Most industrial digital transformation projects stall on a decision nobody writes down: whether the existing control system is the foundation of the new architecture or a tenant inside it. Get that one wrong and you are signing up for another forklift replacement in five years. Get it right and every system you add from here compounds the value of the last.
The ceiling problem
Walk into most plants and the SCADA platform is not just running the process. It is also the de facto integration layer, the historian, the reporting engine, the alarm system, and the place every new request gets bolted onto. That works until you want to do something the platform was never meant to do, like feed a cloud analytics model, stand up a digital twin, or share a clean stream of production data with a partner.
At that point the control system stops being a foundation and starts being a ceiling. Every new capability has to be negotiated through a vendor roadmap, a proprietary connector, or a custom export that breaks on the next upgrade. The operation cannot move faster than the slowest part of its stack, and the slowest part is usually the one holding the most critical data.
What "data source" actually means
The shift is simple to state and hard to do: demote the control system from the top of the architecture to a well-behaved producer of data inside it. CygNet, OASyS, Wonderware, iFIX, whatever you run, keeps doing what it is good at, which is talking to field devices and running the process safely. What changes is that its data flows outward into a shared layer that the rest of the enterprise consumes from, instead of every consumer reaching back into the control system one custom integration at a time.
This is not a rip-and-replace. The control system stays. What you are building is a contract around it, so that the next analytics tool, dashboard, or machine-learning model talks to the contract instead of to the platform.
The control system should feed the enterprise. It should never cap it.
The Unified Namespace as the contract
That contract has a name now: the Unified Namespace. A UNS is a single, structured, real-time representation of everything happening across the operation, organized by where it physically lives rather than by which system produced it. Every producer publishes into it, every consumer subscribes from it, and nobody needs to know which historian or PLC a given value originated in.
The transport that makes this practical is MQTT with Sparkplug B. It is lightweight enough for constrained field links, it carries state through birth and death certificates so consumers always know what is live, and it uses report-by-exception so you are not flooding the network with redundant polls. Your legacy SCADA becomes one more publisher into the namespace, and from the consumer's point of view it is indistinguishable from a brand-new edge device.
How to start without a forklift
The reason this approach wins is that you can begin small and prove value before committing to anything irreversible.
- Pick one site or one asset class and model its data in the namespace. Resist the urge to boil the ocean.
- Stand up a broker and publish the existing control-system data into the namespace alongside the live process, not instead of it.
- Point one new consumer, a dashboard or an analytics job, at the namespace rather than at the control system directly.
- Measure the integration effort for the next consumer. It should be a fraction of the first, because the hard work of modeling the data is already done.
Nothing in that sequence touches the safety-critical path. The control system runs exactly as it did. You are layering a modern data architecture on top of a stable foundation, which is the only responsible way to modernize an operation that cannot afford downtime.
The payoff
Once the namespace exists, the economics of every future project change. Integration stops being a series of brittle point-to-point connections and becomes a publish-and-subscribe pattern that scales linearly. Adding Ignition, a cloud historian, a Kafka pipeline, or a data lake becomes a matter of subscribing to a stream that is already clean, contextualized, and live. The operation can finally move at the speed of its ambitions instead of the speed of its oldest platform.
None of this requires abandoning the systems your operators trust. It requires a single decision, made deliberately and early: that your control system is the best data source in the building, and that its job is to feed everything above it rather than to be the thing everything has to route around.