires of retailers.
Stephan Jacob, chief global officer and a co-founder at Cotopaxi, told Digital Commerce 360 that’s what enables contextual — versus static — discovery. Static discovery is keyword driven, whereas contextual discovery uses natural human language and phrasing (which can include full sentences) to describe products.
“We’re fully mindful and aware of the shift from traditional SEO, traditional product discovery, into agentic discovery and are embracing that in ensuring that the metadata that we surface is structured the right way, so that it can be ingested and reproduced accordingly in the different agentic environments,” Jacob told Digital Commerce 360.
He said Cotopaxi currently leverages partners to do that. It has not directly plugged into any of the LLMs so far.
Cotopaxi is No. 1,415 in the Top 2000 Database. The Top 2000 database is Digital Commerce 360’s ranking of the largest North American online retailers.
How Cotopaxi prepares for integrations with marketplaces vs. agentic AI
A key difference between preparing Cotopaxi product data for an LLM versus a marketplace, he said, is the type of search the retailer optimizes for. Preparing for LLMs requires more context-specific information, whereas marketplaces and product detail pages (PDPs) are more keyword-driven, Jacob said.
As such, Cotopaxi is enriching its data from keyword discovery to contextual discovery, Jacob said. That involves identifying use cases for products, “explained in human language as opposed to an algorithmic discovery tool,” he said. It also means ensuring Cotopaxi can provide back-end data, which the LLMs in turn can surface.
“It’s a lot of — guided through our providers — what are we missing currently in terms of these product feeds? And then ensuring that we enrich it accordingly, and actually leverage AI extensively to create some of that content,” Jacob said.
In part, Cotopaxi is using technology from Mirakl. Mirakl provides solutions that enable companies to create their own marketplaces or sell on existing ones, in addition to other ecommerce-related offerings.
Cotopaxi also uses Shopify to power its point-of-sale (POS) system and its ecommerce sales. In late March, Shopify debuted an integration with ChatGPT to make brands shoppable within the platform.
“We’re doing what needs to happen on our end to ensure our product feed is clean on that front and are proactively listening and learning every day,” Jacob said. “What else can we do? And what else are the opportunities to ensure that our products show up where they need to show up. It’s a learning process.”
Enhancing product data for different integrations
Scott Eckert, Mirakl CEO of the Americas, told Digital Commerce 360 that much like marketplaces, LLMs treat certain product catalog data as uniquely important. Each marketplace or LLM might value some attributes more than others, or require an attribute that another does not.
As Jacob put it, “Every marketplace is a little different, and their requirements are a little different.”
Mirakl’s process involves two interrelated steps, Eckert said: optimizing and enrichment. Optimizing is using Mirakl’s knowledge of what is uniquely important for each space where a retailer is connecting its product catalog.
“And, of course, that changes each time they release a new update,” Eckert told Digital Commerce 360. “So that is a very dynamic space, as the way the LLMs work is changing pretty rapidly.”
Enrichment is multifaceted. It involves making sure that for each product in a retailer’s catalog, the retailer has the right type of image and attributes extracted from that image. It also means using Mirakl’s proprietary system to scan the image and identify product details. Eckert gave an example of Mirakl’s system looking at a picture of a Cotopaxi backpack and autonomously labeling it as multicolored.
From there, it can help to build product descriptions “that are all AI-enabled, so the end result is a richer set of data than what a customer might see when they’re surfing the web and looking at a product data page,” Eckert said. “Behind that is actually a whole — that you don’t see because it’s designed to be machine readable, not human readable — is a whole rich set of data that’s been optimized for the different AI agents.”
Do you rank in our databases?
Submit your data and we’ll see where you fit in our next ranking update.
Sign up
Stay on top of the latest developments in the online retail industry. Sign up for a complimentary subscription to Digital Commerce 360 Retail News. Follow us on LinkedIn, TikTok, X (formerly Twitter), Facebook and YouTube. Be the first to know when Digital Commerce 360 publishes news content.
Favorite