The problem: root cause analysis on scattered data
Production issues show up in SPC charts, but the investigation often moves to other tools. Keeping the analysis in the QMS keeps the evidence together.
The SPC chart shows that something changed, but the investigation needs lot genealogy, raw material data and batch history. Exporting all of that to a spreadsheet or a separate RCA tool breaks the link to the records it came from.
What this demo shows
- View an SPC chart showing issues with key performance indicators on a production line.
- Perform the root cause analysis in Uncountable's QMS.
Why it matters
Closing the CAPA record isn't the same as fixing the underlying problem. A root cause analysis built on connected batch and lot data is more likely to find the real cause, such as a raw material lot change.
Keeping the analysis, its plots and its conclusions in the QMS means the corrective action and any follow-up can point back to the evidence that justified them.
See it with your own data
This video is part of QMS. To see how it would work with your own formulations and records, request a demo. Our team replies within 30 minutes during business hours.
Read the transcript
Today, we'll be taking a look at Uncountable's AI capabilities. We'll be using AI to perform a root cause analysis on a production line that seems to be experiencing some issues. To start, we're looking at a dashboard for our tomato soup production line, and we can see some key performance indicators for batches coming out of the factory.
We can see that the most recent batches are out of specification as their mass deviation, nozzle flow rate, and batch viscosity seem to be varying significantly compared to the previous batches. Naturally, we'd like to know why this is happening, and we can start off by opening our AI agent and just asking it to take a look.
Here, we specify the problematic batches, the necessary outputs that are having issues, and we're asking it to compare it to some of the previous batches to identify what might have changed that caused this out-of-spec problem. After a few moments, the LLM pulls in the relevant batch information and then generates an RCA for us to review.
In this case, it suggests that the most likely cause is a set of raw material lot changes between the earlier in-spec batches and the newer batches that are out of spec. It also highlights specific lot numbers involved, and from here, we can take the analysis one step further by asking the model to create a report.
Based on the completed RCA, we want a clear notebook that captures the key findings in a form we can easily share with colleagues and use to investigate why this out-of-spec issue occurred. After a few moments, we see that the previous RCA has been reviewed and that the AI has produced for us a notebook report.
Looking at the notebook, we can see the visualizations that the AI generated for us, including plots for nozzle flow, viscosity, and net weight deviation. These clearly show the step change between the earlier in-spec batches and the later out-of-spec batches. The notebook also includes an executive summary that identifies the affected batches and highlights the key finding, a raw material lot change that coincides with the onset of the issue, with the tomato concentrate identified as the most likely contributor.
We also see comparisons between the out-of-spec and in-spec batches, along with a breakdown of the input changes, including the ingredient lot IDs used before and after the shift. Finally, the report recommends next steps to confirm the RCA, such as verification testing, checking previous lots, and reviewing process parameters. In this demonstration, we started with a problematic production line, used Uncountable's AI platform to perform a root cause analysis, and generated a report with clear comparisons and visualizations to help guide the next steps.
