💪 This project was an effort to automate much of the manual processes in subscriber forecasting after relocations and layoffs reduced headcount in the group from 6 to me 😶

⚙️️ To build it required working Teradata SQL knowledge to import a lot of data and advanced R knowledge to process and analyze it all. Also it required the use of Facebook’s prophet package 🙏 which performed much better than any of the ensemble methods.

😌 The result was a fully automated forecast built from over 70 highly granular forecasts vs a handful of manual high-level forecasts. In fact, this forecast picked up on trends deeper in the data we wouldn’t have seen until much later when they’d grow large enough to show up in high level data.

My favorite part about this forecast was using looping and some crafty R techniques to really scale our ability to work 📈🚀. Furthermore, I was stoked to uncover hidden trends deeper in the data! 🕵️‍♂️