GE Meridium APM

The challenge

Field engineers and plant managers were receiving hundreds of alerts per day with no prioritisation, no suggested action, and no context about asset history. Alert fatigue had made the existing system nearly useless in practice. The product had 340+ screens built over 8 years with no consistent design language — creating a steep learning curve and frequent user errors in high-stakes environments.

The solution

We built a predictive alert hierarchy that surfaces the three most critical issues requiring attention, with suggested remediation steps drawn from engineering knowledge bases. A new design system — 180+ components — created visual consistency across all 340 screens. Context panels on every alert surface the asset's full maintenance history, failure probability curve, and recommended action ranked by estimated cost-of-inaction.

Our process

1. 01 Field Research

   Embedded with engineers at two refinery sites for 3 weeks. Observed actual alert-handling workflows. Identified the 'alert flood' as the primary failure mode.

2. 02 Predictive Model Integration

   Worked with GE's data science team to design the UX surface for ML-driven failure probability scores and confidence intervals.

3. 03 Design System Creation

   Built a GE-branded design system from scratch: 180 components, dark-mode-first (control room requirement), touch-optimised for tablet field use.

4. 04 Alert Architecture Redesign

   Replaced chronological alert lists with a criticality-ranked action queue. Suggestive guidance panel added to every alert.

5. 05 Validation & Rollout

   Piloted with 3 facilities, 40 engineers. Iterated on two major rounds of feedback before phased global rollout.

Outcomes

• 73%Reduction in mean time to acknowledge critical alerts
• 180+Design system components delivered
• 40%Reduction in operator errors on high-stakes tasks
• 3 wksNew engineer onboarding time (from 3 months)

Technology

• Angular
• HTML
• CSS
• REST APIs
• IoT Sensor Feeds