Most manufacturing IoT initiatives fail because they start too large. The team selects a vendor platform, designs a comprehensive architecture covering predictive maintenance, quality control, and energy management simultaneously, then spends 18 months in implementation before producing any measurable value. By month 12, executive patience is exhausted and the project is defunded. The correct approach is the opposite: start with the smallest possible scope that delivers measurable financial return within 90 days, then expand based on proven results rather than projected ones.
Start with energy monitoring. It has the fastest ROI of any manufacturing IoT use case for three reasons: the data is simple (current, voltage, power factor), the sensors are non-invasive (clamp-on current transformers require no machine downtime to install), and the savings are immediate (operational changes based on visibility, not capital investment in new equipment). A single production facility typically achieves 10-15% energy cost reduction within six months of deploying continuous monitoring. No other IoT use case delivers this combination of low complexity and high return.
Your 90-day implementation framework has three phases. Days 1-30: deploy sensors on main electrical panels (not individual machines), establish connectivity to cloud, validate data integrity. Days 31-60: establish energy baselines by shift, line, and product, configure anomaly detection, build operational dashboards, train floor supervisors to interpret data. Days 61-90: implement operational changes based on data (shutdown schedules, load balancing, compressed air leak detection), measure savings against baseline, produce executive report with documented ROI. At day 90, you have a working system, trained staff, and financial evidence to justify expansion.
Avoid over-engineering the pilot. You do not need a digital twin. You do not need machine learning. You do not need a custom mobile app. You need sensors, a gateway, a time-series database, and a dashboard. The technology stack for a pilot is deliberately simple: current transformers ($40-$80 each) connected to an industrial IoT gateway ($500-$1,500) running AWS IoT Greengrass, publishing to AWS IoT Core, storing in Amazon Timestream, visualised in Amazon QuickSight. Total cloud cost for a single facility pilot: $200-$400/month. Total hardware cost: $15K-$30K depending on facility size.
What to monitor first depends on your cost structure, but the universal starting points are: main electrical feeds (total facility consumption by time of day), HVAC systems (typically 30-40% of non-production energy in manufacturing), compressed air systems (notoriously inefficient, often 25-30% waste from leaks), and lighting circuits (easy wins from schedule optimisation). Do not attempt to monitor individual production machines in your pilot — that requires machine-specific knowledge, often proprietary protocols, and significantly more sensors. Panel-level monitoring captures 80% of the insight at 20% of the complexity.
AI Readiness Checklist
Assess whether your enterprise is ready for production AI — the same framework we use in discovery calls.
The expansion path after energy monitoring follows a predictable sequence. Phase two (months 4-8): predictive maintenance on critical rotating equipment. Install vibration sensors and temperature sensors on motors, pumps, and compressors that cause production downtime when they fail. The ROI model shifts from energy savings to avoided downtime — if an unplanned failure costs $50K-$200K in lost production, and a $2K sensor package provides two to four weeks of advance warning, the economics are compelling. Phase three (months 9-14): quality control integration, connecting IoT sensor data with production quality metrics to identify environmental conditions that correlate with defects.
Organisational readiness matters more than technology readiness. Your pilot needs three roles filled: a project sponsor (plant manager or VP operations with budget authority), a technical lead (controls engineer or IT/OT convergence specialist who understands both factory systems and cloud architecture), and a data champion (operations manager who will use the dashboards daily and drive behaviour change on the floor). If you cannot identify these three people, you are not ready to start. Technology without organisational commitment produces dashboards that nobody looks at.
Connectivity in manufacturing environments is the most underestimated challenge. Factory floors have metal structures, electromagnetic interference from motors and drives, and often no existing WiFi coverage in production areas. Plan for wired Ethernet to gateway locations (most reliable), with cellular backup (4G/5G) for resilience. Do not rely on factory WiFi for IoT traffic — it is typically managed by IT with policies (firmware updates, security scans, bandwidth limits) that conflict with real-time sensor data requirements. A dedicated IoT network segment, even if it is just a few Ethernet drops to gateway locations, eliminates 90% of connectivity issues.
Security concerns from your IT team are legitimate and should be addressed proactively. IoT devices on your factory network represent new attack surface. Mitigate this with network segmentation (IoT devices on a dedicated VLAN with no route to corporate systems), device authentication (X.509 certificates via AWS IoT Core, not shared passwords), encrypted transport (TLS 1.2 minimum for all cloud communication), and firmware update management (AWS IoT Greengrass handles OTA updates with rollback capability). Present this security architecture to your IT security team before deployment, not after.
Budget the pilot realistically. Hardware: $15K-$30K (sensors, gateways, network infrastructure). Installation labour: $10K-$20K (electrician for CT installation, network drops, gateway mounting). Cloud services (first year): $3K-$5K. Integration and configuration labour: $20K-$40K (internal or contractor time for setup, dashboard development, training). Total pilot cost: $48K-$95K. Expected annual savings for a facility spending $2M on energy: $200K-$300K. This is not a speculative investment — it is an operational improvement with a measurable, predictable return that compounds as you expand to additional facilities and use cases.
Define success criteria before you start, not after. For a 90-day energy monitoring pilot, success means: all sensors reporting reliably (>99% data availability), baselines established for all monitored circuits, at least three actionable findings identified, at least one operational change implemented, and measurable energy reduction documented (even 2-3% in 90 days validates the approach). Avoid the trap of defining success as 'full platform deployed' or 'all machines connected' — those are scope definitions, not value definitions. Value is measured in dollars saved, downtime avoided, or defects prevented.
Want to discuss these ideas?
We're always happy to talk shop about cloud, AI, and what it takes to move from pilot to production.