Preparing Retailers for ChatGPT – Structure product relationships such as “Complete the Look”

In this post, we address the importance of structuring product relationships for LLMs to align with how consumers reason, enter their consideration phase, and buy online.

This is part 2 in the ChatGPT series for retailers and B2B's. Click here to see Part 1, which talks about the importance of embedding customer reviews into product master data.

LLMs like ChatGPT are designed to include product relationships, such as cross-sells and substitutions, in their generative product recommendations for consumers who intend to purchase a product.

If this type of product data is appropriately structured, it's considered a "first-class signal" to LLMs, meaning it doesn't need to guess from other content sources. For obvious reasons, LLMs favour first-class signals.

LLMs always want to offer alternatives if the primary choice is not quite right, is out of stock, or is outside the shopper’s consideration set for whatever reason (for example, it may be too expensive).

Other examples of product relationships that LLMs are looking for…

• “Similar to”
• “Accessory of”
• “Works with”
• “Frequently bought together”

This creates a powerful cross-sell and substitution logic that LLMs can use to meet a wide array of consumer needs. And retailers who empower LLMs with this type of first-class signalled data will be recommended more often than those that do not.

Why are these first-class signals favoured?

This type of data enables the LLM to keep the conversation moving. LLMs want to hold on to the consumer as much as any other selling system. If a product is unavailable, too expensive, or mismatched, proposing similar alternatives is how LLMs become "sticky" and enable engagement.

The same old best practice UX/UI principles also apply to LLMs. The longer you can hold on (engage) with a consumer when he/she have buying intent, the greater the likelihood they purchase.

The THREE BIG takeaways from this are…

Tip #1 - Lose the "Notes":

If you have product relationships in your database “Notes” or free-text fields, this is no longer sufficient. Activate a plan to embed this relationship information into your PIM structurally! If you don't use a PIM we need to talk:)

Tip #2 - Don't Rely on your eCommerce Platform:

If you rely on your eCommerce platform to define these product relationships (such as "Complete the Look"), LLMs prefer structured data over dynamic logic. The recommendation is to use a hybrid approach.

You can activate relationship automation for the low-risk relationships, but be strategic in manually curating meaningful product relationships. The risk is that automated rules may not add value to consumers.

Tip #3 - Be Empathetic:

The strategic planning behind enabling LLMs (like ChatGPT) to sell your products is the same as planning your own eCommerce channel. You MUST take an empathetic view of what's essential to your customer in the consideration phase of online buying.