Success Story
How engineers diagnose and optimize pharmaceutical manufacturing processes 10-100x faster with Visplore
A multinational pharmaceutical company had digitalized data collection from manufacturing (AVEVA PI), but the use of data by process experts was still in early stages. Real-time monitoring and dashboards existed, yet deeper troubleshooting and optimization with historical data were rare and required hours of spreadsheet work.
Challenges to broader data use included:
- Products difficult to standardize and know-how scattered across teams
- Complex batch processes with multiple phases, varying lengths, and parallel asset production
- Inconsistent data availability and product contextualization (e.g. unlabeled batch phases)
- Limited analytics expertise among product experts
Use Case 1: Long-term Process KPI Monitoring
CHALLENGE: While quality KPIs were monitored, the batch process was much harder to quantify and track. Deviations in cycle times per phase or rising trends in maximal temperature were typically detected too late—only after poor result KPIs surfaced.
SOLUTION: Visplore enabled engineers to derive KPIs directly from batch curves like temperature or flow rate. Batch phases were extracted in real time using rules or pattern search, allowing automatic process KPI tracking and trend analysis. When issues arose, engineers were alerted with actionable graphics.
RESULT: Early detection of long-term deviations cut time-to-action, preventing wasted batches and maintaining low cycle times—all achieved using only process data, even without batch event frames or added context
Use Case 2: Optimizing Yield by Comparison of Campaigns
CHALLENGE: Engineers observed varying yields between productioncampaigns, despite using the same recipe and equipment. Identifying the causes of these differences was time-consuming and often relied more on intuition than on data analysis.
SOLUTION: Visplore enabled engineers to merge historian data with MESyield data to compare process curves from high- and low-yield campaigns across reactors. An automatic algorithm identified the process signals and batch phases with the biggest differences. Visualizing aligned patterns per phase revealed key deviations—like slower reactions and extended hold times—helping pinpoint root causes of low yield in minutes.
RESULT: This approach enabled fast, data-driven improvements in process consistency and product yield, without requiring deep data science expertise. The outcome: higher output, less waste, and better return on existing production capacity.
