With so many behavioral changes as a result of Covid-19 and market irregularities, historical forecasts have never been more vulnerable. It’s near impossible to predict future spend based on a superficial analysis alone. We partnered with a fortune 500 manufacturing company to identify at-risk customer spend by drilling into root causes.
We combined the power of natural language processing on qualitative customer feedback with historical spending and other operational metrics like time to response and fulfillment to predict sales trends at both the individual customer level and at an aggregated division level. In addition, we focused on interpretability, highlighting the most critical metrics that affected sales.