
Manufacturing

Increase in network-wide factory utilisation
Reduction in unplanned production delays
14%
Reduction in changeover and re-tooling time
Seven Billion gave us the intelligence layer to route orders across 100+ factories with zero manual intervention.

Chris Mullins
SVP
MacroFab

About
An electronics manufacturing platform routes PCB orders across 100+ factories in real time: utilisation up 21%, delays down 18%, changeovers down 14%.
Industry
Manufacturing
Company size
100 – 500 employees
Founded
2013
Business Context
Business context and operating model
A digital platform sitting on top of a very analog industry The client operates North America's largest technology platform for electronics manufacturing, connecting hardware companies to a network of more than 100 partner factories spanning prototype runs through high-volume PCBA production. Each factory carries different equipment, capacity, certifications, and lead times. As order volume grew, the business needed every job routed to the factory that could deliver it fastest and cheapest without breaching quality or capacity constraints — a matching problem with far more variables than any planning team could hold in their head order by order.
Problem Definition & Framework
Current State
Failure Modes
Decision Criteria
KPI Baseline
Problem definition and framework
Order-to-factory matching had outgrown manual planning Production planners were assigning jobs to factories based on relationships and rough capacity estimates rather than live data. Two failure modes resulted: high-demand factories ran over capacity and slipped delivery dates, while comparable capacity sat idle elsewhere in the network because no one had visibility into it in the moment a job needed placing. Changeover sequencing within each factory compounded the problem — jobs arrived in whatever order they were quoted, not in the sequence that minimised setup and re-tooling time across a facility's queue.
Decision framed: Which factory should receive each PCB assembly order when cost, lead time, equipment compatibility, certification, capacity, and changeover constraints all matter at once?
Framework used: Mixed Integer Programming for constrained order-to-factory assignment
Baseline mapped: current process, failure modes, handoffs, data owners, and decision latency.
Success criteria tied to business KPIs, adoption behavior, and operating cadence changes.
How it works
Before / After

Architecture

Process Flow

Data Pipeline

Results

Delivery Timeline

Engagement Method & Process Diagram
01 Discovery
02 Data Audit
03 Baseline
04 Prototype
05 Pilot
06 Production
07 KPI Monitoring
Engagement method and process diagram
Process diagram: Open orders + BOM/PCB requirements -> live factory capacity telemetry -> MIP optimizer in Gurobi -> recommended route and sequence -> planner override workflow -> utilization/delay tracking
Engagement method: Constraint discovery with planners; historical routing baseline; telemetry feed design; optimizer prototype; planner override design; rollout by factory network segment.
Delivery pattern: discovery, data audit, baseline measurement, prototype, pilot workflow, production rollout, and KPI monitoring.
A Mixed Integer Programming router with live factory telemetry We built a real-time order-routing optimiser using Gurobi, connected to a live factory-capacity feed so the model always solves against current, not stale, capacity. Given every open order, the model simultaneously solves for lowest landed cost, fastest feasible delivery, and minimum changeover time within each factory's queue, subject to hard constraints on certification, equipment compatibility, and committed capacity. Planners see the recommended routing through a dashboard, can override for relationship or strategic reasons, and the system automatically recalculates downstream impact of any override — so a manual change to one order's routing doesn't silently break the plan for the rest of the queue.
Architecture, Results & Measurement
Solution architecture, results, and measurement
21% more network capacity without adding a single factory Network-wide utilisation rose 21% in the first two quarters — capacity that already existed but had been invisible to planners routing by memory. Unplanned production delays fell 18% as jobs were matched to factories that could actually deliver on the committed date rather than the factory a planner happened to think of first. Changeover time within individual factory queues fell 14% through better job sequencing, and planning effort dropped enough that the team redirected its time toward partner development rather than day-to-day order placement.
Measured outcome: 21%: Increase in network-wide factory utilisation
Measured outcome: 18%: Reduction in unplanned production delays
Measured outcome: 14%: Reduction in changeover and re-tooling time
Measurement loop: compare weekly operating behavior against baseline, surface exceptions, and review model/workflow drift with business owners.

The optimiser did not replace our planners' judgment on partner relationships — it just stopped us leaving capacity on the table because nobody happened to think of the right factory that day.
VP Manufacturing Operations, North American electronics platform
KEEP READING
Explore more studies on AI, analytics, and enterprise intelligence.








