For Mercado Libre
Craft a big data infrastructure
with Data Science on the top

keywords: ecommerce, Big Query, Hadoop, DoubleClick, Data collection, Data exploration, Data Impact

The Challenge

MercadoLibre (literally “free market” in Spanish) was founded in 1999 in Argentina and is now operating in 16 countries. MeLi is Latin America’s most popular e-commerce site by number of visitors (200+ million registered users, as of Q3 2018).

As a highly sophisticated online marketplace dedicated to e-commerce and online auctions, Mercado Libre is always striving for offering the best experience to its customers but also for its analytics teams so that technology is no obstacle.

The challenge was here to attribute conversions to the right channel in order to distribute the budget efficiently through all MELI’s departments, with the idea to optimize media and marketing expenses.

The Approach

The three most important challenges in this project were to:

1.Craft a big data infrastructure to provide a 360° understanding in ROI and acquisition cost per channel based on the original various set of data;

2.Develop and test with MeLi a best-in-class attribution process;

3.Use an agile framework with multiple iterations (sprints) to provide quickwins and incremental improvements (on the data granularity, the content sources to integrate,…)

  • New sessions 60%
  • New transactions 49%
  • Media savings (through a better media allocation) 8%
  • Increase in revenues from paid channels 65%

The Results Were Amazing

In a winner-takes-all context in LATAM, the number of users and of transactions were key KPIs. Using the infrastructure, MELI could experience :

  • 61% more sessions
  • 49% more transactions
  • 8% Media savings (through a better media allocation)
  • 65% increase in revenues from paid channels

This was possible through the design of:

  1. a Big Data Infrastructure
    (Big Query, Double Click, Hadoop, …)
  2. an attribution process based on various algorithms (Navigation Path Reconstruction, Map Path to coalitions, …)
  3. Related reportings.