Four Gates a Measurement Passes Before Any Sensor Is Priced

A sensor is a claim about the future.

Buying one asserts that some decision, some day, will be made differently because this measurement exists. That is a strong claim, and the useful thing about it is that it is testable — completely, in minutes, before a single wire is run or a single permit is raised.

Name the decision. Name who makes it. Name what value would change it, and roughly what changing it is worth. The sensor either has a job or it does not.

This test kills more bad projects than any technology review, and it costs nothing, because it is cheaper to fail a measurement on paper than in a junction box.

The four gates

Each gate is a question with a stop attached. A stop is a success of the review, not a failure of the project.

Gate one wants a named, recurring decision. Schedule the bearing renewal. Adjust the setpoint. Release or hold the batch. Dispatch the crew. Recurring matters, because a one-off investigation is answered by a portable instrument and a fortnight, not by a permanent point in a historian.

Gate two wants a role, not a department. The named role has to actually receive the data and have the authority to act on it. Two failure shapes hide here: the recipient who has the data and no authority, and the authority who never sees the data because it lives in a system they do not open.

Gate three wants the number that moves the action. This is the gate that exposes measurements bought to satisfy curiosity. If no value would change the decision, the point is decoration with a licence fee.

Gate four wants an order of magnitude, honestly stated. Not a business case — a rough figure in avoided cost, recovered capacity or reduced risk. It is there because the chain that delivers the measurement will have a cost, and the two numbers are going to meet whether or not anyone writes them down early.

The sentence that finishes or does not

The compressed form of all four gates is one sentence, and it is worth writing literally.

“When this value crosses ___, [named role] will ___ instead of ___.”

If the sentence cannot be finished, the sensor is not an investment. It is a souvenir.

The sentence is harder than it looks because it demands the counterfactual — instead of ___. Plenty of measurements survive the first three blanks and die at the fourth, because the honest answer is that the role would do exactly what they already do, just with more confidence. Confidence is worth something. It is rarely worth a chain.

What passing does not buy

A measurement can pass all four gates and still fail later. The sensor may not resolve the effect. The signal may be corrupted before it reaches a converter. The data may arrive too late to act on. The threshold may prove wrong once real behaviour is observed.

Those are honest engineering failures, and reducing them is what the rest of an architecture discipline is for — signal families, sensor selection against duty, wiring and isolation, acquisition rates, network capacity, timestamps, storage settings.

A measurement that fails the test now has nowhere to fail later, because it was never going to change anything.

The census that makes the queue honest

Running the gates on one candidate is easy. Running them across the wish list is where a programme is actually decided, and the frame that survives is a portfolio rather than a coverage percentage.

Every candidate carries a decision, an owner and an indicative value basis, and the list is sorted by value. Below the funded rows sits one more, and it is the most useful line on the page: everything else — queued until a decision and an owner exist.

That row works only if the waiting list is public. Anyone may promote a candidate by supplying the missing decision, owner and value. Nobody may promote a candidate because the hardware is inexpensive.

There is a second census worth running alongside it, and it usually pays faster than any purchase. Most brownfield plants own smart instruments used as if they were dumb — transmitters whose digital diagnostics are stranded at the loop, meters and drives with registers nobody reads, whole motor control centres that never appear on an instrument index. At Meridian, the invented plant the underlying reference runs on, the illustrative census found 140 HART transmitters of which precisely zero had their digital diagnostics read by anything.

Whether liberating that data is worth the integration cost is a real question with a real answer. But no answer is possible until the census makes the stranded value visible, and the census costs a spreadsheet.

An exercise that costs an afternoon

Take the three measurements most recently added to your plant — anything installed or connected in the last year, whether or not it was called a data project.

Run all four gates on each, retrospectively, using only what is written down.

• ✓ Gate one: write the decision as a verb with an object. If the strongest available answer is a noun like 'visibility' or 'monitoring', record that as the answer rather than improving it.
• ✓ Gate two: write a person's role. Then check whether that role has opened the display in the last month — the telemetry usually exists and is rarely consulted.
• ✓ Gate three: name the value that would change the action, and find out whether anyone has ever crossed it. A threshold never crossed is either a healthy plant or a wrong number.
• ✓ Gate four: write one figure with a unit. Order of magnitude is fine; a blank is the finding.
• ✓ Then finish the sentence: when this value crosses ___, [role] will ___ instead of ___.

Two outcomes are common. The measurements pass gates one and three and stall at gate two, because the named role turns out to be whoever asked for it rather than whoever decides. Or all four gates pass on paper and nobody has ever acted, which points at the response path rather than the measurement — and that is a Part 8 problem, not a Part 1 one.

Neither finding costs capital. Both change what the next purchase should be.

For readers still deciding which layer of a modernisation programme to fund at all, the IoT, IIoT and Industry 4.0 decision framework separates control reliability from data capture from transformation before any of these gates apply.

Which of your funded measurements would fail gate two if you asked the named role by name?

The four-gate decision test, the decision sentence, the portfolio framing with its public waiting list, and the stranded-diagnostics census are from Industrial IoT and Data Architecture: From Sensor to Historian to Dashboard (Part 1: The Sensing Layer, sections 1.1–1.2).