A large, publicly traded technology company sought to replace ineffective, legacy revenue forecasting models.
We developed two Gradient Boosted Tree models that predict quarterly revenue (up to 8 quarters out). One model could predict granular revenue by:
- Region and Territory
- Business Unit
- Product Line
- Market and Segment
A second model was focused on explainability and predicts the error rate per quarter and category.
We achieved over 18% greater accuracy predicting revenue while also providing granular predictions, up to two years into the future.