Onboarding a brand’s product catalog to a new marketplace is typically slow and error-prone, but with generative AI, this process is being revolutionised, reducing the time from 80 days to just hours, while ensuring 95%+ accuracy in mapping products to marketplace taxonomies.
Onboarding a brand’s product catalogue onto a new marketplace is a time-consuming and often frustrating task. Successfully onboarding all the SKUs can stretch up to 80 days. The primary bottleneck is mapping the brand’s product data to the marketplace’s taxonomy, which can be a highly manual and error-prone process. Only 20% of SKUs typically pass the first round of upload; the rest require extensive human intervention to ensure that each product is categorized according to the retailer’s specific guidelines.
This slow, tedious process occurs because most taxonomies consist of thousands of categories and types, many of which are similar in name and description. One retailer reported that 70% of midmarket brands dropped out during onboarding due to the complexity of this process. The result is not only operational inefficiency but also a missed opportunity for brands to start selling sooner.
Taxonomy mapping involves translating a brand’s internal product data (like descriptions, attributes, and categories) into a format that aligns with the retailer's or marketplace’s product classification system. This process requires a detailed understanding of both systems and extensive human effort to ensure that every product fits neatly into the right category. Given that some taxonomies can contain 5,000+ categories, and many of them overlap in subtle ways, it’s a daunting task for even the most experienced human operators.
The result is a significant investment of time and labor—and, ultimately, frustration on both sides. For marketplaces, an inefficient onboarding process means lower conversion rates and fewer brands willing to commit to the platform. For brands, the extended onboarding time delays revenue generation and increases operational costs.
Enter Generative AI, specifically large language models (LLMs), which are now poised to revolutionize the way product data is mapped to marketplace taxonomies.
Handling unstructured data at speed: LLMs have the unique ability to handle unstructured data—the long-form text, descriptions, and other rich content that often populate product listings. Traditionally, this kind of data required extensive manual review, but AI models excel in parsing and interpreting this information quickly and accurately, essentially mimicking the cognitive process of a human but at much faster speeds. This allows them to handle much more of the catalogue-taxonomy mapping task than prior AI approaches.
Self-critiquing selections: LLMs can be trained to evaluate the relationships between various categories and types within a marketplace’s taxonomy. This means they can understand which product description corresponds to which category, even when the categories are worded similarly. Unlike earlier AI models, LLMs are not limited to rigid, predefined rules. Instead, they can adapt to the nuances of different marketplaces and continuously improve through self-critiquing. If an LLM’s initial mapping isn’t perfect, it can iteratively refine its decisions, much like a human would adjust their thinking based on feedback.
Thanks to their ability to process large volumes of product data and make intelligent mapping decisions, AI models can now achieve 95%+ accuracy in aligning products with a marketplace’s taxonomy. This reduces the typical onboarding time from 80 days to just a few hours—and in some cases, it’s as simple as a single click.
The speed and accuracy of this process not only accelerates the time to market for brands, but it also dramatically cuts down on the costs associated with human support. Retailers no longer need to rely on manual intervention for every product upload, significantly reducing labor costs.
In practice, these generative AI models need to be part of a larger integration (or suite of integrations) from a source to the target marketplace. Whether the product data is stored in a Product Information Management (PIM) system, a commerce platform, or a third-party marketplace, once embedded in the integration, AI can automatically map incoming products to the taxonomy.
Once integrated, these models take onboarding from a multi-month, multi-team process to a single click. The result is a more streamlined, efficient experience for both brands and marketplaces, with fewer drop-offs and faster time-to-market.
The advent of generative AI is fundamentally changing how product data is mapped to marketplace taxonomies. What once took months of manual effort can now be accomplished in hours with a high degree of accuracy. As more retailers and brands adopt AI-powered solutions, we can expect onboarding to become faster & more cost-effective.
For brands looking to scale across multiple marketplaces, generative AI is not just a game-changer—it’s the key to unlocking the full potential of their product catalogs, reducing friction, and speeding up the path to market success.
