A smart thermometer is a small sensor that reads water temperature and reports it to a dashboard on its own, day and night. For Legionella control that turns a once-a-month clipboard reading into a continuous record. The catch is the part the brochures skip: the sensor only measures a number. It does not run your control scheme, and it does not pick up the phone when something goes wrong.
So the real decision is narrower than “should we go digital”. It is which checks are worth automating, where the sensors should actually sit, and who acts when a reading goes red. Get those three right and the technology is genuinely useful. Get them wrong and you have bought a very expensive way to document your own neglect.
What a smart thermometer actually measures
Most systems are one of three things. A clip-on probe strapped to a pipe at the calorifier flow and return. A sensor at a sentinel outlet — the nearest and furthest taps on each hot and cold run, the points UK guidance treats as representative of the whole system. Or a probe in stored water, in the cold tank or the cylinder. Each one wakes up on a schedule, takes a reading, and sends it over a low-power radio link (LoRaWAN and NB-IoT are common in plant rooms) to a gateway, which forwards it to a dashboard. Set a threshold and the platform raises an alert when a reading falls outside it.
That continuous record is the real prize. Manual monitoring on hot and cold systems centres on temperatures at sentinel outlets and at the calorifier, with the figures and frequency set by your risk assessment [1][2]. The broad expectation is cold water staying genuinely cold — below roughly 20°C — and hot water reaching disinfection temperature at the tap soon after you open it, commonly cited around 50°C, with storage kept hotter still [2]. A monthly check tells you the temperature was right on the day someone visited. A sensor tells you whether it was right every hour in between.
One nuance decides whether the data is worth anything. A clip-on probe reads the pipe surface, not the water you would actually draw from the tap, and a sensor fitted downstream of a thermostatic mixing valve reads blended water — which can sit comfortably in range while the hot supply feeding it has quietly gone cold. Put the sensor where it proves the supply, not where the cable was easiest to run.
Where the brochure and the building disagree
Vendor decks and site reality part ways in fairly predictable places. These are the ones worth fixing in your head before a sales call:
| What the brochure implies | What actually holds on site |
|---|---|
| Installing sensors makes the building compliant | Sensors produce evidence. Compliance is the management system around them — risk assessment, written scheme, a competent person reviewing the data, and action on alerts [3] |
| Remote monitoring lets you drop the manual regime | It covers temperature at fixed points only. Tank inspections, showerhead cleaning and descaling, sampling, and flushing of little-used outlets still need a person |
| A wall of green dashboards means you are in control | Green means the measured points were in range. The value of the system is entirely in what happens when one goes red |
| More sensors equal more safety | A sensor at the wrong point, or a hundred dashboards nobody reads, adds noise. Placement at meaningful points beats raw count every time |
| Digital readings are automatically audit-proof | Only if they are calibrated, time-stamped, attributed to a specific point, and tamper-evident |
The mistake that turns a good sensor into a liability
In my experience the single most common error is buying the hardware before deciding who owns a red reading and exactly what they do about it. It feels backwards, but the alert is the easy bit. The hard bit is the closed loop: alert fires, named person sees it, person acts within a defined time, action gets recorded, repeat exceedances trigger a deeper look.
Skip that loop and you have built something worse than a paper gap. An unactioned alert is a time-stamped, attributed record that the temperature was wrong and nobody responded. A missing entry in a paper logbook is ambiguous. A logged exceedance you ignored for three weeks is not. If you are not ready to staff the response, you are not ready to install the sensors.
Where remote monitoring earns its keep
The technology pays back hardest in a few specific situations. Large or spread-out estates, where manually visiting dozens of sentinel points eats days and still leaves gaps between visits — continuous logging closes the gap and produces the monitoring records duty holders are expected to keep anyway [3]. Hard-to-reach or unpleasant points, like roof tanks and locked plant rooms, where a fixed sensor beats sending someone up a ladder every month. And early drift, where a calorifier slowly losing output or a cold feed warming through summer shows up in the trend line weeks before a quarterly check would have caught it.
It earns much less on a small site with a handful of outlets and a reliable person who already walks them. And it does nothing for the outlets that need physical flushing: a sensor will faithfully tell you a low-use shower is stagnant, but someone or something still has to run the water. That is where pairing monitoring with automated flushing turns a notification into an actual control. The root problem behind most of those alerts is the stagnation covered in — the sensor finds it faster, it does not fix it.
Calibration, signal and the parts vendors gloss over
Sensors drift. A probe that read true on install day will not necessarily read true a year later, so you need a verification routine — periodically check each one against a calibrated handheld at the same point and keep the comparison records. Without that, your “evidence” is just confident-looking numbers. Signal bites too: concrete plant rooms and steel risers swallow radio, and a sensor that cannot phone home is a sensor you are not monitoring. Ask how you export your data if you change provider, because the history is the asset, not the box on the wall.
None of this displaces a competent, site-specific assessment. A reading is one input. The temperatures you accept, the alert thresholds, and what counts as an adequate response to an exceedance are decisions for your risk assessment and a competent person — not defaults a supplier picks for you, and not figures to lift from an article. Treat any number here as orientation and confirm it against current HSE guidance for your own system; sampling, where you use it, follows the same rule, with frequency driven by the system and the assessment rather than the dashboard [4].
Where to start this week
You do not need a purchase order to make progress. You need your existing risk assessment and an hour.
List the points that already matter on it — the sentinel outlets, the calorifier flow and return, the storage tanks. For each one, write down the temperature you expect, the threshold that should raise concern, who would receive that alert, and the exact action they would take. If you can fill that table in honestly, you have a specification, and you will buy the right number of sensors for the right places. Then run a small pilot: instrument a handful of those points and read them alongside your manual checks for a month. If the two broadly agree and every test alert reaches the right person, you have proof the system works before you scale it across the estate. That single step also feeds straight into a wider water safety plan.
FAQ
Can remote sensors replace our monthly temperature checks?
For the specific points a sensor watches, continuous logging is generally stronger evidence than a single monthly reading — but only for temperature, and only if the device is calibrated and the data is reviewed by someone competent. Physical inspections, cleaning, descaling, flushing of little-used outlets and any sampling still need a person on site. Your risk assessment decides what can move to sensors and what stays manual [1].
Where should we actually place the sensors?
At the points that prove your control rather than the points that are convenient: sentinel outlets, calorifier flow and return, and stored water. Remember a clip-on probe reads the pipe, not the delivered water, and a sensor downstream of a mixing valve reads blended temperature — so position them where they reveal whether the supply itself is in range [2].
Will an inspector accept smart thermometer data as evidence?
It can carry real weight if the data is calibrated, time-stamped, attributed to a named point and tamper-evident, and if you can show you acted on the exceptions it flagged. Records of monitoring and your management arrangements are exactly the sort of thing duty holders are expected to keep and be able to produce [3].
Sources
[1] HSE, “Legionnaires’ disease: Technical guidance (HSG274)”. https://www.hse.gov.uk/pubns/books/hsg274.htm [2] HSE, “Hot and cold water systems”. https://www.hse.gov.uk/legionnaires/hot-and-cold.htm [3] HSE, “Legionnaires’ disease. The control of legionella bacteria in water systems - Approved Code of Practice and guidance (L8)”. https://www.hse.gov.uk/pubns/books/l8.htm [4] HSE, “Testing and monitoring your water system for legionella”. https://www.hse.gov.uk/legionnaires/testing-monitoring-water-system.htm