After working with Alteryx automation for over a year, GC’s team of engineers have begun to leverage the insights powered by Alteryx to drive organization-wide advantages. Olefin and phenol production, GC’s area of expertise, require large amounts of chemical feedstock. Chemical feedstocks are raw, untreated petroleum oil materials. Using their learnings from over the year and Alteryx’s predictive suite of tools, engineers have been able to predict their chemical output, or “yield” as it is more commonly known in the scientific community, to make better feedstock purchasing decisions.
The process was as follows: Leverage sales, production, pricing data, and plant conditions from disparate systems like CRM’s and ERP’s to create a better algorithm that is user-friendly and allows for advanced analytics reskilling and upskilling, and provides deeper insights. The objective was to create a model to predict the price of phenol, which is an organic compound used as an intermediate for industrial synthesis. The team embarked on a year-long proof-of-concept period encompassing:
- Data training
Leveraging 19 parameters overall, the team evolved from utilizing one Python-driven Linear Regression Model to building an Alteryx driven Random Forest model which allows them to leverage up to 50 parameters with a high confidence interval above 0.95 and an error rate of 2.6%.
Prior to Alteryx, feedstock was purchased when needed and not in a strategic manner. The team now knows when it is best to purchase feedstock and how much, predict their yield, and drive value back to the business while doing so in an innovative and environmentally friendly way. This new process has driven $1.5M to the phenol businesses’ bottom-line.