One area where we applied analytics is for the toner we manufacture for our printers. Developing toner is a complex, multi-step chemical process overseen by our team members. The process itself is an attempt to destabilize materials in a controlled manner. It’s a difficult, delicate process that can be thrown off by even a minor variation.
Because of the nature of human decision making – with differing opinions and judgments – the process wasn’t always precise. Compounding the situation, the human decisions were made with limited, real-time insights, and they didn’t tap into the full breadth of our analytics capabilities. As a result, we often saw a wide range of outcomes, from uncontrollable particle growth to long, inconsistent cycle times.
To address this, we embarked on a data-driven exploration to improve process control and reduce cycle times. Our first step was to leverage the power of the vast amounts of data we create.
Our manufacturing system generates and saves large amounts of data from each batch of toner. Yet, we were not actively analyzing and extracting insights from it. This oversight and lack of knowledge sharing was the first thing we looked to address.
To start, we combined, cleaned and analyzed historical manufacturing data sets. An estimated 20 million data points from 200 toner batches informed our analysis.
Through this research, we found the source of our process control issues. Unbeknownst to us, one minor element had a huge impact on outcomes. Armed with this new knowledge, we were able to adjust our approach and build more accurate algorithms to drive decisions for consistent outcomes.
This and other discoveries — and their resulting improvements — were only made possible because of the efforts of one of our team members, Catherine Randolph. Formerly an R&D scientist at Lexmark, Randolph joined the toner development team, empowered to follow her passion for data analysis with this project.
She and other Lexmark team members reviewed scientific literature to better understand the toner manufacturing process and how it could be improved. While there’s a lot of research on manufacturing, the process models described were typically based on simpler systems and none directly focused on the process for toner. Randolph’s particular interest in a data-driven approach, plus our need to address process control, created the perfect environment for her to develop and drive this project.
With the new data-driven model for toner production in place, we have made notable improvements in process control, significantly reduced overall cycle time and slashed our manufacturing expenses. Our efforts won a Manufacturing Leadership Award for outstanding achievement in engineering and production technology.
This project will continue in phases, and we expect to see an even greater improvement across the board when finished. This process showed us how powerful data can be and the importance of not letting any insights go to waste.
Throughout this process, we discovered three key findings that you can apply to your business: