Estée Lauder How AI-Driven Feed Optimisation Delivered 178% Revenue Growth Across 8 Brands

“As AI-driven search continues to evolve, we knew our product feeds needed to work harder and smarter. The Sparro team helped us move beyond manual optimisation and build a scalable, future-ready approach that is delivering meaningful commercial results across multiple brands. The performance uplift speaks to both the strength of the model and our partnership. It’s an important step forward in modernising our organic shopping strategy.” – Kate Gildea, Marketing & Communications Director, Australia & New Zealand

A scalable approach combining AI automation and manual QA to transform Organic Shopping performance for nearly 4,000 SKUs in just under 3 months. 

THE CHALLENGE

Manual optimisation was no longer sustainable at scale

AI-driven search and conversational shopping continue to transform the way consumers discover products. Visibility now depends on more than traditional SEO; product feeds need to be accurate, detailed, and comprehensive enough for AI platforms to correctly interpret and recommend products.

With nearly 4,000 SKUs across 8 brands, each product required unique and search-optimised titles, detailed descriptions, and a thorough attribute audit. Managing this at an individual level would have significantly slowed delivery and limited the ability to scale.

OUR APPROACH

A 4-phase system combining AI automation and manual QA

The system prioritises high-impact products, detects feed issues, optimises content using AI-powered keyword research, and validates everything through a multi-step QA process.

1. Prioritise

First, we created a data-driven prioritisation model in Google Sheets, scoring products by revenue potential and focusing on high-conversion products with low impressions. This ensured we optimised products that would deliver the greatest impact first.

2. Detect

Next, we built a semi-automated detection system to identify feed issues across both the GMC account and individual products, creating a clear issue list for resolution.

3. Optimise

To scale our optimisations, we developed an AI-powered workflow using GPT and Claude to identify high-volume keywords relevant to each product, and generate optimised titles and descriptions aligned with real-world search behaviour. Account-level issues were allocated to ELC’s development team and their feed agency, Feedonomics.

4. Validate

Implemented a 4-stage QA process: automated checks → team review → client approval → feed sync. One product per group manually QA’d before sync.

THE RESULTS

The results speak for themselves.

Optimised products achieved 178% revenue growth, 100% session growth, and 87% cart add growth. Meanwhile, non-optimised products saw a 2% revenue decline, making optimised products perform 90x better.