When to Scan, Automate or Integrate

Why this lesson matters

Most monitoring programmes fail on scope, not on technology. They either automate everything, stall on cost, and deliver nothing; or they manually capture everything, including points where a week of undetected failure is genuinely unacceptable. This lesson is the sorting method.

The three paths

Scan. Structured manual capture at the instrument. Identity, value, unit, timestamp, source image, person. Works on any instrument with a visible face. No power, no network at the asset, no certification, no modification to the instrument.

Automate. A sensor or retrofit device produces the value without a person. Pulse outputs, clamp on ultrasonic flow, current transformers, wireless temperature and pressure transmitters, optical register readers, LPWAN loggers.

Integrate. The value already exists in another system and simply needs to be brought in. BMS and SCADA points over BACnet or Modbus, PLC historians, utility interval data from the metering provider, the CMMS, submetering platforms.

Integrate is listed last but should usually be investigated first, because it is often the cheapest source of data on site and it is regularly overlooked. Half hourly electricity interval data for a NMI is frequently already available and unused.

The decision variables

For each asset, establish six things.

1. Required frequency. Match to the time constant of what you are monitoring. If the phenomenon moves faster than your sampling interval you will not see it, regardless of how good the record is.

2. Tolerable latency. How long can a problem go undetected before the consequence becomes unacceptable? This is the single most powerful discriminator and the one most often skipped.

3. Consequence of undetected failure. Safety, compliance breach, product loss, plant damage, energy cost, reporting gap. Quantify it roughly; rough is enough to sort.

4. Physical constraints. Hazardous area classification, metrological sealing, power availability, network reach, pit or roof access, landlord or lease restrictions, asset ownership.

5. Existing data sources. Is this point already in a BMS, a PLC, a submetering system or a utility feed?

6. Cost to deliver, all in. Not the sensor price. Device, mounting, containment, power, certification where applicable, commissioning, integration, and the recurring cost of connectivity and batteries.

The decision table

Condition	Path
Value already exists in BMS, SCADA, PLC or utility data	Integrate
Weekly or monthly frequency adequate, and a week of undetected failure is tolerable	Scan
Reporting obligation only, no operational urgency	Scan
Instrument sealed, in a hazardous area, or otherwise not modifiable	Scan, or a non invasive optical reader
Detection delay must be hours or less	Automate
Consequence of undetected failure is safety, compliance or major loss	Automate
Phenomenon is faster than a manual round can sample, such as short excursions or night leak signatures	Automate
High point density in one accessible location with power and network	Automate, economics improve with count
Remote, unpowered, low consequence, low frequency	Scan

The uncomfortable conclusion for most sites is that the majority of points belong on scan, a minority justify automation, and a surprising number are already available through integration.

Cost shape, honestly

Scan. Near zero capital. Recurring labour proportional to point count and frequency. Cost per point per read is low; total cost scales linearly with frequency. Doubling frequency doubles the labour.

Automate. Substantial capital per point, dominated by installation rather than hardware. Low marginal cost per reading once installed, so frequency is effectively free afterwards. Recurring cost in batteries, connectivity, and eventual device replacement. Devices fail silently, so it introduces a new maintenance obligation of its own.

Integrate. Capital is engineering effort rather than hardware. Point mapping, protocol work, tag naming, and validation. Marginal cost per additional point from the same system is very low, which is why it pays to map a system properly once.

The crossover is driven by frequency. At monthly readings, manual capture is cheaper than automation for a very long time. At hourly readings, automation wins almost immediately. Work out the required frequency first and the economics usually resolve themselves.

Mixing paths on one asset

These are not mutually exclusive, and the strongest configurations combine them:

• Automated electricity interval data for the incomer, plus scanned submeters on individual circuits.
• A continuous pressure transmitter for alarming, plus a scanned reading of the local gauge monthly as an independent check on transmitter drift.
• BMS integration for temperature, plus scanned gas register readings for the same boiler, so energy and condition sit against one asset.

That last pattern is worth noting: an independent manual check on an automated point is how you catch silent sensor drift, and it costs almost nothing.

Practical constraints that decide more cases than technology

• Who owns the meter. Utility revenue meters are usually not yours to modify. Ask before designing a pulse head into a scheme.
• Hazardous area. A certified installation in Zone 1 changes the cost of a point by a large multiple, and it adds inspection obligations under the wiring rules for the life of the installation.
• Network reality. Basements, pits, tank farms and metal clad plant rooms defeat many wireless technologies. Site survey before specifying.
• Battery logistics. A hundred battery devices at a five year life is twenty replacements a year, forever, on a schedule someone has to own.
• Who receives the alarm. An automated point with no defined recipient and no response procedure is capital spent on nothing.

A staged approach that actually completes

1. Inventory every point on one site with the six decision variables.
2. Harvest integration first. Find what already exists in BMS, PLC, submetering and utility data. This is often the single largest step forward for the least money.
3. Scan everything else immediately. You now have full coverage and a structured baseline within weeks rather than quarters.
4. Run for a quarter and let the data identify the automation candidates: the points with real variability, real consequence, or repeated exceptions.
5. Automate that short list, with a defined alarm recipient and response procedure for each point.
6. Re review annually. Plant changes, obligations change, and the classification should change with them.

This sequence has a useful property: every stage delivers usable data, and the expensive stage is specified by evidence rather than by assumption.

Common mistakes at this stage

• Automating before knowing the required frequency. The most expensive error available in this subject.
• Not checking what data already exists. Interval electricity data and BMS points are routinely paid for and unused.
• Costing the sensor rather than the installation. Installation, certification and commissioning usually dominate.
• Treating the three paths as sequential project phases rather than per asset choices.
• Automating a point with no alarm recipient or response procedure.
• Forgetting that automated devices need a maintenance regime, including batteries, calibration and offline detection.