AI Descriptions

Using AI generation to create concise and accurate product descriptions so users can quickly understand if a product is right for them.

Meta is an industry defining tech company with an estimated total market capitalisation of approximately $1.5 trillion. My role was within the Commerce organisation, with an estimated worth of $230 billion. During my time at Meta, I was a leading force in driving revenue, scaling teams, and improving internal processes, as well as leading content strategy for key high-impact features. One such opportunity I identified was to use cutting edge tech innovations to improve Shop features across multiple platforms, resulting in significant revenue growth.

CONCEPT & CREATION

Instagram and Facebook shop users, arriving from online ads, sought quick access to relevant product information. However, sellers provided product descriptions were often suboptimal for our platform - long, dense, not mobile-optimised, or overly focused on marketing content.

Having followed Generative AI trends, I saw an opportunity to address this issue when Meta released their AI tool, Llama.

PROMPT & CIRCUMSTANCE

I aimed to create concise summaries of our product descriptions, focusing on key information so users could make quick and confident purchase decisions. To achieve this, I created a project strategy and implement the following steps:

Created a project brief, recruited and led a team of Designers, Product Managers, Researchers, and Data Analysts and collaborated with C-Suite level stakeholders to set-up a Meta Hackathon project team.
Held research interviews with a range of partner commerce companies, major ones such as Nike, and smaller family-owned businesses.
Collaborated with Shopify stakeholders to obtain learnings from their user base.
Investigated product revenue data, used machine learning tools to analyse the content of product descriptions for top selling items and shops.
Ran user interviews and surveys.

I used this information to generate and test different AI prompts on a set of ten product descriptions, collaborating with legal and transparency teams to ensure compliance.

This resulted in the following prompt...

RESULTS

After testing, iterating, and finalising on the AI prompt, I ran a 2-week experiment that led to the following results:

A significant increase in products added to Meta platforms.
Increased conversion through Meta commerce platforms.
A decrease in business support customer service tickets.
A significant reduction in the time it took sellers to add products to Meta commerce platforms.
An increase in the retention of commerce business accounts.

Conclusion: by making it faster and simpler to add products to Meta, we were able to increase the number of products offered, and sold, through the Meta commerce platforms. The high quality of the AI product descriptions also contributed to increased sales. Both of these factors had a bonus benefit of strengthening the satisfaction and trust of our sellers, increasing retention and improving relationships with a vast majority of our partners, both major companies and smaller family ran businesses.