How OT/IT convergence actually fails in brownfield plants

Every OT/IT programme I have been near started the same way: a pilot on one line, one furnace, one area. The pilot works. Data flows, a dashboard appears, someone senior is pleased. Then the rollout stalls, and the post-mortem blames integration complexity.

Having done this across brownfield steel plants in two countries, I think integration complexity is almost never the real cause. It is the thing that becomes visible when the actual causes are left unaddressed. There are four, and they are all organisational before they are technical.

1. The pilot was chosen because it was easy

Pilots get sited where the equipment is newest, the supervisor is enthusiastic and the tags are already documented. That is a sensible way to prove a concept and a terrible way to estimate a rollout. The second area has a twenty-year-old PLC with undocumented tag names, a sample rate that makes your analytics meaningless, and a supervisor who has watched three previous digital initiatives arrive and leave.

If you want a plan you can trust, pilot somewhere representative and awkward. The number you need is not “can this work” — it is “how long does one difficult area take”, because that is the number the rollout is made of.

2. Nobody owns the tag dictionary

This sounds like a detail and it is the single largest hidden cost in brownfield work. A plant that has grown over decades has instrument tags named by whoever installed them, in conventions that changed three times, with the same physical measurement appearing under different names in different systems.

Until someone owns the mapping — one named person, with the authority to settle disputes about what a signal means — every downstream analysis is a negotiation. I have seen more Industry 4.0 schedules slip on this than on any platform decision. Budget for it explicitly, as a workstream with a name, not as part of “integration”.

3. The safety culture is treated as an obstacle

In a plant, the reflex against unplanned change is not conservatism, it is the reason people go home intact. An IT team that arrives talking about rapid iteration and moving fast is speaking a language that, in that context, sounds like recklessness — and will be resisted for reasons the resisters may not articulate but are entirely right about.

The convergence that works runs in the other direction: adopt the plant’s change discipline for anything touching operations, and keep the fast-moving practice for the analytics layer sitting safely downstream of it. Two different tempos, deliberately separated. Trying to impose one tempo on both is how you lose the operations team, and without them you have a data project, not a transformation.

4. The KPI is agreed after the data arrives

The most common sequence is: build the pipeline, surface everything available, then ask the business what it wants to see. This produces dashboards that are impressive and unused, because the measure that would change a decision was never something the instrumentation captured.

Invert it. Start from the decision somebody wants to make differently — when to schedule the reline, which scrap mix to buy this week, whether that motor is degrading. Work back to the measurement that would change it, then instrument for that. It sounds obvious. It is also the step most often skipped, because starting from available data feels like progress and starting from decisions feels like meetings.

What good looks like

At the plants where this went well, the pattern was consistent: a representative pilot with an honest per-area estimate, a named owner for the tag dictionary, an explicit two-tempo change process, and a KPI agreed before the pipeline was built. None of that is a technology choice. The technology choices — which broker, which historian, which lakehouse — mattered far less than any of them, and were also the only part anyone argued about in the steering meetings.

The uncomfortable summary is that OT/IT convergence is an organisational programme wearing a technical costume. Staffed as a technical programme, it produces a very good pilot.