Data-driven root cause analysis of batch failure

THE PROBLEM

Unexplained batch failures‍

A team we work with were experiencing batch failures at manufacturing scale, despite no significant change to the process. Deep-dives in our platform and across various other databases did not yield any significant insight into where the issue was coming from. Our chemist worked with them to better understand what was going on, and how we could prevent future, costly, failures.

OUR APPROACH

Data consolidation and simplification

The manufacturing, process and scale-up teams were all seperate, and full knowledge share was not in place. We pulled together data from across the sites, including accessing non-digitised records that we deemed essential. This allowed us to see if anything had changed over the time period, despite the overarching process remaining consistent.

THE SOLUTION

Identifying a poor additive and an ageing substrate‍

With as much information as available, it became clear that failures only occured when two conditions were met:

The substrate was more than 6 months old.
The additive was from a new supplier.

Where both of these were not seen all batches passed. We put in processes to mitigate against either of these happening again, and a system to diagnose similar issues in future.