Walk any modern refinery, gas plant, or petrochemical complex and you will find more instrumentation than at any point in the history of the site. Vibration probes, flow meters, temperature points, pressure transmitters, gas detectors, and now cameras and edge devices stream readings by the second. The historian holds years of it. By every count that matters to a procurement team, the plant is well measured.
And yet the people who run it are not better informed. They are busier. The operator on shift has more screens, more alarms, and more dashboards than the operator who held the same seat a decade ago, and no more time to act. The instinct, when a decision is hard, is to add another sensor or another view. The result is a plant that is rich in telemetry and poor in signal. This is the gap that matters, and it is not a hardware problem.
The Telemetry Illusion
Telemetry is a measurement of a thing. Signal is something a person can act on. The two are easy to confuse because they arrive through the same pipe, but they are not the same, and the difference is where the value sits.
A pressure reading is telemetry. The same reading, placed against the asset it belongs to, the equipment history behind it, the operating envelope it is supposed to stay inside, and the decision a named person has to make when it drifts, is signal. The first is a number. The second is a number that has a job. Most industrial data programmes deliver the first at enormous scale and assume the second will follow. It does not follow on its own.
The illusion is that more telemetry must mean more signal. In practice the opposite often holds. Every additional uncontextualised feed raises the noise floor, adds another alarm to acknowledge, and pushes the decision the operator actually needs further down the screen. Volume of data and clarity of decision are not the same axis, and beyond a point they pull against each other.
Telemetry is a measurement of a thing. Signal is something a person can act on. They arrive through the same pipe, but they are not the same.
You can see the illusion most clearly at handover. The outgoing shift has the readings. What it cannot always pass cleanly to the incoming shift is what those readings mean, which ones matter tonight, and what was already tried. The telemetry survives the handover effortlessly. The signal often does not. A site can be fully instrumented and still lose the thread every twelve hours.
An Operating Ontology for the Plant Floor
Closing the gap starts before any automation. It starts with an agreement on what the things are. BOST works through three lenses, and the first, Marsad, is the operating picture: the ontology, the data model, the integrations, and the lineage that turn feeds into a structure a decision can rest on. Ontology before automation is not a slogan. It is the order of operations.
An operating ontology names the entities that the plant actually runs on and the relationships between them. Assets. Equipment. Events. Decisions. A pump is an asset with a model, a service history, and an operating envelope. A vibration reading is an event tied to that asset. An anomaly is an event that crosses a threshold the ontology already knows about. A response is a decision made by a role, against that event, with a reason attached. When the data is mapped onto that structure, a reading stops being a free-floating number and becomes a fact about a specific asset that a specific person is accountable for.
This is unglamorous work, and it is the work that most data initiatives skip. A model trained on telemetry that has no ontology behind it learns correlations without meaning, and it cannot tell an operator which decision a pattern belongs to. Lineage matters here too. When a number is wrong, and numbers are sometimes wrong, the operator needs to trace it back to the instrument, the integration, and the transform that produced it. An ontology with lineage makes the data answerable. Telemetry without it makes the data merely abundant.
The test of an operating ontology is simple. Can you take any reading on the floor and, without a meeting, say which asset it describes, which operating decision it informs, and who owns that decision. If you can, you have signal. If you cannot, you have telemetry wearing a dashboard.
Decision Audit Where the Work Happens
Signal that never reaches the point of action is wasted. The second lens, Maydan, is field action and the decision audit: the record of what was decided, by whom, on what basis, captured where the work happens. On the floor. At the unit. Not reconstructed afterward in a head-office dashboard from memory and a spreadsheet.
The distinction is sharper than it sounds. Head-office systems are good at reporting that a decision happened. They are poor at capturing the decision as it is made, with the operating context intact, because by the time the data reaches them the context has been stripped away. The operator who throttled a unit at 03:00 had a reason. If that reason is captured at the moment and the place, with the reading, the asset, and the alternative considered, it becomes part of the operating record. If it is captured the next morning in a summary, it becomes a number with the meaning sanded off.
A decision audit that lives where the work happens does two things at once. It makes the present decision better, because the operator is acting against structured signal rather than raw feeds. And it makes the next decision better, because the reasoning is now on the record, attached to the asset, available to the next shift and the next investigation. The audit is not paperwork. It is the mechanism by which a plant accumulates judgement instead of repeating it.
This is also where continuity is earned. The third lens, Mashhad, is operating continuity and handover, and it depends entirely on what the decision audit preserves. A site measures its data programme not by how much it collects but by what survives the handover. Telemetry survives by default. Decisions, and the reasons behind them, survive only if they were captured as signal in the first place.
From Signal to Action
The point of all of this is not a cleaner data model for its own sake. It is a shorter, more reliable path from an anomaly to a verified, logged response. That path is where the operating value lives, and it can be measured.
Consider the sequence the right way. An anomaly occurs. The system recognises it as an event against a known asset, inside an ontology that already knows the operating envelope and the responsible role. The signal reaches the person who can act, framed as a decision rather than a raw alert. The person acts, and the action is captured at the point of work, with its reason, against the asset. The response is verified and logged. The next shift inherits not just the reading but the decision and the reasoning behind it.
Each step in that sequence is something the right structure shortens and the wrong structure stretches. The time from anomaly to verified response is a real number a plant can track, and reducing it is a more honest goal than counting sensors or dashboards. A site does not get safer or more efficient because it measures more. It gets safer and more efficient because the distance between knowing and doing gets shorter, and stays shorter through every handover.
None of this requires the site to rip out its instrumentation or buy more of it. The sensors are mostly already there. What is usually missing is the operating ontology that turns their output into signal, and the decision audit that turns signal into action on the floor. That is the order BOST works in, and it is deliberate. Ontology before automation. Decisions before dashboards. Measure by what survives the handover.
The plant floor does not need more telemetry. It needs the few readings that matter, tied to the decision they inform, placed in front of the person who owns that decision, and recorded where the work is done. Sensors everywhere is a solved problem. Signal where it counts is the work that remains.