NESPRESSO: Doubling Traffic and Tripling Visibility Online by Applying GenAI to Product Freed

Perfecting Product Feed Files for Improved E-commerce Visibility

Selling coffee capsules and machines worldwide, Nespresso wanted to improve its e-commerce visibility with a particular focus on product feed. Driving online visibility for a brand’s product, a product feed is a file containing all the details about an item’s properties and attributes to display for shoppers. 

In Nespresso’s case, issues within product feed files meant that search engines would sometimes “misread” prices (e.g., displaying 4,40 euros instead of 0,44 euros for an item price), show competitors first in search results, or even forgo product descriptions altogether.

Building a Custom GenAI Flow to Harmonize, Fix, and Enrich Product Feed Files

Nespresso and fifty-five set out to overcome three specific product feed challenges:

• Adapting to Google’s platform guidelines
• Ensure conformity with Nespresso’s guidelines
• Correct and enrich product descriptions

While the improvements obtained would be individually slight, they can lead to substantial gains down the line when combined.

To achieve the desired results, fifty-five relied on its in-house tool, SyMap, which can combine functions from various business-grade LLMs to create custom flows.

Greatly Improved Visibility and Traffic

Focusing on a specifically scoped PoC (proof of concept), fifty-five combined functionalities from various GenAI models to form a product feed expert custom flow on SyMap. This PoC was built to generate a best-in-class product feed catalog, fixing issues that hindered visibility in Google Merchant Center.

Pre-prompting input by fifty-five combined best-in-class product feed catalog requirements, guidelines from Google Merchant Center, and guidelines from Nespresso. The SyMap flow was also calibrated to identify a list of issues to fix according to product type, generate format suggestions, and justify and evaluate its own output. The resulting product feed was both corrected and enriched to ensure optimal visibility as well as accuracy.

The PoC's new product catalog was subsequently pushed to Google Merchant Center. We ran test campaigns, then measured before and after variations on impressions in CTR. While PoC results should always be read with caution, as their scope is limited by design, the results were already promising, with x3 visibility/impressions and x2 traffic.