Mappy is a software publisher that largely focuses on cartography. Mappy designs maps, calculates routes with directions for cars, public transportation, bicycles and on foot, and enables users to search for points of interest, restaurants, hotels and shops via the Internet, from a tablet, cell phone, or a GPS device.
Visualizing data is important to understand what Mappy’s 10 million unique users are doing each month and how the data behaves. Every day at Mappy, around a dozen people need to find answers to questions about the product, features, and more.
For example, Mappy’s analysts needed to look at how users behave, how many times a point of interest (POI) is viewed on a map, the number of times that users click on a point, and even how many times the POI address is shown.
With 4 million POIs available on Mappy, the team’s ability to answer questions was quickly outstripped by new requests.
The previous system used was long, complicated, and tedious. The team knew it needed to rethink how Mappy was collecting, viewing, and understanding data.
“In the past, we were doing BI like in the 1980s. We used a processing chain with the server logs composed of an aggregation stage written in Python. The aggregated data was then stored on Microsoft SQL Server. To visualize the data, we used Excel, which pointed to SQL server cubes that we had to recompile every night. This process added hours, if not days, to data analysis. Finally, PDF exports were created so analysts could have a direct image in order to visualize the data” – Nicolas Korchia, Business Manager.
Each day this represented 150 GB of data produced by the company’s 120 operational servers. Faced with this continually increasing volume of data, Mappy had reached the upper limit of its vertical scalability with its old system. The platform needed a major redevelopment. The Business Intelligence Manager decided to take a radical turn that would enable analysts to obtain and visualize data in a simple and effective way.