A few weeks ago, we hosted our second annual Customer Meetup in New York City, where we had the honor of connecting and brainstorming with knowledgeable industry experts and our customers on various topics surrounding e-commerce success.
A key moment from the event was the keynote address by Brian Dhatt, CTO of BigCommerce. He walked us through his career journey and extensive experience in online retail, highlighting the evolving industry trends that retailers need to keep up with.
In our previous posts, we discussed the first two segments of his presentation—the role of decision theory in e-commerce pricing optimization and the potential of passwordless one-click checkouts.
In this post, we will break down his thoughts and insights on the last segment of his address – the growing importance of LLMs (such as ChatGPT) in content discovery.
The Impact of LLMs on Online Shopping Behavior
LLMs, or large language models, comprehend and process vast amounts of information to generate human-like text. ChatGPT is an AI-powered LLM that understands data to have human-like conversations and make predictions.
In e-commerce, LLMs enhance customer support, automate content creation (such as writing product descriptions), and aid product discovery. Brian emphasizes that GPTs might sound complicated but at the very basic level, they have a statistical model for predicting the next word in a customer query. Take the following example:
In the above example, the model has predicted that there is a 97% chance that the next word in the sentence – “cat sat on a..” is mat. Essentially, GPTs build a statistical model that excels at deducing the next word in a search query based on context.
Understanding Temperature in GPTs
To ensure data freshness, real-time access, and high ranking, OpenAI has begun licensing data, including a partnership with Stack Overflow and Reddit. This means OpenAI uses these platforms’ real-time data to enhance its ability to answer questions accurately.
How does ChatGPT get the data it needs?
To ensure data freshness, real-time access, and high ranking, OpenAI has now begun licensing their data including a partnership with Stack Overflow and Reddit. This means that OpenAI will now use Stack Overflow and Reddit’s suite of tools to enhance its ability to answer questions using these platforms’ real-time data.
Optimizing Content for Large Language Models (LLMs)
In a world that is now ruled by ChatGPT, Brian recommends investing heavily in optimizing your e-commerce content for LLMs, just like we all once did for SEO. Here are some tips here shared:
- Prioritize websites that are known to be scraped by LLMs
- Ensure product descriptions are clear, concise, and informative
- Include relevant keywords in titles and throughout the content
- Use structured data to help LLMs understand content
- Monitor LLM performance to identify areas for improvement
Brian recommends moving away from a keyword-first approach to more of a question-answer format and relying on more structured/hierarchical data to get your content captured by LLMs.
It’s almost important to note that LLMs update their models once every few months, so it can take a while to see the intended results. Therefore, Brian suggests investing early and monitoring the results regularly over long periods of time to seed your data into various LLMs. The first mover advantage that you can get is similar to the early days of SEO where companies that invested early on could reserve a spot on the first page of Google.
Brian shared some actionable and invaluable insights as part of his keynote presentation. This was just one segment of Brian’s keynote address. Follow the Noibu blog for more insights from his talk and other industry trends and updates.
