Brand ID: Scaling Visual Consistency at Mercado Libre with Multimodal AI

Executive Summary

Maintaining brand integrity across thousands of assets is a monumental task, especially for a multi-brand ecosystem like Mercado Libre. As the company grows, manual brand compliance checks became a bottleneck for creative teams. We partnered with Mercado Libre to build Brand ID, an AI-powered compliance tool that automates the review of image and video ads. By leveraging a multimodal AI stack —including vision models, LLMs, and speech to text models—we transformed a manual process that took hours into an automated check that takes seconds, ensuring 100% compliance for every branded asset.

‍We turned brand governance from a bottleneck into an automated system that scales across 18 countries

About The Company

Mercado Libre is Latin America’s leading e-commerce technology company. Through its primary platforms, Mercado Libre and Mercado Pago, it provides solutions to individuals and companies buying, selling, advertising, and paying for goods online. With sub-brands like Meli+, Mercado Ads, and Mercado Play, the company operates a complex visual identity system across 18 countries.

The Challenge

For the Brand Team at Mercado Libre, "Brand Governance" is critical. Every piece of advertising must adhere to strict guidelines regarding logo placement, color accuracy, and messaging. However, as the volume of creative assets grew, manual oversight became impossible to scale.

The manual review process was:

• Time-Consuming: Reviewing long-form video ads for specific objects and logo proportions took hours of human focus.
• Subjective: Consistency varied between different human reviewers.
• Prone to Error: Small details, such as incorrect contrast ratios or the wrong hex code for a sub-brand logo, were easily missed.

Mercado Libre needed a centralized, automated system that could "read" an image or "watch" a video just as a brand manager would, but at the speed of software.

The Solution

We designed and deployed Brand ID, a standalone internal web application hosted on Fury, Mercado Libre’s cloud platform. The tool provides a comprehensive automated report for every asset submitted, covering:

• Logo & Geometry Analysis: Using a Computer Vision API model, the tool detects logos and verifies their size, padding, and specific color hex codes (e.g., ensuring the Mercado Pago blue is exact).
• Visual Asset Scrutiny: The system analyzes the entire asset’s color palette and checks for text contrast to ensure accessibility and readability.
• Video Content Intelligence: For video ads, we integrated a vision-enabled LLM to detect specific physical objects required by the brand, such as "posnets" (Point of Sale terminals), branded credit cards, or return boxes.
• Multimodal Message Auditing: We utilized a speech-to-text model to transcribe audio from videos and LLMs to analyze the semantic quality of the message, checking for script length, tone, and brand-safe language.

The result is a comprehensive Web Report that gives human reviewers a clear "pass/fail" breakdown across logo, color, contrast, and message analysis sections.

The Results

Since its deployment, Brand ID has become a mandatory step in the creative workflow for all branded assets.

Wrapping Up

By building Brand ID, Mercado Libre has bridged the gap between creative freedom and brand discipline. This project demonstrates how Generative AI and Computer Vision can be used to protect brand equity, allowing the creative team to focus on storytelling while the AI handles the technical rigors of compliance.

Why Mercado Libre Loves Working with Us

Throughout our history, Muttdata has conducted various projects to enhance Mercado Libre’s data capabilities with a special focus on MarTech and AdTech use cases. Mercado Libre has also chosen Muttdata time and again to conduct projects related to Generative AI, Machine Learning, and Data Stacks.

Other joint projects include:

• GenAds: Automating the creation of high-performing banner ads for smaller sellers using Amazon Bedrock.
• Identity Resolution: Centralizing audiences using a Master ID (ML_ID) for improved advertising attribution.
• Multi-Touch Attribution: Implementing deep-learning models to measure the incremental lift of ad touchpoints.
• Bidding System Optimization: Enhancing real-time bidding for promoted items, resulting in a 25% increase in clicks.

Do you want similar results for your brand governance? Schedule a call with us!

‍