Developer of software named Matt Basta recently published an entertaining detailed story about his work experience in Uber, where he was instructed to create an alternative to Excel tables to work with big data.
In 2016, Basta joined the Uber unit called “Crystal Ball”, whose main task was to automate the process of creating models for data analysis. At the time, Uber was actively working on entering the Chinese market and needed reliable tools to work with big data.
“The entire work of the team was supported by Uber China. R-Crusher was a tool that helped us obtain the necessary data to compete with DIDI,” recalls Basta.
Shortly after his employment, Basta was given his first task: to create a user-friendly web interface for working with data from R-Crusher that resembled Microsoft Excel. According to Basta’s colleagues, “Every day without this tool cost us millions of dollars,” prompting him to begin working on the task immediately.
The initial version of the graphic interface for R-Crusher, named Weasley, was completed a few weeks after starting the project. However, it took an additional six months to refine it and ensure it could fulfill its assigned tasks. A team of twelve specialists participated in the development process.
The final tool developed by the Crystal Ball team was a JavaScript-based Excel clone that operated within a browser. It met all the required standards. However, during additional testing, it was discovered that Weasley and the original Excel produced slightly different results when working with the same data.
After extensive investigation into the code, formulas, and discussions with colleagues, Basta discovered that the original Excel would stop calculation when the difference between the last values of a cell fell below a specific threshold value. This algorithm sometimes produced errors, resulting in Excel providing incorrect outputs, which strangely enough, was acceptable.
As a result, Crystal Ball obtained an improved version of Excel, free from its technical limitations and more accurate. However, it was essential for the calculations to match exactly with Excel. Consequently, Basta had to fine-tune his algorithm to ensure precise consistency with Excel’s outputs.
Regardless, the main focus was on achieving results, and in July 2016, the inspired team successfully implemented the browser-based Excel, which even showed some positive outcomes.