Temperature Humidity Monitoring System — Cost and Payback

In Thailand, the metric you have to manage flips twice a year

If you are considering a temperature and humidity monitoring system for a plant in Thailand, there is one fact worth getting straight before anything else. In the same factory, twice a year, the metric that actually causes trouble changes. There is a season in which temperature hurts you, and a season in which humidity hurts you, and the two are cleanly separated.

The Thai Meteorological Department (TMD) issued a warning for 7 to 11 August 2026, covering the North, Northeast and East, for heavy to very heavy rain along with flash floods and overflowing waterways, driven by a monsoon trough and a strong southwest monsoon. The same agency’s outlook for the 2026 hot season said the season would begin at the end of February 2026 and run to mid-May, with maximum temperatures possibly reaching 42 to 43°C in April and May. Average maximum temperatures in the North were forecast at 36 to 37°C against a normal of 35.4°C, with rainfall 30 to 40% below normal.

Put those two statements side by side and it becomes clear that factory environmental control does not have the same face all year. In the hot season, temperature is what bites. Air-conditioning capacity runs out, panel interiors heat up, and heat load on operators rises. In the rainy season, humidity is what bites. Dew point, condensation, moisture uptake. The same rack in the same warehouse is a place that is too hot in February and a place that is too humid in August.

This is why the design habit of picking one threshold and posting it for the whole year fails in one season or the other. A threshold tuned for the hot season stops firing in the rainy season, and a threshold tuned for the rainy season never stops firing in the hot season. It is the first thing projects trip over, and it is the kind of problem that adding more sensors does not solve.

Here is the conclusion of this article up front. The hard part of temperature and humidity monitoring is not measuring and storing data. The hard part is deciding which excursion counts as an anomaly, and deciding who corrects it, within what time, and how that correction gets written into the record. Cost and benefit both concentrate there. In what follows we use a Japanese-owned electronic components plant in Rayong, Thailand as a model, break a 40-point, 1,300,000 baht investment into five layers, and follow the numbers to see where the money goes and where the return actually comes from.

What a temperature and humidity monitoring system is — the three layers of measure, judge and correct

In practice, the phrase “temperature and humidity monitoring system” is used for three completely different things. Quotations get requested while those three are still mixed together, which is why the conversation never converges. Separated out, they look like this.

Layer 1, measure. Temperature and humidity sensors, wireless or wired communication, gateways, and somewhere to keep the data as a time series. This layer replaces the practice of reading a wall-mounted thermo-hygrometer twice a day and copying it onto paper with a mechanism that records continuously and automatically. Technically it is a solved problem, and it goes in fast. Getting sensors installed and a chart on screen can take one to two weeks depending on scale.

Layer 2, judge. This layer decides which of the recorded data counts as an excursion. Is anything above 60%RH an anomaly, or only 60%RH sustained for a defined length of time? Does a spike caused by a door opening count or not? Do thresholds switch by season? This looks like configuration work, but it is actually the quality assurance department’s judgement written down.

Layer 3, correct and record. This layer defines who does what within how many minutes when an excursion occurs, and what gets written into the record. It also puts calibration and traceability in place so that the record means something to a third party. Only at this point does accumulated data stop being a log and start being evidence.

The three layers stack, and they can only be built from the bottom up. Investment logic, however, runs the other way. Layer 1 can be built quickly and returns almost nothing. Layer 3 takes time and costs more, yet nearly all of the benefit originates there. That is why a project that stops at Layer 1 ends up in the strange state of having succeeded as an installation while producing no result.

There is one more practical note. The sensors and gateways in Layer 1 cannot be placed without regard to existing network and power infrastructure. The same design problems you meet when pulling signals off older machines show up here, so a plant already looking at retrofitting IoT onto legacy equipment should design environmental monitoring as part of that plan and get the cabling work done once.

Why so many projects stop at Layer 1

Stopping at Layer 1 is not a failure of will. It is structural, and the way a quotation looks on paper is what causes it.

In the model below, the initial cost for 40 monitoring points totals 1,300,000 baht. Of that, the sensors themselves account for 180,000 baht, or 13.8% of the whole. The remaining 86.2% is communication and power installation, collection and record configuration, corrective-action workflow design and training, and calibration and verification. In other words, the sensor cost that comes to mind first when someone says “the cost of a temperature and humidity monitoring system” is roughly one seventh of the total.

The trouble is that sensor cost is easy to state. Forty points at 4,500 baht, one line. By contrast, “240,000 baht for corrective-action workflow design and training” is hard to explain on a quotation sheet in terms of what is being bought. The result is that the proposal easiest to get approved internally is the Layer 1 only version, and Layers 2 and 3 get deferred to “let us install it first and think about that later.”

And later rarely arrives, because running Layer 1 on its own produces three predictable effects on the floor.

• Charts exist, but nobody goes to look at them. There is no reason inside anyone’s work procedure to go and look.
• Thresholds stay at their provisional values, so alarms either fire constantly or never fire at all.
• Paper records continue in parallel. With no basis for treating the automatic record as authoritative, the paper that audits rely on cannot be thrown away.

The third is the one most often missed. If the automatic record cannot be used as evidence, then records multiply while transcription work stays exactly where it was. Labour savings are zero and the only new line item is a SaaS subscription. That is the anatomy of the state where records pile up while defects and complaints do not fall.

Put differently, Layer 1 is not an investment. It is a precondition. Projects fail to earn a return because they bought only the precondition and then expected a return from it. Whether a team recognises this at the outset largely determines how the project ends.

Monitoring points fall into five categories with different accuracy and record-keeping demands

Treating the task as “monitoring temperature and humidity in the factory” makes design impossible. Monitoring serves five distinct purposes, and each category needs different accuracy, a different sampling interval, a different record retention period, and a different first response when an excursion occurs. The same sensor can measure them, but they are not the same operation.

Using this table is simple. Before requesting a quotation, sort your own monitoring points into the five categories, then count how many fall into categories 1 and 2. Benefit comes almost entirely from those two, so that count is the denominator of the investment decision.

Sorting the model plant’s 40 points gives 12 around the lines (category 2), 18 in the raw material warehouse (category 1), 6 in constant temperature, humidity and dry storage (category 1), and 4 for outdoor reference (category 4). Category 3, the inside of control panels, is currently not monitored at all. Statutory measurement for category 5 cannot be replaced by temperature and humidity sensors, which we return to later.

What deserves attention here is that the category with the most points is not the category with the most benefit. The 18 points in the raw material warehouse are the largest count, but if what is stored there is not actually being damaged by humidity, the benefit from monitoring those 18 points is close to zero. The 6 points in dry storage, on the other hand, may capture most of the available quality benefit if moisture-sensitive devices (MSD) are kept there.

This is why an instruction to “make the whole factory visible” maximises cost and minimises benefit. Sort by category, and start with the places in categories 1 and 2 where something is genuinely being damaged. That sequence is the argument this article keeps returning to. The design of quality records themselves is covered in our article on quality data management systems, which is a good companion read if you want to start from the record format.

Layer 2 — what counts as an excursion, and why instantaneous thresholds stall the floor

Layer 2 is the smallest layer in implementation effort and the largest in consequence of failure. There are only two ways to fail here. Too many alarms, or none.

The classic cause of too many alarms is judging on instantaneous values alone. Open a warehouse door and humidity spikes. A dehumidifier defrost cycle makes it spike. Put a sensor where air moves across it and the reading wanders. Judge on instantaneous values and every one of these non-anomalous fluctuations becomes an alarm. An operation that generates 80 alarms a day is, from the floor’s point of view, identical to one that generates none. Nobody looks, and those who look do nothing. Silencing alarms becomes the job.

The classic cause of no alarms is the opposite, thresholds loosened too far. Someone raises the threshold after a hot season of alarms that would not stop, and then the rainy season arrives with the loose setting still in place. Or the team hunts for a value that works year round and settles on one that fits no season at all. Records accumulate, and excursions go undetected.

There is one direction out of this, and it is to judge on the duration of an excursion rather than on instantaneous values. Define the anomaly not as “exceeded 60%RH” but as “remained above 60%RH for longer than a defined period.” Choosing that period is the design work of Layer 2, and it is where standards enter the picture.

For a plant handling electronic components, the reference is IPC/JEDEC J-STD-033. The standard sets the baseline condition for the floor life of moisture-sensitive devices (MSD) at 30°C or below and 60%RH or below, and under those conditions floor life is 168 hours for MSL3, 72 hours for MSL4, 48 hours for MSL5, and 24 hours for MSL5a. The most recent widely circulated revision is J-STD-033D, published in April 2018.

The important point is that in environments exceeding 30°C or 60%RH, the derating defined by the standard, meaning the reinterpretation of floor life for actual factory conditions, becomes necessary. In an SMT plant in Thailand during the rainy season it is not unusual for measured humidity to exceed 60%RH even with air conditioning running. Which means the assumption that MSL3 parts are good for 168 hours is quietly failing in plants that keep no records. Nobody notices the failure because nobody is measuring the moment it happens.

Translated into practice, Layer 2 design comes down to deciding four things.

• Thresholds per category. Derive categories 1 and 2 from the standard, category 3 from dew point, category 4 from running cost.
• The duration that counts as an excursion. Long enough to exclude fluctuation from door openings and equipment cycles.
• Seasonal switching. Decide in advance whether thresholds or durations change between the hot and rainy seasons.
• Treatment of cumulative exposure. Whether short excursions are accumulated for judgement, or only single events are considered.

Whether those four fit on one sheet of paper is the test of whether Layer 2 exists. Go live without them and you inherit an operation that adjusts alarm setpoints every month.

Layer 3 — corrective action and records, where 41.5% of the initial cost turns records into evidence

Layer 3 has two components. The corrective-action procedure, and the evidential quality of the record.

The corrective-action procedure is the path that follows detection. Who receives the notification. Within how many minutes that person is on site. What they check when they get there. Where the result of that check is written. How the affected lot is handled. Who makes the disposition. If that path is not written into a work procedure, the notification ends at “seen.”

Corrective procedures that work on the floor share one characteristic. The first response is simple and requires no judgement. “On receiving a humidity excursion notification for the dry storage cabinet, first check that the door is properly closed and check the state of the desiccant, then write the time of the check and the condition into the record.” Anyone can do that much. Judgement calls such as lot disposition belong to the next stage. Mix judgement into the first response and the response stops.

Evidential quality is the state in which the record means something to a third party. This is where calibration and traceability become necessary. Practitioner guidance on temperature and humidity mapping for pharmaceutical warehouses holds that data loggers should carry traceable calibration certificates, that calibration should generally be within the last 12 months, and that error should be within ±0.5°C at each calibration point and within ±4%RH for humidity. On placement, the same guidance cites grids at 5 to 10 m intervals in large warehouses, vertical placement at low, middle and high positions such as 0.5 m, 3 m and 6 m, and a minimum of seven days of continuous measurement in ambient warehouses in order to capture operating cycles. These points come from practitioner commentary aligned with GDP and WHO guidance, and the level actually required varies by industry and by the authority involved. For calibration itself, it is enough to plan around ISO/IEC 17025 accredited calibration.

Why spend money here? The cost breakdown makes it obvious. Of the model plant’s 1,300,000 baht initial cost, calibration and verification account for 300,000 baht and corrective-action workflow design and training for 240,000 baht. Together that is 540,000 baht, or 41.5% of the total. That 41.5% is what it costs to turn accumulated data from a log into evidence.

Read the other way, a quotation with that 41.5% stripped out is a quotation for data with no evidential value. The data exists but cannot be shown in an audit. Because it cannot be shown, the paper records stay. Because the paper stays, the labour does not fall. Whether that chain is about to happen is visible at the quotation stage.

For food plants, the record requirement arrives more explicitly from the regulatory side. Thai Ministry of Public Health Notification No. 420 (B.E. 2563) sets GMP requirements for food production methods, production equipment and storage, applying from 11 April 2021 for new manufacturers and importers and from 7 October 2021 for existing operators. Records must be maintained on the prescribed forms, numbered TorSor 1 to 5, and annual testing records from an accredited laboratory are also required. This article does not address specific numerical temperature or humidity requirements in that notification, but the existence of record-keeping and storage requirements is clear, and it is safer to design automatic environmental records so they can connect to those forms. Record design for food plants is covered in more depth in our article on food factory traceability.

The five-layer cost breakdown (Scenario A, 40 points, 1,300,000 baht)

Now to the numbers. The model is a Japanese-owned electronic components plant in Rayong, Thailand. Two SMT lines, a raw material warehouse and a constant temperature and humidity storage room. Twenty-five operating days a month on two shifts. Forty monitoring points, allocated as 12 around the lines, 18 in the raw material warehouse, 6 in constant temperature, humidity and dry storage, and 4 for outdoor reference. Today a wall-mounted thermo-hygrometer is read twice a day and copied onto paper. The internal labour rate used throughout is 450 baht per hour.

Scenario A monitors all 40 points across the whole site at once. Breaking the initial cost into five layers that match units of installation and work gives the table below. The mapping back to the three-layer model is this. Cost layers 1 and 2 are “measure” at 38.5% combined, cost layer 3 is “judge” at 20.0%, and cost layers 4 and 5 are “correct and record” at 41.5% combined. To avoid confusion, each cost layer below carries the name of its content.

The most striking figure in this table is that Layer 1, sensors and measuring ends, is only 13.8%. Negotiating 10% off the sensor unit price moves the total by 1.4%. Delete Layer 4, corrective workflow design and training, and Layer 5, calibration and verification, entirely and 41.5% disappears. Where the negotiating room sits and where the results come from are two completely different places.

The 320,000 baht for Layer 2, communication and power work, may feel unexpectedly large. But in Thai factories, monitoring points frequently land between metal racks where wireless does not propagate, or high in a warehouse where there is no power nearby. Gateway placement, access point tuning and power routing determine cost more than the point count does. This is the part that cannot be quoted from a desk without a site survey, and equally the layer where survey quality moves the number the most.

Next comes annual running cost. Judging this project on initial cost alone guarantees a misjudgement.

An annual 189,200 baht becomes 946,000 baht over five years. Added to the initial 1,300,000 baht, the five-year total is 2,246,000 baht, of which the initial cost is 57.9% and the remaining 42.1% flows out through operations. This is precisely why the approach of approving capital expenditure and leaving running cost to the site budget breaks down.

Annual calibration covers 20 points rather than all 40 because the intent is to prioritise points in categories 1 and 2 and exclude items such as outdoor reference from the calibration scope. This varies with industry and audit expectations, so decide explicitly at quotation time how many points will be calibrated each year.

Payback and Scenario B (12 points) — the more points you monitor, the further payback recedes

After cost comes benefit. Splitting Scenario A’s annual benefit into four items gives the following.

The 3.95 baht per kWh used for the electricity conversion comes from reporting that the tariff for the May to August 2026 billing period was set with an Ft of 0.1623 baht per unit and an average rate of 3.95 baht per kWh, reported as a decision of the Energy Regulatory Commission. Government support measures holding the first 200 units below 3 baht per unit have also been reported, but this model uses 3.95 baht per kWh consistently as the effective factory rate. If the goal is to optimise air conditioning and dehumidification operation itself, it is faster to get the electricity breakdown first through a factory energy monitoring system than through environmental monitoring.

Of these four items, the quality-driven benefits, 168,000 for re-baking plus 180,000 for complaints, come to 348,000 baht, which is 72.4% of the total. Labour and electricity savings together are only 27.6%. This investment is therefore fundamentally a quality investment, not a headcount-reduction or energy-efficiency investment, and the internal proposal should be written accordingly.

Now the payback. Annual net benefit is 480,480 minus 189,200, which is 291,280 baht. Simple payback is 1,300,000 divided by 291,280, roughly 4.5 years. Five-year ROI is (291,280 × 5 − 1,300,000) ÷ 1,300,000, which is +12.0%. Barely positive after five years. As a piece of factory capital expenditure, that is not an attractive number.

So for comparison, consider Scenario B with a reduced point count. The scope is 6 points in dry storage plus 6 points around the SMT lines, 12 in total. In other words, only the places where humidity is genuinely damaging something.

What stands out is that cutting the point count by 70%, from 40 to 12, only cuts initial cost roughly in half, from 1,300,000 baht to 664,000 baht. Sensors scale with point count, but collection and record configuration and corrective workflow design and training are the cost of designing the factory’s own operating procedures, and they barely scale with point count at all. That non-proportionality produces the next conclusion.

Annual cost for B is 60,000 for SaaS and maintenance plus 15,000 for calibration (6 points × 2,500) plus 21,600 of internal labour (4 hours per month × 12 × 450), giving 96,600 baht. The five-year total for B is 664,000 plus 96,600 × 5, which is 1,147,000 baht. Benefit for B is 168,000 in reduced re-baking plus 90,000 in reduced complaints, half of A, plus 30,000 in audit labour plus zero for electricity, giving 288,000 baht. Net benefit for B is 288,000 minus 96,600, which is 191,400 baht. Payback for B is 664,000 divided by 191,400, roughly 3.5 years, and five-year ROI is (191,400 × 5 − 664,000) ÷ 664,000, which is +44.1%.

The investment differs by a factor of about 2.0, 1,300,000 baht against 664,000 baht, yet net benefit differs by only about 1.5, 291,280 baht against 191,400 baht. That is why A, with more points, pays back more slowly. Saying that the more points you monitor, the further payback recedes is just another way of stating this asymmetry.

Benefit exists only near the thing that is breaking. Monitor 18 points in the raw material warehouse and no benefit appears unless humidity-driven defects are occurring there. So before adding points, identify what is breaking. Reverse the order and only cost scales with the point count.

Sensitivity analysis — a drop to 70% realisation alone pushes the five-year ROI negative

Every calculation so far carries one implicit assumption, that the benefits in the table are realised in full. In reality they are not. Alarms fire without corrective action following, and corrective action happens without translating into fewer defects.

Call that proportion the benefit realisation rate and flex it for Scenario A.

This table is the core of the article. A drop in realisation to 70% alone stretches payback from about 4.5 years to about 8.8 years and drops the five-year ROI from +12.0% to −43.4%. At 50%, payback is about 25.5 years and five-year ROI is −80.4%, close to a total write-off.

Why is sensitivity this high? Because the 189,200 baht annual running cost leaves the business every year as a fixed cost. When benefit falls by 30%, the 144,144 baht of lost benefit comes straight off net benefit, which nearly halves from 291,280 baht to 147,136 baht. Simple payback is initial cost divided by net benefit, so halving the denominator doubles the years. In an investment with heavy fixed costs, a small fall in realisation pushes payback out a long way.

So what determines the realisation rate? Going back to the definition, it is the proportion of alarms that are actually followed by corrective action which in turn leads to fewer defects. Which means the number is a direct reflection of how well Layer 3, correct and record, has been built. In a deployment that skips this layer, assume the proportion sits below 50%. Notifications go out but there is no path to follow, so all that ends up in the record is the fact that an excursion occurred.

One practical conclusion follows. Decide who will run the corrective-action workflow before you request a sensor quotation. Can one person in quality assurance be the receiving point? What happens on night shift? Will corrective records be appended to an existing form or will a new one be created? If those are unsettled, expected value is higher if you cut the point count and shift budget toward Layer 3.

Conversely, in a plant where Layer 3 already exists, realisation goes up. If an existing defect-response workflow is functioning and all that is needed is to feed environmental excursion notifications into it, a realisation rate close to 100% is reasonable to assume. The same capital expenditure lands at +12.0% or at −80.4% depending on the organisational capability of the plant. That is the real shape of the investment decision in environmental monitoring.

Three issues specific to Thailand

Everything so far holds as a general argument. Here are three issues specific to factories in Thailand.

First, dew point and condensation in the rainy season. What causes problems in a humid environment is not humidity itself so much as temperature difference. When outdoor air with a high dew point reaches a chilled metal surface or an air-conditioned room, it condenses on the surface. The inside of a control panel is a place where this readily happens. Operation stops overnight, the panel interior cools, high-humidity morning air enters, and water forms on the board. Condensation inside a panel is invisible from outside, so the event is usually filed as an unexplained stoppage.

Countermeasures include panel heaters, vents and gasket degradation management, but many plants have never measured whether it is happening at all. Treated as category 3, equipment protection, rather than category 4, facilities, a temperature and humidity sensor inside the panel makes the condensation risk windows visible through dew point calculation. Point counts here are low and the benefit shows up as reduced downtime, which makes the return relatively easy to read.

Second, MSD floor life failing quietly. As noted, J-STD-033 sets the baseline condition for floor life at 30°C or below and 60%RH or below. In the Thai rainy season it is not unusual for measured humidity to exceed 60%RH even with air conditioning. If there are periods above that, the assumption that MSL3 gives you 168 hours does not hold.

The awkward part is how quietly it fails. Components do not break the instant humidity passes 60%RH, so nothing happens on the floor. Weeks later popcorning appears at reflow and the cause is filed as lot variation. With no records, there is no way to test it against humidity. Continuous records earn their value precisely in this ability to go back and verify after the fact. Which also means that without a work procedure for going back and verifying, records produce no value.

Third, WBGT cannot be substituted with temperature and humidity sensors. This is worth stating explicitly as a caution. Thailand’s thermal environment standard is set by the Ministry of Labour’s ministerial regulation on heat, light and noise, B.E. 2559 (2016), which uses WBGT, the wet bulb globe temperature, as its index and sets limits according to workload. For moderate work, WBGT 32°C is cited as the corresponding value. Where the standard is exceeded, engineering controls such as ventilation and cooling are expected, and where those are difficult, warning signage and provision of protective equipment are required.

An academic review in Safety and Health at Work in 2021 reports that this standard was first promulgated in 1976 and amended in 2016, that compliance would protect four fifths of workers while non-compliance drops that proportion to roughly half, and that gaps and ambiguity in the law are cited as reasons for non-compliance.

The critical point is that WBGT is an index that includes globe temperature, meaning radiant heat. Ordinary temperature and humidity sensors do not measure radiant heat, so they cannot compute WBGT. Installing a temperature and humidity monitoring system therefore does not constitute compliance with heat regulations. It can be used to understand trends in perceived conditions on the floor and to prioritise heat-illness countermeasures, but statutory measurement is done with a WBGT meter. If you receive a proposal that conflates the two, scrutinise it.

The implementation sequence in four steps, and five items to settle before requesting a quotation

Get the order wrong and cost scales with point count while benefit does not. The sequence that works in practice is these four steps.

Step 1, identify what is breaking. Allow two weeks. Cross-reference existing wall-mounted thermo-hygrometer records against defect and complaint records. The goal is to express the annual cost of humidity- and temperature-driven defects and complaints as a single number. Precision is not needed. Getting the order of magnitude right is enough. This work requires no new investment.

Step 2, start with around 12 points. Place monitoring points only in the locations identified in Step 1. Build Layer 2, the excursion definitions, and Layer 3, the corrective workflow and calibration, at the same time. The temptation to expand across the whole site appears here. Do not expand. The scope over which realisation can be held near 100% is the ceiling at this stage.

Step 3, run it for three to six months and measure realisation. Count the excursions detected, how many of those had corrective action recorded, and how many led to fewer defects. The realisation rate that comes out is the basis for the expansion decision. If it is below 50%, rebuild Layer 3 before adding points.

Step 4, expand only where realisation is confirmed. Going to 40 points can wait until realisation is confirmed at 12. Follow this order and the first stage pays back at Scenario B economics while still scaling toward the equivalent of Scenario A. If you want to think about IoT monitoring rollout from the overall architecture down, see our article on factory IoT implementation as well.

There are also five items worth settling internally before requesting a quotation. Anything left unsettled hands that decision to the vendor.

• The result of sorting monitoring points into the five categories, and the count in categories 1 and 2. This is the denominator of the benefit.
• The excursion definition. The threshold, and the duration that counts as an excursion. Whether thresholds switch by season.
• The first responder for corrective action, and the response time limit. Decide this including night shift.
• The number of points subject to annual calibration and its frequency. Tie this to audit expectations.
• Whether the automatic record becomes authoritative or paper continues in parallel. Labour savings are decided here.

The last of the five moves the money most. If paper is going to continue in parallel, the 75,600 baht in audit and record-preparation labour savings cannot be counted. That removes 15.7% of the total benefit, and payback visibly recedes.

Frequently asked questions

What is a temperature and humidity monitoring system?

It is the general term for a mechanism that measures temperature and humidity continuously, records them, issues a notification when conditions fall outside a predefined range, and keeps the record of corrective action through to completion. In practice it is easier to design when split into three layers. Layer 1 measures, covering sensors, communication and the record destination. Layer 2 judges, deciding what counts as an excursion. Layer 3 corrects and records, defining who does what and when, what gets written into the record, and how calibration gives the record evidential value. Many proposals use the term to mean Layer 1 alone, so when comparing quotations, check layer by layer how far the scope extends.

Should temperature and humidity records be automated?

Automation on its own has limited return. In the model calculation, savings in audit and record-preparation labour are 75,600 baht a year, only 15.7% of total benefit. Eliminating transcription work by itself produces about that much and no more. Automation becomes meaningful when the record carries evidential value so paper can be retired, and when the data can be cross-referenced against defects retrospectively. The first requires calibration and traceability, the second requires a work procedure for doing the verification. Rather than asking whether to automate, first decide whether the automatic record can become authoritative so paper can stop.

How densely should temperature sensors be placed?

The more points you add, the further payback recedes. In the model calculation, the 40-point Scenario A pays back in about 4.5 years with a five-year ROI of +12.0%, while the 12-point Scenario B pays back in about 3.5 years with a five-year ROI of +44.1%. The reason is that investment differs by about 2.0 times while net benefit differs by only about 1.5 times. In practice it is rational to first identify where humidity or temperature is actually causing defects or stoppages, and to begin only around those locations. Practitioner guidance on mapping cites grids at 5 to 10 m intervals and vertical placement at three heights in large warehouses, but that is temporary measurement for verification purposes and should be kept separate from the number of permanent monitoring points.

How much does a temperature and humidity monitoring system cost?

For a 40-point model, the scale is 1,300,000 baht initial, 189,200 baht a year, and 2,246,000 baht over five years. Narrowed to 12 points it is 664,000 baht initial, 96,600 baht a year, and 1,147,000 baht over five years. What matters is the breakdown. The sensors themselves are 180,000 baht for 40 points, only 13.8% of the initial cost. Meanwhile calibration and verification at 300,000 baht plus corrective workflow design and training at 240,000 baht come to 540,000 baht, which is 41.5% of the initial cost and is what turns records into evidence. A quotation with that 41.5% removed looks cheaper, but it has removed the benefit along with it.

Is calibration required every year?

Not necessarily for every monitoring point. The common design is to prioritise points tied to product quality and process conditions and to exclude reference-only points such as outdoor measurement. The model calculation puts 20 of the 40 points into annual calibration scope, budgeting 50,000 baht a year, being 20 points × 2,500 baht. Practitioner guidance holds that data loggers should carry traceable calibration certificates, that calibration should generally be within the last 12 months, and that error should be within ±0.5°C at each calibration point and within ±4%RH for humidity. Perform calibration as ISO/IEC 17025 accredited calibration, and set the number of points and the frequency to match audit expectations.

Conclusion — identify what is breaking before you start measuring

A temperature and humidity monitoring system can be taken as far as sensors on the wall and a chart on screen in a short time. Stop there, though, and records accumulate while defects and complaints do not fall. The reason is that only Layer 1 of three, measure, has been built. In cost terms, sensors and measuring ends are 13.8% of the initial cost, and adding communication and power work brings it to 38.5%. Buying only the measuring layer is not an investment. It is buying a precondition.

Cost and benefit both concentrate in Layers 2 and 3. Calibration and verification plus corrective workflow design and training make up 41.5%, and that is the cost of turning records into evidence. And a drop in benefit realisation to 70% alone takes the five-year ROI from +12.0% to −43.4%. Since realisation is nothing other than the quality of Layer 3, correct and record, the corrective-action workflow owner should be decided before the sensor quotation.

One more thing. The more points you add, the further payback recedes. The 40-point Scenario A pays back in about 4.5 years, the 12-point Scenario B in about 3.5 years. Benefit exists only near the thing that is breaking.

So the right question is not whether to install a temperature and humidity monitoring system. It is how much money humidity and temperature are costing you per year. Two weeks of cross-referencing existing wall-mounted thermo-hygrometer records against defect records gives you an estimate. If that figure is below 348,000 baht a year, the 40-point Scenario A does not merit consideration. Start with 12 points, measure realisation, then expand. That sequence is the only one that makes this investment recoverable.

TOMAS TECH designs and delivers everything from factory automation and control through IoT to production management systems for Japanese manufacturers in Thailand and across ASEAN. On environmental monitoring, rather than opening with a quotation for point counts and hardware, we start by working through existing records with you to find out how much humidity and temperature are costing per year. There is no need to have reached a conclusion. If you are at the stage of asking how many points you have in each category, or who could own Layer 3, please get in touch through our contact page. We can also propose an approach that includes on-site measurement and mapping.

References

• Nation Thailand, “Monsoon brings heavier rain and flash-flood risk to parts of Thailand” https://www.nationthailand.com/news/general/40069528
• Nation Thailand, “Thailand’s 2026 summer forecast to reach temperatures over 42°C” https://www.nationthailand.com/blogs/news/general/40061952
• Food Division, Thai Food and Drug Administration, “GMP 420” https://food.fda.moph.go.th/gmp-head/420/
• PCBSync, “J-STD-033 Guide: MSD Handling, Floor Life, Baking and Dry Storage Requirements” https://pcbsync.com/j-std-033/
• Safety and Health at Work, “Climate Warming and Occupational Heat and Hot Environment Standards in Thailand” https://www.sciencedirect.com/science/article/pii/S2093791120303383
• Labour Protection and Welfare Office, “Ministerial Regulation on Heat, Light and Noise, B.E. 2559 (2016)” https://ubonratchathani.labour.go.th/en/2018-08-09-04-58-32/96-2559
• Thai PBS, “ERC sets the May to August electricity tariff at 3.95 baht per unit” https://www.thaipbs.or.th/news/content/504125
• DicksonData, “8 Steps to Compliant Temperature and Humidity Mapping for Pharmaceutical Warehouses” https://dicksondata.com/en-my/8-steps-to-compliant-temperature-humidity-mapping-for-pharmaceutical-warehouses