Choose Language Hide Translation Bar
phil_kay
Staff
Saving Big Money in Pharmaceutical Manufacturing with Data Analytics

A pharmaceutical company was losing money and credibility due to a quality issue with a tableting process. In this case study you will see how they found a robust solution to fix the problem for good, increased their credibility with externals and saved many $100k per year in scrapped batches.

 

 

This case study shows how you can solve scientific and engineering problems by exploring data with the interactive visual and statistical tools in JMP.

This is a schematic of the multi-stage process for producing tablets from the Active Pharmaceutical Ingredient:Tablet process.jpg

 

What were the problems they faced?

  • Scrapping 15% of batches due to non-conformance to release specification.
  • The process owners breathed a sign of relief whenever a good batch was released.
  • QA would investigate failing batches but conclude there was no assignable cause.
  • This pattern continued over many years.
  • Then, after an audit, they received a letter warning that license to manufcature could be removed.
  • This threat had serious financial consequences as the product was off-patent and still generated significant revenue. 

How did they solve the problem?

By visually exploring the process data for clues to causes of out-of-spec output they gained a rich understanding of the process.

Using this approach they were able to find a robust solution to fix the problem for good. This increased credibility with externals and ultimately saved many $100k per year in scrapped batches.

They identified that just a handful of the process variables were important drivers of variation. A statistical model showed them how to control the process to consistently produce quality product:

 

Profiler.png

 

Try this live version of the profiler to see how you can interact with the model to find the bext solution:

This interactive profiler is driven by JMP Live. Change the settings, either by applying the different remembered settings, or by dragging the broken red lines within the profiler plots. You can see this and other examples in JMP Public.

 

Find out more about improving your processes with visual and statistical exploration in JMP.

 

Article Labels

    There are no labels assigned to this post.

Article Tags