What to actually look for in a production monitoring & scheduling system
A practical framework for evaluating IoT platforms. The questions to ask, the red flags to catch, and the demos that separate real capability from slideware.
Last updated: April 22, 2026
- Configurable KPIs beyond OEE. Formula builder for process-specific metrics like tons per hour, linear feet per hour, or spindle utilization. OEE matters, but it's rarely what your operation actually runs on.
- Two-tap operator downtime classification. Context captured at the source, on a shop-floor kiosk, at the moment of the stop. Not reconstructed from memory at the end of the shift.
- Closed-loop scheduling. Live schedule projection against actual production, not a static plan. The supervisor sees "30 minutes behind" at 10am, not at 4pm.
- Standard industrial hardware you own. $500 to $1,000 per machine for edge gateways or bolt-on sensors. Not proprietary hardware that locks you to a vendor after you've spent time and money installing it.
- Open APIs, full data export, customer-segregated cloud. No vendor lock-in on your production history. When you ask where your data lives, you should get a direct answer.
- Day-one implementation and self-service after. A working system at handoff, not login credentials. Admin UI handles machine, KPI, and dashboard changes without vendor calls.
- Roadmap toward maintenance and cost-to-serve. The vendor is building where your operation is going. Point solutions get stitched together later or start over when complexity arrives.
OEE gets the attention. It shouldn't get all the focus.
OEE matters, but operations actually depend on business-specific metrics. Steel service centers track tons per hour and linear feet per hour. CNC shops track spindle utilization and parts per shift. Packaging lines track cases per minute and changeover frequency.
The distinction that matters: "linear feet per run hour" (machine capability) versus "linear feet per shift hour" (operational efficiency). A system that can't separate those two numbers can't tell you where you're losing time.
What you need
- Configurable KPIs. Build formulas for metrics like "footage produced per run hour" or "run time as percentage of total shift time."
- Downtime tracking with classification. "Two hours down" becomes 45 min waiting for material, 30 min blade change, 45 min unexplained. Operators classify in seconds from a kiosk, feeding Pareto analysis.
- Chart builder, not pre-built screens. Supervisors need custom trend views: daily production, weekly throughput, downtime patterns over months, shift comparisons.
- Operator interface for actual use. Large touch targets, dark mode for visibility, minimal clicks for common tasks.
Red flags
What to ask in the demo
The closed loop is everything.
A monitoring system disconnected from scheduling cannot indicate ahead/behind status. The "closed loop" (schedule plan vs. actual production vs. gaps) drives improvement.
"We thought we were running to plan. Now we can see exactly where and why we're not."
Core capabilities to require
- Live schedule projection. "Order 1234 is 60% complete, projected to finish at 2:15pm, 30 minutes behind the plan." Requires integrated monitoring and scheduling.
- Capability-aware scheduling. Machine constraints (max width, max weight, material types) defined once and enforced everywhere.
- Sequence-dependent changeover optimization. Steel-to-steel changeover differs from steel-to-aluminum (10 min vs. 45 min). The system groups similar jobs.
- Multi-operation routing. Dependency chains (slit, cut to length, press brake) across machines. Operation 2 can't start before Operation 1 completes.
- Predicted durations from actual data. Historical data weighted by material, dimensions, machine, recency. Not manual estimates.
- Time-weighted schedule attainment. Measures by hours, not count. "8 of 10 jobs completed" hides the reality if eight were 1-hour jobs and the two 8-hour jobs got bumped.
Red flags
What to ask in the demo
The biggest ROI is in the connections.
System modules should work independently initially, but when connected, data flows automatically with no manual re-entry.
When monitoring detects a breakdown, it auto-creates a maintenance work order (pre-filled with machine, timestamp, operator description). When maintenance schedules a PM, it appears as an immovable block on the production schedule. A mid-job breakdown triggers automatic schedule adjustment: order block splits, maintenance block inserts, downstream orders shift. Planner sees impact immediately.
No manual re-entry. No spreadsheet gymnastics.
Simpler and cheaper than vendors make it sound.
Two practical approaches, and the right one depends on what's already on your machine. Both use off-the-shelf hardware. No proprietary anything.
Off-the-shelf industrial edge gateway reads PLC tags directly (Modbus, EtherNet/IP, PROFINET, OPC UA). Gateway publishes data outbound to platform. One gateway connects multiple PLCs.
~$1,000 per machine · Best for machines built in the last 20 years
Add simple sensors (current transformers, proximity sensors, pulse counters) to existing electrical signals, wired to a compact IoT gateway. No PLC access required. Works on any machine age.
~$500 per machine · Best for older legacy equipment
Cybersecurity: data flow is outbound-only. Gateway initiates outgoing encrypted connection. No inbound connections, open ports, VPN, tunnel, or remote gateway access. For sites without network access, gateways support cellular connectivity.
What to avoid
Who controls where your data lives?
Many IoT platforms route production data through vendor cloud. Machine telemetry, order data, and production history reside in vendor infrastructure. Worth asking early since vendors typically don't volunteer this.
Look for platforms with open APIs, full data export, and an architecture built for integration. Your production data should be accessible to any service you connect, not locked in a vendor's cloud.
Cost is part of it. What they build next is the rest.
Good vendor relationships start with active implementation, not just login credentials. Day-one specialist configures machine types, field mappings, state rules, KPI formulas, dashboards, and operator kiosk with your team. Your team walks away with a working system, not a project plan.
Post go-live, your team should run the system independently. Adding machines, changing KPI formulas, building dashboards, modifying downtime codes through an admin UI without vendor calls.
If the system can't provide scheduling capability, CMMS, and cost-to-serve analytics, you'll end up stitching together point solutions or starting over when complexity increases. Evaluate whether the vendor focuses on manufacturing and invests in next-year capabilities, not just today's needs.
Crawl, walk, run.
Phase 1: Weeks 1-2 (First machines live)
Connect 3-5 critical machines. Configure machine types, field mappings, state rules, KPIs. Deploy operator kiosk with downtime classification. Build first dashboards, validate data. Within the first week, downtime Pareto reveals actual time losses. First data-driven conversation occurs.
Phase 2: Weeks 2-8 (Expand and refine)
Add machines in waves. Refine KPIs and dashboards based on Phase 1 discoveries. Learn which metrics drive action vs. look good in theory. Build audience-specific views (executives: plant-level; supervisors: workcenter drill-down). Team comfort with data enables better questions and new view requests.
Phase 3: When ready (Schedule and optimize)
Connect monitoring to scheduling for closed-loop operation. Track schedule attainment. Enable duration predictions from accumulated history. Connect maintenance events to schedule. Become best-in-class.
Buyer evaluation questions.
What should I look for when evaluating a production monitoring platform?
Seven key criteria: configurable KPIs beyond OEE, two-tap operator downtime classification, closed-loop scheduling with live production progress, standard industrial hardware you own ($500 to $1,000 per machine), open APIs and full data export, day-one implementation with self-service admin after, and a vendor roadmap toward maintenance and cost-to-serve analytics.
What are red flags to watch for in a production monitoring demo?
Watch for: proprietary sensors or gateways that lock you to a vendor, KPIs limited to OEE without custom formula builders, pre-built dashboards with no chart builder, inbound network connections to PLCs required, no per-machine overrides for same-model equipment, and a schedule disconnected from live production progress.
How long does production monitoring software take to deploy?
A functional 5-machine pilot reaches production in about 90 days. A full 30-machine facility typically completes in 90 days for setup plus one more quarter for adoption and KPI tuning. Hardware installs during planned maintenance windows with no production disruption.
What is the cost difference between production monitoring and a full MES?
A typical MES implementation runs $1M to $3M plus 12 to 24 months based on industry-reported TCO benchmarks. A direct-to-PLC monitoring deployment runs a fraction of that and reaches a functional pilot in about 90 days. For most mid-market manufacturers, production monitoring plus scheduling replaces the 80% of MES capability that actually gets used.
The question isn't whether you need better visibility into your operation. The question is whether the system you're evaluating will actually give it to you.
Ask harder questions. Demand live demos. Trust the data, not the deck.
See MACH run on a real floor.
Not a polished demo environment. Not sample data. We'll show you exactly what the platform does, configured for your machine types, your KPIs, your operation.
Book a Diagnostic