AI-Powered Brand Detection for Greater Accuracy and Reliability

At Launchmetrics, delivering accurate and reliable data is essential to helping fashion, beauty and lifestyle brands measure the true impact of their communication strategies.

To further enhance data quality, especially for brands with complex or ambiguous names, we are introducing an AI-powered approach to brand detection.

The challenge: complex brands, ambiguous contexts

Many brands operate across multiple languages and industries, or share their name with cultural references, places, or concepts. Historically, this has required:

• long and complex keyword lists
• detailed inclusion and exclusion rules
• frequent manual updates to maintain accuracy

While this rule-based approach has delivered strong results, it can become increasingly complex and harder to scale for brands with ambiguous names, potentially leading to data noise or missed mentions.

A shift from rules to context

Our new approach marks a fundamental shift in how brand coverage is detected.
Rather than relying on complex query logic, it uses contextual understanding powered by AI.

This evolution is built on four key principles:

1. Simplified brand queries

Brand detection starts from a simplified base query, without multiplying keywords or creating rigid inclusion and exclusion rules. This reduces operational complexity and makes the system easier to scale and maintain over time.

2. Brand descriptions to provide context

Each brand is enriched with a clear, structured description that provides the AI model with essential context, including:

• what the brand represents
• its product categories and universe
• what should not be associated with the brand

These descriptions follow a structured and standardized format to ensure consistency across brands. This allows the model to correctly interpret mentions, even in ambiguous situations.

3. AI-driven relevance decisions

Using this contextual information, the AI model (LLM) determines whether a brand mention is genuinely relevant to the brand. Relevance is no longer determined solely by static keyword rules, but by semantic analysis guided by clearly defined brand parameters.

Text:

“The exhibition featured archival sketches from Dior.”

Semantic analysis understands: This refers to the luxury fashion brand.

Text:

“Maria Dior won the local art prize.”

Semantic analysis understands: This likely refers to a person, not the brand.

4. Manual quality checks

To maintain the highest level of quality, we also perform manual checks on top-performing and high-impact documents, adding an additional layer of validation where it matters most.

Hybrid system

It is important to note that AI is applied where it adds measurable value. For straightforward, non-ambiguous brands where our existing methodology already delivers excellent results, we continue to rely on the traditional process.

Brand detection operates as a hybrid system: brands can be tracked either through AI-powered enrichment or through our traditional methodology, depending on what ensures the highest accuracy and efficiency.

What this means for brands

This new AI-driven approach delivers tangible benefits:

• Fewer missed mentions, thanks to better contextual detection
• Less noise, with more accurate exclusion of irrelevant content
• Higher overall data quality, especially for brands with ambiguous names
• Simpler upkeep, with fewer manual updates

The result: more reliable insights and greater confidence in brand performance measurement.