An IIoT readiness checklist for industrial operators.

Most stalled IIoT projects fail at the cable, not the cloud. Segregate OT from IT, profile every device, and design for deterministic latency before adding a single sensor.

• Plant network segmented into Purdue-style zones with documented conduits between Levels 2, 3 and 3.5.
• Industrial DMZ (iDMZ) terminating every north-south flow — no direct PLC-to-cloud connections.
• Wireless coverage surveyed for 2.4 GHz / 5 GHz / private LTE blind spots in process areas.
• Failure-mode tested: what happens to control loops when the uplink drops for 60 seconds?

An IIoT rollout doubles your attack surface overnight. Treat IEC 62443 as the baseline, not the ceiling, and make sure incident response covers the production network too.

• Asset inventory complete for PLCs, RTUs, HMIs, drives and gateways — including firmware versions.
• Role-based access on engineering workstations; no shared admin credentials on the plant floor.
• Patch and backup cadence agreed with operations — including offline golden images of every PLC.
• IR runbook explicitly names OT containment steps (isolate, freeze, fall back to manual control).

Data without context is noise. Standardise tag naming, units and metadata at the source — before the first dashboard goes live — or you will rebuild the historian twice.

• Tag naming convention published (ISA-95 / ISA-88 aligned) and enforced in engineering reviews.
• Unified Namespace (UNS) or equivalent semantic model agreed with process and reliability teams.
• Time synchronisation via NTP/PTP across PLCs, gateways and historian — drift under 100 ms.
• Retention policy defined per tag class: high-resolution, aggregated, and cold archival.

The edge is where physics meets software. Choose protocols that survive intermittent links and gateways that can be remotely managed, updated, and observed.

• MQTT Sparkplug B (or equivalent) for store-and-forward telemetry with birth/death certificates.
• OPC UA for richer asset models where legacy DCS / PLC vendors support it natively.
• Edge gateways provisioned via zero-touch enrolment with signed firmware updates.
• Local buffering proven for at least 24 hours of disconnected operation.

Telemetry only creates value when it lands in the systems that drive decisions — CMMS for maintenance, ERP for production, EMS for energy. Plan the integration before procurement.

• CMMS (e.g. SAP PM, Maximo) integration scoped for work-order generation from condition alerts.
• ERP touchpoints identified for OEE, batch genealogy and material consumption.
• Energy management system fed sub-metered loads for ISO 50001 and Section 12L reporting.
• Event bus or middleware (Kafka, MQTT broker, iPaaS) chosen — not bespoke point-to-point.

Pilots die when nobody can defend their ROI. Pick two or three use cases with hard baselines and a named operations sponsor before any sensor is procured.

• Top three use cases prioritised by annualised value (downtime, yield, energy, safety).
• Pre-rollout baseline measured for at least 90 days on the target asset or line.
• Success metrics agreed with finance — measured in rand, not dashboard views.
• Stage-gate plan defines what triggers scale-up from pilot to plant-wide rollout.

Technology is the easy part. Without a control owner, a data steward and an empowered reliability engineer, the platform decays within twelve months.

• Named platform owner sitting between IT, OT and operations — with budget authority.
• Reliability and process engineers trained on the analytics tools, not just the dashboards.
• Change management process covers tag additions, model retraining and dashboard publishing.
• Vendor lock-in risks reviewed: data export, model portability, contractual exit clauses.