Architecture of Precision: Inside Mercado Libre’s CDP

About the Company

Mercado Libre is Latin America’s largest e-commerce and fintech ecosystem. Operating in 18 countries, it empowers millions of buyers and sellers through its marketplace, payment platform (Mercado Pago), logistics network (Mercado Envíos), and ad-tech solution (Mercado Ads).

Mercado Ads, Mercado Libre’s advertising arm, enables brands to reach highly relevant audiences directly within the ecosystem. Using first-party data, it offers advanced segmentation, campaign automation, and measurable results — supporting both small sellers and large-scale advertisers.

As Mercado Libre’s ad business grew, so did the complexity of managing audience data — setting the stage for the creation of a modern, Composable CDP.

‍

Executive Summary

Muttdata collaborated side by side with Mercado Libre to build a robust, scalable Composable CDP designed to revolutionize how user data is managed, activated, and leveraged for marketing across Latin America.

The challenge was clear: audience data was fragmented across systems. Through a close collaboration with Mercado Libre’s teams, we delivered a platform that enables real-time audience creation, precise identity resolution, and cross-device targeting — all while ensuring privacy compliance and future-proofing Mercado Libre’s advertising capabilities.

This new Composable CDP powers hundreds of campaigns, improves user experience through personalization, and helps Mercado Libre unlock new revenue streams by connecting behavioral and transactional insights at scale.

The Challenge

As Mercado Libre scaled its advertising ecosystem across Latin America, it ran into an obstacle: audience fragmentation. With millions of users interacting through multiple devices, accounts, top-of-the-industry media content streaming partners, and even third-party platforms, the company struggled to build granular segments and execute precise ad targeting. This made it difficult to attribute campaign performance, personalize engagement, and optimize marketing investments.

On top of this, Mercado Libre’s existing analytics tools needed to boost the speed of insight delivery. It was during the early stages of our strategic partnership, around 2021, when Muttdata and MELI started to think about building a Composable CDP platform. At this time, teams worked using non-scalable ETLs that required manual intervention to build audiences. There was no centralized system to automate audience creation, track users across channels, or activate audiences in real time, all of which limited the effectiveness of both internal campaigns and off-platform media buying.

To meet these challenges, Mercado Libre side by side with Muttdata set out to build a robust, scalable Composable CDP one that would centralize all user data (behavioral, transactional, and profile-based), power real-time segmentation, and support business use cases like:

• Attribution across devices and sessions
  
  
• Lookalike modeling and audience enrichment
  
  
• First-party data onboarding
  
  
• Retargeting for Mercado Pago users
  
  
• Privacy-compliant audience activation on and off-platform
  
  

The goal was not only to improve how audiences were built and activated, but to fundamentally transform how Mercado Libre made marketing decisions faster, smarter, and at scale.

The Solution

To help Mercado Libre transform its marketing capabilities, Muttdata collaborated closely with Mercado Libre’s internal teams to design and implement a full-scale Composable CDP a foundational layer to power audience intelligence, campaign optimization, and customer personalization.

We approached the project through two complementary tracks: building a powerful audience engine and developing a robust identity resolution system.

1. Audience Engine

We engineered a self-service platform that empowers Mercado Libre’s marketing and business teams to create and manage audience segments without writing code. This frontend tool, backed by a flexible query builder and orchestration layer, allows users to define complex rules combining user behavior, purchase history, demographic attributes, and ad engagement data.

At its core, the platform relies on a Java-based API responsible for the creation and dynamic management of audiences through simple configuration files. These configurations define both the data sources and the criteria used to build and refresh segments. The API orchestrates automated processes that keep each audience updated with a user-defined refresh rate, ensuring data freshness and reliability at scale. Initially exposed through a command-line endpoint, the API later evolved to power a user-friendly backoffice, allowing non-technical users to easily visualize, create, and modify their audiences.

The underlying data warehouse powering the Composable CDP was modeled to optimize read performance, a key requirement for the API’s audience creation and update workflows,  while enforcing strict data quality rules on writes. A dedicated data quality engine and alerting system continuously validate input data, preventing incomplete or erroneous datasets from propagating to production segments.

To expand the reach and flexibility of campaign creation, machine learning models were developed to generate lookalike audiences. Starting from a “seed” audience, these models analyzed demographic attributes and user activity signals — such as clicks, views, and purchases — to identify users with similar behaviors. The resulting expanded segments provided marketing teams with new, data-driven audiences for campaign activation, complementing manually defined targeting rules.

These audiences are continuously refreshed through automated ELT pipelines, served both in batch and near real-time. Through this combination of a flexible API, continuous data refresh, robust data quality checks, and machine learning–driven segmentation, the Audience Engine became a central piece in Mercado Libre’s marketing ecosystem, activating audiences across internal tools, on-platform campaigns, and external DSPs like Google Ad Manager with up-to-date and consistent logic.

‍

2. Identity Resolution

To unify user interactions across sessions, devices, and third-party platforms, we implemented a scalable Identity Resolution system centered around a unified user identifier built to consolidate signals from Mercado Libre’s internal systems and external partners .

We designed and deployed an Identity Resolution API for fast, secure lookups. Personally identifiable information (PII) is hashed and salted to meet Mercado Libre’s stringent privacy requirements. The architecture uses BigQuery and Cloud Spanner to orchestrate massive-scale data synchronization, ensuring low-latency access and high availability.

To maintain data integrity, we also implemented daily quality checks across accuracy, uniqueness, and consistency enabling Mercado Libre to trust the platform as a single source of truth for audience-based decision-making.

Together, these solutions gave Mercado Libre a centralized, extensible platform for marketing intelligence reducing manual work, improving data governance, and accelerating experimentation with new audience strategies and use cases.

The Impact

With the new Composable CDP in place, Mercado Libre transformed how it builds, analyzes, and activates audiences — across both owned and external marketing channels.

The biggest shift? Speed and scale. Teams that once depended on manual processes can now generate audiences and activate them quickly. This agility has directly improved time-to-market for campaigns and enabled more rapid testing of audience hypotheses.

The platform also improved marketing attribution and personalization, centralizing user data across sources and enabling more accurate customer segmentation. By consolidating identities through ML_ID and enriching user profiles with behavioral and transactional data, Mercado Libre now reaches users more precisely — across all stages of the funnel.

The impact of smarter audience building has been felt across the business:

• Increased campaign effectiveness, with better alignment between messaging and user intent
  
  
• Reduced media waste, by avoiding duplication and targeting inactive users
  
  
• Improved cross-channel consistency, powering cohesive experiences across web, app, and external ads
  
  
• Empowered non-technical teams to act on data autonomously, lowering dependency on engineering
  
  
• Enabled first-party data strategies to future-proof Mercado Libre’s ad business in a privacy-first world
  
  

The Results

Expanded Reach: Successfully integrated over 415 million users into audience recalculations, significantly enlarging the addressable market for targeted initiatives.

Massive Scalability: Achieved a daily processing volume of over 180,000 unique audiences across nine different sites, demonstrating robust and scalable computational power. More than 280,000 audiences were created over the last year, with a rate of more than 34,000 audiences in the last 6 months.

Enhanced Data Freshness: Enabled near real-time decision-making by ensuring all audience data is recalculated at the precise frequency that the product team requires.

Self-service audience creation: Empowered business and marketing teams to independently define and deploy new audiences using pre-modeled data segments, eliminating the need for technical intervention and accelerating campaign setup.

Optimized data processing: By modeling the data warehouse for efficient reads, the system ensured that only relevant data was processed during each audience refresh, improving overall performance and resource utilization.

‍

Wrap-Up

Mercado Libre set out to solve a critical challenge: how to unify fragmented user identities and activate audiences at scale, without compromising data privacy. 

With millions of interactions occurring across devices, accounts, and partners, understanding users holistically was essential to improving marketing performance and growing ad revenue.

Muttdata partnered with Mercado Libre to design and implement a robust Data Management Platform that addressed these needs head-on. At the heart of the solution was a new identity layer, powered by the ML_ID,  a unique identifier that consolidates data from multiple systems and touchpoints. 

Through a combination of API development, scalable data infrastructure, and strict PII handling, we enabled Mercado Libre to connect signals, orchestrate secure data flows, and unlock self-serve audience creation across the business.

What started as a technical solution became a strategic capability: a platform to power smarter decisions, faster marketing, and better user experiences. And it’s just the beginning.

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 once and again to conduct projects related to Generative AI, Machine Learning, and Data Stacks. Mercado Libre and Muttdata enjoy working with each other due to their teams’ expertise and professionalism. So the Paid Online Media Optimizer for Mercado Libre is not our only lap with this crew. Other joint projects include:

GenAds

Mercado Libre, AWS and us joined forces to tackle the challenge of producing high-performing banner ads with generative artificial intelligence (Gen AI), for dozens of thousands of smaller sellers. Using Amazon Bedrock, Amazon S3, and Amazon DynamoDB, we developed GenAds, an innovative solution that automates the creation of compelling product images, streamlining the process for sellers and enhancing their market presence with image-generating AI. Through GenAds, Mercado Libre helped small sellers through image generation for advertising, enabling sellers of all sizes to attract customers, increase click-through rates (CTR) by an average of 25%, and boost sales effectively.

Multi-Touch Attribution

Mercado Libre needed to improve its online attribution system to measure the incremental lift of each ad touchpoint in conversion journeys, overcoming the limitations of traditional multi-touch attribution (MTA) models. Muttdata collaborated with Mercado Libre to design and implement a new MTA platform using advanced deep-learning models. This platform accounted for channel interactions, time dependencies, user characteristics, and other business variables, enabling more accurate measurements of each ad touchpoint's true impact. The new attribution model allowed for real-time, campaign-level measurement closer to actual lift estimations. This improved the accuracy of marketing spend optimization, enhanced campaign performance, and informed better decision-making by providing insights into the effectiveness of various marketing strategies across customer journeys.

Boosting Mercado Pago's Customer Base

MercadoPago aimed to increase its Total Payment Volume (TPV) by expanding the range of products used by sellers and introducing new services. The challenge involved both identifying effective recommendations and implementing them through the best channels. Muttdata developed a comprehensive recommendation framework using Machine Learning models to predict sellers’ propensity to use specific services. This system tailored recommendations based on delivery timing, channel interaction, service relevance, and message content. We significantly accelerated the ML model deployment process. The project led to a 4% nominal increase in sellers selecting MercadoPago as their primary platform, growing from 6% to 10%. Additionally, the implementation of monitoring best practices enabled real-time metric tracking through user-friendly dashboards, further optimizing the platform's performance and decision-making.

Mercado Libre’s Bidding System

Mercado Libre’s Product Ads Team needed to enhance their internal bidding system for promoted items. The existing batch process was suboptimal, leading to inefficient billing and revenue loss, especially for lower bidders. Muttdata developed a real-time bidding system that optimized Click-Through Rate (CTR) and Conversion Rate (CVR) estimations for each item during searches. The system balanced the impact on Gross Merchandise Volume (GMV) while maximizing ad placement efficiency. It resulted in a more than 25% increase in advertising clicks without significantly affecting organic GMV, demonstrating the effectiveness of the real-time bidding system.

Mercado Libre’s choice to work with us on multiple projects is a testament to the trust we’ve built, thanks to our ability to deliver data-driven, measurable improvements across complex marketing and tech landscapes.

Do you want similar results for your e-commerce platform? Schedule a call with us!

‍