An Owner-Operator set out to build a predictive analytics program for their flagship refinery — a $20M-per-year opportunity stalled by data collection. Populating the required data templates manually was estimated at 16 weeks per process unit: over 2.5 years for the full refinery.

This case study covers how the team ran a proof of concept with the Intelligent Drawing Platform, aligning process variables with the instrument loops measuring them on the P&IDs, then cross-referencing instrument tags against the data historian to map their data streams. The full-scale rollout processed 5,100 facility P&IDs and compressed the timeline from 2.5 years to six months — with an estimated $3M return on the data collection and contextualization effort alone.

The numbers that matter: 16 weeks per process unit reduced to a six-month facility-wide execution, higher-quality inputs for the predictive models, and additional value outputs generated along the way — equipment inventories and instrument-to-historian mappings the refinery kept using well beyond the analytics project.

It's a pattern we see across data consulting engagements: the value of predictive analytics is decided upstream, in the quality and speed of the facility data feeding it.