Echotraffic Holding: From Raw Data to Production, 7x Faster

With just a three-person data team but a solid ambition to tap the market for extended payment terms and working capital, Echotraffic envisions data science and machine learning as a frictionless part of their product and organizational processes. We sat down with Lead Data Scientist Adrien Basso-Blandin and Data Scientist Hayet Bezzeghoud at Echotraffic to talk about the projects they’re working on and how Dataiku helps them achieve their goals.

Background

Before Dataiku, it was a long and painful process for the Echotraffic data science team to both build data projects and to put them into production — a challenge that many small and medium businesses (SMB) still face today. The Echotraffic data science team, made up of three data scientists, was primarily using Python in notebooks and a bit of C# to automate processes, but they didn’t have any visual tools for building data pipelines or to conduct on-the-fly data analysis.

As is the case with many small teams, this method was scrappy, yet ultimately functional. However, it was also extremely tedious, and in the long run — especially with the company’s growth and plans for future products, expansions, etc. — they realized it was not sustainable.

The Project

In July 2020, Echotraffic launched Finexpay, a new service that provides— among other things —  a new machine learning-based service that helps B2B e-commerce or marketplace operators (such as METRO FRANCE) offer their clients longer payment terms in order to increase their key performance indicators (conversion rate, average basket, user experience, etc.) The extended payment terms module adds up to 90 days on top of existing terms and is based on a client proprietary score.

In order to be more precise, Echotraffic built instantaneous client scoring, which allows the clients to define their refer limit according to parameters that go beyond financial stability. For example, they can automatically take into account phenomena impacting entire areas of activity, like the current global health crisis. 

The Finexpay client score is generated by Dataiku, from which the team built the entire project end-to-end. The team chose Dataiku for its:

• User-friendly interface
• Easy data exploration and analysis capabilities
• Flexibility, including the capacity to extend core capacity with Dataiku Plugins (both pre-built and custom-developed)
• Integrated notebooks connected directly to datasets
• Visual recipes, which even though the team is technical, save a lot of development time
• Ability to facilitate quick and easy project deployment to production
  

Results

From connecting to various data sources to pushing models to production, the team at Echotraffic is seven times more efficient with Dataiku than they were using a notebooks-only approach: