Predictive intelligence for Europe's fastest-growing cold chain operator.

NewCold operates highly automated deep-freeze warehouses across Europe. As their network scaled, so did the cost of unplanned downtime. Qubiz built the platform that changed that.

The challenge

Six facilities. One unresolved question: why does equipment fail when it does?

NewCold's automated warehouses run at -25°C. Every unplanned equipment failure means halted throughput, emergency engineering callouts, and — in the worst cases — product loss. As their European footprint grew from two sites to six in under three years, the reactive maintenance model they had inherited from smaller-scale operations began to break. The data existed — thousands of PLC sensors, WMS event logs, and environmental monitors — but it lived in isolated systems per site with no intelligence layer capable of surfacing early warning patterns.

The strategy

Unified data first. Prediction second. Operator trust last — and hardest.

Three phases, each with a hard exit criterion before the next began.

01
Data unification
Cataloguing every sensor type across every site, building a canonical event schema, and establishing a streaming pipeline ingesting 40,000+ data points per minute without loss — without interrupting live 24/7 operations.
02
Model training
An ensemble of gradient boosting and LSTM networks trained on two years of maintenance records. Feature engineering identified 18 failure signatures accounting for 74% of unplanned downtime across the estate.
03
Deployment
Alert logic embedded directly into the workflow tooling operators already used — not a new dashboard to check. Six-week calibration at the first site before rolling to the remaining five.
04
Continuous improvement
Post-rollout model retraining cadence established, with NewCold engineers trained to tune thresholds. Qubiz remains the embedded engineering partner for the platform.

WHAT WE BUILT

Predictive maintenance model

End-to-end design and build — from sensor ingestion to operator-facing alert logic. Trained on two years of historical maintenance records aligned to live sensor states. The model surface 18 failure signatures that account for 74% of unplanned downtime.

Unified operational data layer

Harmonising PLC telemetry, WMS events and environmental sensors across six warehouses into a single queryable source of truth — ingesting 40,000+ data points per minute without interrupting live 24/7 throughput.

Legacy system integration

Phased SCADA and proprietary WMS integration over six months across live facilities, embedded directly into the workflow tooling operators already used — no parallel systems, no shadow IT.

Operator trust and rollout

Six-week calibration at the first site before rolling to five more. Alerts surfaced inside existing screens — not a new dashboard no one opens. Qubiz engineers remained embedded until adoption metrics were confirmed.

The outcome

Numbers that appear in board reports.

34%

Reduction in unplanned downtime across all connected facilities within eight months of go-live.

22%

Uplift in inbound slot utilisation — recovered through better load visibility and turnaround prediction.

€2.1M

Estimated annual operational savings attributed to predictive maintenance and reduced emergency callouts.

Facilities now running on the unified data platform — with consistent KPIs across the European footprint.

“

Qubiz understood from day one that we couldn't afford a big-bang approach. They delivered incrementally, kept our engineers in the loop on every decision, and built something our team actually wants to use.