Routine water testing for Thai factories: what changed in 2025

If your water spec sheet cites Notification No. 61, it is out of date. Ministry of Public Health Notification No. 462 B.E. 2568 replaced it in December 2025, and No. 61 now carries the status "Repeal" on the Thai FDA's own page. Most Thai lab-supply sites still quote the old limits, which is worth knowing before you copy a table off one of them into a work instruction.

This is about the water tests a factory runs on its own bench — which ones earn their keep daily, which ones are a waste of a technician's morning, and which ones you should stop trying to do in-house at all.

What actually changed

The current standard for drinking water in sealed containers is Notification of the Ministry of Public Health (No. 462) B.E. 2568 (opens in a new tab), given on 28 November 2025 and in force from 12 December 2025. It repealed No. 61 B.E. 2524 along with four later amendments. The FDA's drinking-water index page (opens in a new tab) now lists it as the only current notification, and the page for No. 61 (opens in a new tab) shows the repeal.

One correction worth making early, because it circulates: No. 462 did not repeal Notification No. 220. That one was revoked by GMP 420 back in 2021. If a supplier's documentation cites 220 as live, they have two separate updates to catch up on.

Licensees producing carbonated water under the old beverage notification have two years from the in-force date to comply. Everyone else complies now.

The numbers you will actually use

From clause 6, for water in sealed containers. These are the ones a plant checks against, in the order a technician meets them:

• pH must not exceed 8.5 — but water used in food production, as an ingredient or in direct contact with food, must sit between 6.5 and 8.5. The two-sided limit is the stricter one and the one most factory water falls under.
• Turbidity: not more than 4 NTU. Colour: not more than 15 Hazen.
• Total dissolved solids: not more than 500 mg/L. Total hardness as CaCO₃: not more than 100 mg/L.
• Coliforms and E. coli: less than 1.1 per 100 mL by MPN, both.
• Water may be odourless only — except food-production water, which may carry a chlorine odour provided free chlorine stays at or below 5 mg/L.

The contaminant schedule behind those runs to 44 entries. The ones that come up most in Thai groundwater: arsenic 0.01 mg/L, lead 0.01 mg/L, manganese 0.08 mg/L, nitrate as NO₃ 50 mg/L, nitrite as NO₂ 3 mg/L. Where the notification is silent it defers to the current WHO drinking-water guidelines and to Codex.

A trap in the annex worth naming, because getting it backwards costs money in both directions. Footnote 1 says surveillance "may" be conducted by source risk and that analysis "may be limited to" a listed set. That is permission to narrow the scope, not an instruction about what you must test. Reading it as a mandatory list means paying for analyses nobody asked for; reading it as blanket permission to skip means missing the one limb that does say "must". Read the footnote itself before you design a sampling plan around it.

The split: bench today, lab once a year

The deciding factor is what the result is for. A number that changes what the line does in the next hour belongs on your bench. A number that goes to a regulator, a customer auditor or a certificate belongs at a government or accredited laboratory, because what you are buying there is the accreditation and the signature, not the measurement.

Daily, on the bench

pH, free chlorine, turbidity, TDS or conductivity, and total hardness. All five are strip- or meter-work, all five move day to day, and all five tell an operator something actionable before the batch moves. A softener that has stopped regenerating shows up in hardness within a shift; waiting for a quarterly external report to find it means a quarter of scaled equipment.

Free chlorine is the one people most often under-instrument. If you chlorinate, you are managing a number with a legal ceiling of 5 mg/L for food-production water and an effectiveness floor set by your own process — a test kit that reads in the right range, used at the right point in the loop, is not optional equipment.

Annually, or to an outside lab

Heavy metals, the full contaminant schedule, and microbiological confirmation. GMP 420 (opens in a new tab) clause 3.3.3 requires that water and ice which are an ingredient or contact ready-to-eat food be tested against the drinking-water notification, with results analysed at least once a year by a government laboratory or an accredited laboratory, and the records kept. Clause 3.9.1 says the same for finished product. Its Specific Requirements add annual physical and chemical testing of raw water, which exists so you can design a treatment system against real numbers rather than assumptions.

I would not try to bring coliform confirmation in-house for a plant that does not already run a micro bench. The plates are cheap; the incubator, the aseptic technique, and an analyst who can defend a result to an auditor are not, and a contract lab has already paid for all three.

Two things GMP 420 does not say

Both get sold to factories, and neither is in the text.

1. It does not require you to own a laboratory. Clauses 3.3.3 and 3.9.1 are explicitly satisfied by results from a government or accredited laboratory. The notification contemplates outsourced testing on its face; there is no clause establishing an in-house lab, in-house methods, or analyst qualifications.
2. It does not require an ISO/IEC 17025-accredited calibration provider. Clause 2.8 sets a frequency — appropriate frequency, and at least once a year — and a duty to act when a result falls outside the acceptable range. That is all. Accreditation language appears in the notification only where it talks about testing laboratories, never in the calibration clause. Traceable calibration is good practice and I would still buy it; it is not what 2.8 says.

What GMP 420 does not do either is tell you how often to test. Annual is a floor, written as "appropriate frequency and at least once a year". The real frequency for your product comes from its own commodity notification and from your HACCP plan. A general-GMP answer to a commodity-specific question is exactly the kind of thing an inspector catches.

Where HACCP fits, and where it does not

HACCP appears nowhere in Notification 420 — not the word, not "hazard analysis", not ISO 22000. I checked the full text. In Thailand HACCP is a customer and export requirement and a voluntary system built on Codex CXC 1-1969 (opens in a new tab), not a general legal one. Anyone telling a factory that Thai law obliges it to run HACCP is selling something.

Where it does bear on water: if a water parameter is a critical limit in your plan, Codex expects the monitoring device to be checked for accuracy and calibrated at regular intervals or as needed. Codex sets no number. Your interval is yours to justify, and the justification is the deliverable an auditor wants — not a certificate with a convenient date on it.

What to keep, and for how long

GMP 420 clause 3.12 requires records and reports kept at least one year past the date shown on the product label. Codex is looser — it says records should be retained for a period exceeding the product's shelf life, or as the competent authority determines, which in Thailand means clause 3.12. There is no fixed number of years in Codex, whatever a template tells you.

Practically: your daily bench log and your annual external certificates are the same record set to an inspector. The FDA's own audit checklist turns "no quality analysis result at least once a year" into a defect against 3.3.3 directly. An uncalibrated meter producing tidy daily numbers is worse than no log at all, because it looks like control.

What we checked, and what we did not

The limits above were read from the Thai FDA's own English text of Notification 462 on 1 August 2026, and cross-checked against the FDA index page. That English file is headed "Unofficial Translation" and the Thai version prevails in any discrepancy — if a number is going into a specification or a contract, quote the Thai. It also contains at least one date typo of its own.

The in-force date of 12 December 2025 comes from the FDA index page; we did not verify it against the Royal Gazette. And we read the drinking-water notification, not your commodity notification. If you produce something with its own quality standard, that document sets your parameters and we have not read it for you.