Solving escalations on machine-level

At one of our manufacturing customers, support-line engineers spent around 80% of their time on escalation solving, and always reactively, after the fact. These escalations are costly, unpredictable, and time-consuming – hampering scalability and growth.

The only way to counteract this reactive way of working is to reduce escalations. That’s where our customer’s Proactive Escalation Monitoring (PEM) team comes in. The PEM team builds models to proactively track new escalations and prevent reoccurrence, by monitoring machines on fleet-level in order to move from unscheduled downs to scheduled downs (USD2SD).

Think of models that predict certain failure modes and lifetime of components, or models that monitor sensor trends and trigger engineers in the field to timely order new sensors and prepare for scheduled maintenance.

Building models using the Self Service Analytics platform 

The PEM team is continuously looking for new escalation use cases to model, monitor, and prevent from reoccurring. They pick these cases using an escalation afterlife tool. In this tool, support-line engineers document findings and solutions for each escalation, and share ideas on how to proactively monitor the escalation. System engineers of the PEM team go through this database of escalation-cases and select those that are potentially promising, looking at business value, cost, effort, and impact.

As soon as they pick a new case, they use our Self Service Analytics (SSA) platform to build a model, turn it into a KPI, visualize it in a dashboard, and define actionable triggers for engineers in the field. The field engineers use that to prepare for swaps, solve sensor drifts, or foresee machine failures, preventing unscheduled downs and thus escalations. When the model turns out highly valuable, the team industrializes it to guarantee 24/7 monitoring and continuous value.

PEM Engineer: “By using SSA instead of local analysis, we’ve reduced diagnostic times from 72+ hours to minutes in certain cases”

Delivering measurable value

PEM currently has over 10 models running on SSA, and already industrialized five. From simple mathematical models to Machine Learning (ML) pipelines. Within six months, these models have sent 53 triggers to field engineers, of which 30 actually prevented an escalation. This resulted in preventing 26 unscheduled downs, a stunning achievement!

With a growing number of PEM models implemented, support engineers can spend more and more time on enabling proactive monitoring, preventing downtime, and potentially saving millions of escalation costs. This will enable our customer to further expand their factory and significantly enhance user experience.