Mercado Libre: Enhancing the User Experience With Lookalike Audiences

Discover how Mercado Libre uses Dataiku to unlock the power of lookalike audiences, simplifying complexity for code-free marketers.


New users conversion with the lookalike tool built with Dataiku


Improved conversion rate with the lookalike tool built with Dataiku

Just Days

For POC development and model training, versus weeks before Dataiku


The data and analytics team has been a part of online marketplace company Mercado Libre for nearly a decade and currently consists of 400 people divided into different teams according to their function. 

On one hand, there are the verticals that support the most important business units within Mercado Libre, such as fintech and commerce. There are also the horizontals responsible for facilitating capabilities to the entire data team and to all of Mercado Libre. They own administration and development of data tools, generation of new data products, data governance, and data culture. 

Today, Mercado Libre has more than 2,000 data scientists (some of them sit within business units) who impact its business with over 100 initiatives on a daily basis. When it comes to Generative AI, Mercado Libre held a hackathon that resulted in over 200 ideas, 10 of which are already being explored more seriously. 

With an already sophisticated analytics and AI practice evidenced by the information above, discover how Mercado Libre is driving innovation and growth via a very specific initiative designed to enhance the discovery of potential new customers: Lookalikes. 

What Is the Lookalikes Tool and How Does It Work?

Every day, the marketing and business teams send many communications (across email, real estate, and push notifications) to acquire new customers, prevent churn, and generate cross-sell opportunities. To accurately perform these actions and develop impactful contact strategies, Mercado Libre needed a deep understanding of their users’ profiles, in order to optimize overall campaign conversion rates.

The key use case the Mercado Libre team uses Dataiku for is the development and implementation of their Lookalikes tool. The application serves both users without experience in development or analysis as well as those with expertise in these areas, offering them valuable tools for execution.

Specifically, Dataiku enables Mercado Libre to leverage an initial user base to generate a predictive model that identifies the most representative characteristics of these users. This type of predictive model (also known as a lookalike model or similar audience model) allows the team to anticipate who could be potential users in the future, facilitating targeting marketing efforts and improving customer acquisition strategies.

End users were involved in the project’s evolution, providing feedback that was instrumental in guiding the enhancements and updates of the project. Dataiku’s capabilities enable the team to tailor the Lookalikes application specifically for use with Mercado Libre data, simplifying implementation for users and providing them with a user-friendly interface for autonomous and hassle-free testing.

The Right Value Proposition to the Right User at the Right Time

Before Dataiku, the process of achieving the use case resembled a variety of methods commonly used at Mercado Libre, primarily based on machine learning, to generate marketing audiences. However, Lookalikes offers a distinct approach by rapidly cloning target audiences using comprehensive customer insights, providing a 360-degree view.

As the application of Lookalikes has progressed, it has garnered increasing adoption due to its ability to deliver superior results within a reduced development timeframe, even for users who are not data scientists. This surpasses the current metric value of campaigns, indicating its effectiveness in enhancing marketing strategies.

Since integrating Dataiku into our ecosystem, we’ve observed a significant acceleration in generating insights and analyses. Mariano Hombre Senior Manager of Data & Analytics, Mercado Libre

The Lookalike tool has positively impacted several projects where the team managed to exceed the BAU conversion rate by more than 50%. It has also driven a +40% new users conversion. Currently, the output of these models is consumed by business and marketing teams that send daily communications to Mercado Libre users, helping them to better select their audiences.

Plus, the team has significantly improved their machine learning POC development and propensity model training from weeks to days with Dataiku.

Key Dataiku Features for Lookalikes

Dataiku was chosen as a platform for Mercado Libre for a compelling combination of factors:

  • User-Friendliness: Dataiku’s intuitive interface and drag-and-drop functionality make it accessible to a broader range of users, including those with less technical expertise. This is crucial for fostering collaboration between data scientists and business analysts within the Mercado Libre team.
  • Customization for Diverse Needs: Dataiku’s powerful workflow capabilities allow Mercado Libre to tailor experiences for both simple and highly complex use cases. They are able to build streamlined processes for routine tasks while maintaining the flexibility to handle advanced data manipulation and modeling.
  • Educational Resources: Dataiku Academy provides a wealth of learning materials for the Mercado Libre team members to continuously develop their skills and stay current with the latest advancements in the field.
We were impressed with Dataiku's commitment to customer satisfaction. Their high-quality support ensures we have the resources needed to overcome challenges and maximize the value we get from the platform. Mariano Hombre Senior Manager of Data & Analytics, Mercado Libre

The Right Balance Between Communication and Conversion

The Lookalike model in Dataiku has enabled Mercado Libre to discover valuable insights about their user segments and apply them to various business decisions (i.e., attracting new users without being intrusive or completely losing their attention).

Overall, the team is in an “AI-first” mindset and is learning about and experimenting with Generative AI to prepare for the future of GenAI-orchestrated workflows and products. They believe that the orchestration and automation of AI agents will significantly impact systems design, which prompts them to challenge their traditional practices and ways of thinking. They acknowledge that “Simplicity is the ultimate form of sophistication,” signaling a move towards innovative but also elegantly simple and user-friendly experiences for their users.


Marc O’Polo: Optimizing Print Campaigns With Uplift Modeling

Learn how Marc O’Polo leverages uplift modeling for print ad optimization. Gain key insights from the Dataiku Everyday AI Conference.

Read more

MandM: Lightning Fast Value From Data Science With Dataiku

MandM uses Dataiku’s deployment infrastructure and MLOps capabilities to amplify the impact of their modest data science team, delivering impactful use cases across the company.

Learn More

Action: Powering Data-Driven Decisions & Company-Wide Trust

Dataiku scientists at Action worked with Capgemini and Dataiku to develop more accurate and transparent forecasting models, faster, ultimately achieving a 900% improvement on forecast runtime.

Learn More

LVMH: Centralization & Personalization — A Hybrid Approach to AI

Discover how LVMH centralized and customized deployment of AI algorithms for its luxury goods houses.

Learn More

Showroomprivé: Putting ML-Powered Targeting in the Hands of Marketers

Showroomprivé leverages Dataiku to innovate across their business, including for machine learning-based targeting to build marketing campaigns that are 2.5x more effective.

Learn More