As published at Beyond the Catalog.

The Generative AI Revolution Begins

GPT-1, rooted in the Transformer architecture introduced in 2017, was launched by OpenAI in 2018 as a groundbreaking proof of concept. Five years and four versions later, GPT-4 emerged in 2023, offering significant advancements in multilingual understanding, contextual coherence, and handling complex queries. In short, GPT models continue to evolve, becoming remarkably efficient and accurate.

Two aspects that make GPT models truly special are their unique abilities to engage in context-aware, human-like conversations and generate content. As processing costs decrease and new solutions are developed to enhance e-commerce applications, these AI capabilities are poised to revolutionize the industry. Consequently, both existing and emerging solutions are racing to leverage these technologies and play a pivotal role in transforming digital commerce.

To see how these capabilities will change ecommerce in the future, let’s review some of the problems traditional product discovery has suffered for decades.

Catalog Data Deficits

Having collaborated extensively with over 200 of the nation’s leading eCommerce retailers over the past decade, I’ve noticed a recurring challenge they often face: their product catalog. Specifically, the root of the issue lies in the product feed that is ingested and mapped to their catalog. It’s important to recognize that all product discovery begins with this data, and nearly all discovery-related issues stem from it. Ensuring the accuracy and efficiency of this initial step is crucial for a seamless product discovery experience.

Most catalogs unfortunately suffer from missing, inaccurate, or misattributed data, leading to subpar search results and recommendations. Given that many catalogs contain over a million products, addressing these data discrepancies is a complex task. To manage this challenge, most product discovery platforms have developed an array of manual merchandising tools to implement workarounds. This involves employing strategies such as keyword synonyms and redirects, product and category burying, blacklisting, and other techniques to mitigate these data-related issues.

Despite the time-consuming nature of manual workflows, they cannot address all issues, especially as catalogs are continuously updated or expanded, often on a daily basis. Additionally, customers may use keywords that don’t align with specific catalog attributes or search for brands that retailers do not carry. This often leads to the dreaded “No Search Results” page—something every retailer wants to avoid.

Imagine if there were a technology that could reach “beyond the catalog” to intuitively understand customer queries and seamlessly identify matching products within a retailers catalog.
Imagine an end to “No Search Results” pages. Imagine that!

Imagine no more. AI is now solving for decades old data deficiencies.

Retailers looking for better results, a better customer experience, and better performance are now implementing new solutions that solve the problems of traditional product search and discovery with AI.

Vector & Conversational Search

Vector search is an advanced search technology that moves “beyond the catalog” by offering a more intelligent approach to retrieving information. It excels at providing contextual and semantic relevance, capturing relationships between keywords based on their co-occurrence within data. This capability allows vector search to match catalog products to a customer’s query even in the absence of the exact keywords in the product catalog.

Vector search also supports the use case of conversational search. Conversational search refers to a type of search experience that aims to understand and engage in natural, human-like dialogue with users. Unlike traditional keyword-based search, conversational search is designed to interpret the user’s intent more deeply, respond in a way that feels natural, and maintain context throughout the interaction.

While vector search offers substantial benefits in terms of relevance and contextual understanding, its current processing demands and cost may limit its practicality as an immediate replacement for the faster traditional keyword-based search systems.

Catalog Enrichment

Traditional keyword-based search systems offer the distinct advantage of speed, as they utilize indexed catalog data to enable fast processing and retrieval of search results. The primary challenge with current search platforms lies in the data itself, rather than the search technology. To address this, some solutions are beginning to incorporate the concept of catalog enrichment.

In this approach, AI technologies are deployed to continuously modify and update product catalogs (typically daily upon ingestion). This enrichment process enhances the catalog data, allowing search systems to deliver more accurate and relevant results based on the improved catalog information.

AI is enriching product catalogs in various ways, including:

  • Data Gap Analysis: AI identifies missing or incomplete data fields and fills in these gaps, ensuring more comprehensive catalog information.
  • Attribute Extraction: AI extracts new product attributes from unstructured data, adding valuable details that enhance product descriptions, categories, and other descriptive attributes.
  • Attribute Synonyms: AI recognizes synonyms for key attribute values, improving search results by accounting for different terminologies that users might employ.
  • Data Normalization: AI standardizes product data formats across the catalog, ensuring consistency and easier data management.
  • Image Enrichment: AI enhances images with additional metadata to support image-based searches and visual recommendations, providing a richer visual catalog experience.
  • Multi-Language and Locale Support: AI enriches catalog data by translating key product attributes, tags, and descriptions into multiple languages and local dialects, catering to diverse global markets.

AI will soon revolutionize ecommerce. Are you ready for it?

The impact of these new technologies will be transformational. Soon, we will be able to engage in verbal dialogue with an AI chat “beyond the catalog” that will assist us in finding and purchasing the products we want. This is not a distant, utopian vision; the technology for these use cases is already available today. However, these changes will not happen overnight, as customers will need time to learn how to use and adapt to this new technology.

There is no question about it. Generative AI will revolutionize the shopping experience. And sooner than you might think. Are you ready for it?