Increasing paste-ink throughput with predictive risk scoring
A weekly risk score that predicts which ink orders will need post-test adjustment, with about 90% accuracy, so planners schedule around them.
Challenge
Paste-ink production for banknotes and secure documents involved frequent post-test adjustments that consumed about 42% of workstation time, each averaging 300 minutes. The risk per order sat in fragmented data across more than ten systems and over a thousand materials.
Solution
Arundo implemented a site-configurable logistic-regression risk model inside a prototype planning tool, combining a produced-material risk term with the contributions of the consumed materials after filtering the top 200 candidates. Risk scores are generated weekly from order details before production and turned into prioritised planning recommendations.
Impact
About 90% accuracy overall at a 30% false-positive tolerance and about 70% on previously unseen materials; adjustment time turned into proactive scheduling.