Checkweigher Selection: Cutting Giveaway with the Right RFP and Acceptance Tests in Thai Factories

“After packing, QC pulls samples every 30 minutes and weighs them. We are worried about underweight complaints, so we set the fill on the generous side.” We often hear this from production engineering managers and quality assurance managers at Japanese food, beverage, household-goods and pharmaceutical plants in Thailand. That is usually the point at which checkweigher selection comes onto the agenda.

Here is the conclusion up front. What determines the payback of a checkweigher is not “how many defective packs it can reject.” It is “how far you can safely bring the average fill toward the declared quantity,” in other words, how much giveaway you can cut. That is why, before comparing machine speeds, you measure the average and the variation of your current fill. Selection comes down to five decisions: (1) measure your current fill, (2) decide the accuracy and speed you need, (3) design the outlet, (4) design for legal metrology and record-keeping, and (5) confirm the numbers in acceptance testing.

Note that all production volumes, fill weights, costs, investment amounts, reject rates and payback periods in this article are this article’s own estimates and assumed values (placeholder values) based on the model factory described later. They are neither industry averages nor survey data. No primary vendor pricing information has been used either. Please read them as a “calculation template” to be replaced with your own measured values and quotations.

Why checkweigher selection matters now for factories in Thailand

The limits of spot checks and “generous filling”

In plants without an in-line checkweigher, net-content control relies mainly on sampling inspection. For example, QC pulls 10 bags every 30 minutes, weighs them, and adjusts the filler by hand if the average has dropped.

This approach has two weaknesses. The first is that almost all bags go unweighed. A line running at 40 bags per minute for 16 hours a day produces 38,400 bags a day. With 10 bags every 30 minutes, you weigh 32 times a day, or 320 bags. That is 1 bag in 120; nobody weighs the rest. If the fill drifts between two sampling rounds, you will not notice until the next round.

The second is that this “blind time” is covered by adding extra product to every fill. Because you fear underweight bags reaching the market, you set the average well above the declared quantity. It is a cautious decision, but the extra product goes to the customer free of charge in every single bag. This extra is called “giveaway,” and is also commonly referred to as overfilling.

Headwinds for food exports make “1 g of product” count

According to a September 17, 2026 article in the Thai business newspaper Thansettakij, Thailand’s food exports for January–June 2026 were THB 693,376.26 million, down 7.8% from THB 752,034.07 million in the same period a year earlier. The article cites a strong baht (31.62 baht/USD versus 33.44 baht/USD a year earlier), geopolitical conflicts affecting shipping routes, and rising ocean freight as headwinds. The person quoted is Wisit Limluecha, Vice Chairman of the Thai Chamber of Commerce. For the full year, growth of 3–5% is said to be possible if markets improve in the fourth quarter, but this is a projection that assumes such an improvement. This is a news report, and this article has not been able to check the materials of the agency that originally compiled the statistics.

When sales are hard to grow, the parts of cost that you can control yourselves become important. Giveaway is a prime example. We will run the numbers in detail later, but on a line making 11,520,000 bags a year, cutting the extra fill by 1 g per bag reduces product use by 11,520 kg a year. At an assumed cost of THB 60/kg, that is THB 691,200 a year. Extra fill that you thought of as “only about 1 g” adds up to hundreds of thousands of baht a year on a single line.

New checkweighers keep coming in 2026

There is movement on the equipment side too. Here are a few examples (all are vendor announcements, not recommendations or comparisons).

• METTLER TOLEDO: According to a September 22, 2026 preview article in the trade magazine Processing Magazine, the company’s Product Inspection Group will exhibit at PACK EXPO International 2026, held in Chicago on October 18–21, 2026, and will launch a program called “LightningBuild” with the C31 checkweigher and the X12 X-ray inspection system. It says selected systems will be delivered within four weeks from purchase order to the customer’s dock. This is a conditional vendor-stated figure limited to “selected systems,” and it concerns a trade show in the United States. The article does not mention delivery times to Thailand.
• Minebea Intec: On August 19, 2026, the company announced the “Essentus Advanced L” checkweigher. Its predecessor, Econus, is said to have more than 2,000 units installed worldwide, and the new model has a weighing range of up to 12 kg. It is said to offer Ethernet, Profibus, Profinet and serial communication, as well as a function for viewing production data in a browser. The company says it will first sell the model in Asia, India, the Middle East and Africa, but it does not state whether “Asia” includes Thailand. The company also exhibited Essentus Advanced at Anuga FoodTech India 2026, held in Mumbai from September 29 to October 1, 2026.
• Fortress Technology: At interpack 2026, the company launched the “Raptor Flex” checkweigher. It is said to handle packs of up to 4 kg and to weigh at up to 200 ppm (200 packs per minute). This is also a vendor-stated figure and varies with pack conditions. The announcement does not give an accuracy figure.

In addition, a trade magazine’s product news listing includes announcements of new checkweighers from manufacturers in the United Kingdom and China, so new products keep appearing. More choice is a good thing, but lining up catalog speeds and new features does not tell you whether the investment will pay back on your own line. This article works through, step by step, what you should decide first.

What a checkweigher (automatic weight checker) does

The diagram above shows the basic configuration of a checkweigher. Products enter from the left; the infeed sets their orientation and spacing, the weigh table measures their weight, and on the outfeed a rejector removes out-of-spec items. The measured weight data can also be sent back to the upstream filler to correct the fill amount automatically (feedback control).

Basic configuration

A checkweigher weighs products one by one as they move along the line, without stopping them, and automatically rejects anything outside the set weight range. It is also called a checkweighing machine or an automatic weight checker. Recommendation R 51-1 of the International Organization of Legal Metrology (OIML) defines a checkweigher as a device that subdivides prepackages of different mass into two or more subgroups according to the difference between their mass and the nominal set point.

According to Thermo Fisher Scientific’s practical guide, a checkweigher consists of a controller and a weigh frame, and the weigh frame is usually divided into the following three sections.

• Infeed: stabilizes the product and feeds it into the weighing section
• Weigh table: measures the weight while the product is on it
• Outfeed: sends weighed products onward and, if necessary, rejects out-of-spec items

The company says the infeed and weigh table conveyors should run at the same speed and that products should be properly spaced. With adequate spacing, the zero point can be re-taken at moments when no product is on the scale, which the company says gives the highest accuracy.

On conveying methods, the company explains as a general rule that belts are the most accurate, chains the most robust, and skates (sliders) give the best performance with cans. There are also auxiliary mechanisms, such as side belts that keep tall products from tipping over, worm screws for spacing products, and accelerating conveyors.

It does more than “reject underweight packs”

Thermo Fisher lists uses for checkweighers such as the following.

• Detecting underfilled and overfilled packs
• Combining with metal detectors and X-ray inspection systems
• Grading by weight zone
• Checking counts by weight (for example, a box of chocolates missing one piece)
• Detecting damage, leaks and wrong fill product
• Detecting missing components such as leaflets, spoons and straws
• Comparing lot averages and standard deviations with targets before shipment
• Sending weight data back to the filler to control the fill amount in a closed loop

This article puts particular weight on the last two. With weight data for every pack, the average and variation of each lot become visible as numbers. If you send that data back to the filler, you can keep the average close to the declared quantity.

Relationship with metal detectors and X-ray inspection

A checkweigher looks at weight, not at foreign bodies. Detecting foreign bodies is the job of metal detectors and X-ray inspection systems. Thermo Fisher also lists combination with metal detectors and X-ray systems as one use of checkweighers, and there are combination units that mount a checkweigher and a metal detector on a single frame. How to decide which to install, in what order and how many, is covered in Metal Detector Selection: X-ray Inspection and Thai Plant RFP.

Payback depends on “how far you can move the average,” not “how many packs you reject”

The chart above compares, for the “current state,” “Option A” and “Option B” of Model Factory Q (explained below), the average fill per bag and the amount of giveaway above the declared quantity of 200 g. The current state averages 206.0 g with 6.0 g of giveaway, Option A averages 203.0 g with 3.0 g, and Option B averages 201.5 g with 1.5 g.

From here, we estimate costs and benefits for a fictional model factory. To repeat, all of the following figures are this article’s own estimates and assumed values (placeholder values). They are neither industry averages nor survey data, and no primary vendor pricing information has been used.

Common assumptions (Model Factory Q)

• A Japanese seasoning and snack plant in central Thailand. The target is a line for a single SKU of 200 g declared-quantity pouches
• Line capacity of 40 bags per minute, running 16 hours a day (two shifts), 300 days a year
• Output: 40 bags × 60 minutes × 16 hours = 38,400 bags a day, × 300 days = 11,520,000 bags a year
• Product cost of THB 60/kg (= THB 0.06/g, assumed)
• Currently there is no in-line checkweigher. QC pulls and weighs 10 bags every 30 minutes and adjusts the filler by hand. For fear of underfill, the average fill is set at 206.0 g (giveaway of 6.0 g per bag, assumed)
• Handling cost for each rejected bag (labor to open, reweigh and repack, plus packaging material) is THB 2.0 (assumed)
• Labor savings from sampling inspection are not counted as a benefit (to stay conservative)

First, let us put a money value on the current giveaway. 11,520,000 bags × 6.0 g = 69,120 kg; multiplied by THB 60/kg, that is THB 4,147,200 a year. This is the product that leaves this line every year as a “free bonus.”

Option A: one checkweigher (reject only, no feedback)

• Initial cost of THB 1,200,000 (machine, rejector, installation and training, assumed); annual maintenance and calibration cost of THB 60,000 (assumed)
• Because every bag can be weighed, the average fill is lowered to 203.0 g (giveaway from 6.0 g to 3.0 g, a reduction of 3.0 g per bag, assumed)
• Product benefit: 11,520,000 bags × 3.0 g = 34,560,000 g = 34,560 kg. × THB 60/kg = THB 2,073,600 a year
• Reject handling cost: at a reject rate of 0.8% (assumed), 11,520,000 bags × 0.8% = 92,160 bags. × THB 2.0 = THB 184,320 a year
• Annual net: 2,073,600 − 184,320 − 60,000 = THB 1,829,280
• Payback: 1,200,000 ÷ 1,829,280 = about 0.7 years

Option B: checkweigher plus feedback control to the filler

• Initial cost of THB 1,800,000 (Option A plus communication and control modifications to the filler, assumed); annual maintenance and calibration cost of THB 80,000 (assumed)
• Because fill drift can be corrected automatically, the average is lowered to 201.5 g (giveaway from 6.0 g to 1.5 g, a reduction of 4.5 g per bag, assumed)
• Product benefit: 11,520,000 bags × 4.5 g = 51,840,000 g = 51,840 kg. × THB 60/kg = THB 3,110,400 a year
• Reject handling cost: at a reject rate of 0.5% (assumed), 57,600 bags. × THB 2.0 = THB 115,200 a year
• Annual net: 3,110,400 − 115,200 − 80,000 = THB 2,915,200
• Payback: 1,800,000 ÷ 2,915,200 = about 0.6 years

The reject rate for Option B is set lower than for Option A on the assumption that, because fill drift is corrected quickly, fewer bags fall below the lower limit.

Sensitivity: what if measurement shows the current average is already 203.0 g?

The most uncertain assumption is “what the current average fill actually is.” Even if the sampling records suggest it “should be around 206 g,” continuous weighing often shows something different. So we calculate the case where measurement shows the current average is already 203.0 g (giveaway of 3.0 g).

• Option A: Even with every bag weighed, without feedback the average stays at 203.0 g and there is no room to lower it. The product benefit is 0. Annual net is 0 − 184,320 − 60,000 = THB −244,320, so the investment cannot be recovered from product savings alone.
• Option B: The average can be lowered from 203.0 g to 201.5 g, or 1.5 g per bag. 11,520,000 bags × 1.5 g = 17,280 kg, × THB 60/kg = THB 1,036,800. Subtracting 115,200 and 80,000 gives an annual net of THB 841,600. Payback is 1,800,000 ÷ 841,600 = about 2.1 years.

However, the fact that Option A cannot pay back on product savings alone does not mean it has no value. Because every bag is weighed, it prevents underweight bags from reaching the market, and it leaves weight records for each lot. These are separate forms of value that reduce the risk of complaints and legal violations. This estimate does not put a money value on them.

Summary of the estimate

Takeaway 1: almost all the benefit is product, and rejects are actually a cost

In both Option A and Option B, nearly all of the benefit comes from product (giveaway reduction). Reject handling sits on the cost side, not the benefit side, because rejected bags have to be opened, reweighed and repacked.

In other words, if you see a checkweigher as “a machine that rejects defective packs,” the more it rejects, the more it costs you. The right way to see it is as “a machine that weighs every pack so you can move the average toward the declared quantity.” The reject function is a safety device that keeps bags below the lower limit off the market when you move the average down.

Takeaway 2: the difference comes from whether you have feedback

If the current average is 206.0 g as assumed, Option A’s payback of about 0.7 years and Option B’s of about 0.6 years are almost the same. Either one pays back within a year on these numbers.

But if the current average is already 203.0 g, Option A cannot pay back on product savings alone, while Option B pays back in about 2.1 years. The smaller a plant’s current giveaway, the more the outcome depends on whether it has feedback control.

Incidentally, Option B costs THB 600,000 more up front than Option A, and the difference in annual net is THB 1,085,920 both under the base assumptions and in the sensitivity case. Looking only at the incremental portion, that extra is recovered in about 0.6 years in either case.

Takeaway 3: the first step is measuring your current fill, not comparing models

The benefit in this estimate is set by the gap between “what the current average is” and “how far it can be lowered.” Depending on whether the current average is 206.0 g or 203.0 g, Option A’s annual net swings from THB 1,829,280 to THB −244,320.

That is why, before comparing models, you measure the average and variation of your current fill with continuous samples. The first task in selection is to get a numerical grip on how many grams of extra fill you are using to protect your net content. How to measure is explained in the next section.

Takeaway 4: how far you can lower the average depends on “accuracy”

How close you can bring the average to the declared quantity depends on the variation of the filler itself and on the accuracy of the checkweigher. If you lower the average while variation remains large, more bags fall below the lower limit and reject costs swell.

This is where the trade-off between speed and accuracy matters. As explained later, even on the same machine, running faster tends to shorten weighing time per item and reduce accuracy. If you choose a model on catalog speed alone, accuracy may fall short and you will not be able to lower the average. In that case, the benefit in this estimate does not materialize at all.

Takeaway 5: lowering the average assumes you still meet legal requirements

Lowering the average is only possible within the range that meets legal requirements. OIML Recommendation R 87 (2016 edition), an international reference, requires for prepackages that “the average actual quantity is at least equal to the declared quantity.” In addition, for a declared quantity of 200 g, the tolerable deficiency T is 9 g (in the R 87 table, the 200–300 g band is 9 g and the 100–200 g band is 4.5%; 4.5% of 200 g is also 9 g). The tolerances in Thailand need to be checked in the Ministry of Commerce notification described later.

Takeaway 6: the numbers are placeholders, so recalculate with your own

All the numbers so far are placeholder values. Please replace cost, output, current average, achievable reduction, reject rate, handling cost and investment with your own measurements and quotations and recalculate. In particular, “achievable reduction” is a value that should be decided only after confirming actual accuracy in acceptance testing.

Five decisions in checkweigher selection

(1) Measure your current fill: average, variation and change over time

The first step is to turn the reality of your current fill into numbers. Even if you have sampling records, 10 bags every 30 minutes cannot show you what happens in between.

The basics of measurement are as follows.

• Take continuous samples: for a set period, weigh the bags coming off the filler one after another without skipping any. Do this several times, across different times of day (just after start-up, midday, around shift changes, at night)
• Calculate the average and variation: calculate not only the average but also the standard deviation and minimum. How far the minimum sits above the declared quantity is your current “safety margin”
• Look at change over time: see how the average moves when the raw-material lot changes, when the hopper level changes, just after a shift change, and so on
• Calculate the giveaway: the average minus the declared quantity is your current giveaway. Multiply this by annual bag count and product cost to get the money value of your current giveaway

If the “change in the average over time” found here is large, that indicates feedback control will have a large effect. Conversely, if the average is stable but variation is large, adjusting or servicing the filler itself may need to come first.

If there is room to review the capacity or type of your upstream filler or bagging machine, please also see Bagging Automation: URS, FAT/SAT and Business-Case Guide.

(2) Accuracy and speed: load cell vs EMFR, weigh conveyor length, single lane vs multi-lane

The faster you run, the more accuracy tends to drop

A technical document from Anritsu (Technical Note Vol.6, September 2012) presents a comparison in which only the length of the weigh conveyor was changed, with the same weighing unit and the same speed. Weighing accuracy was 2.3 g with a 270 mm conveyor and 1.0 g with a 495 mm conveyor, so the longer one was 1.3 g better. The document explains that this is because the product is weighed for longer on the longer conveyor, making the measurement more accurate. On the other hand, it also states that short weigh conveyors are suited to high-volume production lines.

The important point here is that these values come from a vendor’s in-house experiment in 2012, and the document states neither what “weighing accuracy” means statistically (standard deviation, or three times that) nor the product weight and speed conditions. They therefore cannot be compared on the same scale as the OIML “standard deviation” values introduced later. What you can take from them is only the tendency that “the longer the weighing time, the higher the accuracy, and this trades off against speed.”

Load cell vs electromagnetic force restoration (EMFR)

Weighing cells broadly fall into two types: strain-gauge load cells and electromagnetic force restoration (EMFR).

The same Anritsu document states that, with the same weigh conveyor length, weighing accuracy was ±2.3 g for the strain-gauge type and ±0.8 g for the electromagnetic type. It then says the high-accuracy electromagnetic type suits products such as individually wrapped candy and chocolate before packing, while the less accurate but lower-cost strain-gauge type is recommended for checking missing items in bags after cartoning. The idea is that you should choose the type based on the product and where on the line the machine sits. These too are values from the same 2012 in-house experiment, with unknown statistical definition and conditions. Please do not generalize the ratio and read it as “the electromagnetic type is about 3 times more accurate.”

WIPOTEC, a manufacturer specializing in electromagnetic force restoration, explains the EMFR principle as follows. When a load is applied to the balance mechanism, a position sensor detects the displacement, and a controller adjusts the coil current to maintain equilibrium. Because that current is proportional to the load, it is converted into a digital signal and used as the weight. The company claims this gives very accurate weights faster than strain-gauge cells and that, with no springs or mechanical wear parts, it is well suited to dynamic weighing. It also describes a vibration compensation function that removes external vibration, such as from passing forklifts, from weighing results. Please note that these are all vendor claims and do not mention the advantages of load cells, such as cost and robustness.

How “accuracy” affects product use on a 200 g bag

The Anritsu document also includes a simplified example showing why accuracy affects product use. For a 200 g coffee pack, a machine with ±1 g accuracy needs a minimum fill of 201 g, while a machine with ±0.5 g accuracy needs 200.5 g, so the former uses 0.5 g more product per pack. This is an example to illustrate the concept and is separate from the legal average requirement. Even so, it captures the relationship that “the better the machine’s accuracy, the closer you can bring the average to the declared quantity.”

Single lane vs multi-lane

Thermo Fisher gives the following reasons for using multi-lane checkweighers: a single lane may not keep up with the filler’s capacity; multiple lanes coming off a multi-lane filler can be weighed as they are; speed per lane drops, so accuracy improves; and the straight-line space required can be greatly reduced.

Checkweigher comparison table by weighing technology

Do not compare on speed alone

Vendor catalogs are full of big throughput numbers. For example, Thermo Fisher’s guide lists values such as up to 700 packages per minute for a chain-type model, up to 550 ppm for 2 g–500 g on a pharmaceutical model (depending on pack size), and up to 350 packages per minute for packages of up to 3 kg on a general-purpose model. However, it does not say which accuracy is achieved at which speed.

Model Factory Q’s line runs at 40 bags per minute. What you need to know is what accuracy you get on your own product at your own speed. In the RFP, ask not for “maximum speed” but for “the accuracy when running this product at this speed.”

(3) Design the outlet: rejector type, reject confirmation and feedback control

Rejector types

Thermo Fisher lists air blast, pusher, bopper, drop flap and diverter as rejector types, and notes that other types exist depending on the product and customer requirements. It also lists the factors that affect accuracy as product speed, product weight and size, transfer onto and off the checkweigher, and the reject method and design. Which type fits depends on the product and speed, so decide by testing with actual product.

Reject reliability: underfill prevention so no underweight bag gets through

The rejector fails to fire, yet the machine counts the bag as “rejected.” This is the most dangerous failure. A useful reference is the set of basic requirements in a Thermo Fisher document for systems that meet the foreign-body detection code of practice of the major UK retailer Marks and Spencer (M&S).

• All sensors fail to the closed state and raise an alarm (fail-safe)
• A photo-eye at the infeed registers each product
• A lockable reject bin, and full enclosure from the inspection point to the bin
• A reject confirmation sensor
• Bin-full notification, and an alarm if the bin stays open or unlocked
• A low air-pressure switch
• Faults and alarms remain even after power cycling and can be cleared only by a key holder such as a QA manager

These are requirements for foreign-body detection systems such as metal detectors and X-ray inspection systems, not requirements for a standalone checkweigher. Still, the ideas of “confirming that a reject actually happened,” “not letting rejected bags be put back on the line at will” and “stopping when air pressure drops” apply directly to checkweighers. We recommend including reject confirmation, a locked bin, bin-full detection and low air-pressure detection in the RFP. Ishida Europe’s DACS-GN series, announced in 2024, is also said to incorporate reject confirmation technology for retail and technology to counter floor vibration, and is an example of functions based on the same idea.

Filler feedback control for giveaway reduction: keeping the average on target

An article contributed to a trade magazine by a Mettler-Toledo Garvens representative (June 2025) describes feedback control as a communication loop that connects the checkweigher with the upstream filler. It says that as soon as a deviation from the target weight is detected, data is sent upstream and the filler adjusts its output automatically, so product weight can be kept constant without manual work or production stops. No figures such as reduction rates are given.

The assumption that Option B can bring the average 1.5 g lower than Option A at Model Factory Q is based on this mechanism. To introduce feedback, the filler must be able to change its fill amount in response to an external signal, and you need to decide the communication method and the content of the signal (how the correction is calculated, the size of each correction, the upper and lower limits of correction, and so on). Also decide at the RFP stage which party, the filler manufacturer or the checkweigher manufacturer, is responsible for what. The basics of connecting scales and control panels (PLCs) are covered in Industrial Scale PLC Integration: RFP and FAT/SAT for Thai Factories.

(4) Design for legal metrology and records: OIML R 51 / R 87, the EU example and Thailand’s Weights and Measures Act

Two international recommendations: machine accuracy and net content of prepackages

There are two main international references on net content. One is OIML R 51-1, which sets the accuracy of the equipment (automatic catchweighing instruments); the other is OIML R 87, which sets the net content of prepackages.

OIML R 51-1 (2006 edition) divides automatic catchweighing instruments into two categories by use. Category X applies only to checkweighers that check prepackages subject to the requirements of R 87, and Category Y applies to everything else (such as weight-price labelers and postal and shipping scales). A checkweigher that checks the net content of prepackages in a factory falls under Category X in the OIML classification.

Maximum permissible errors for Category X have two parts: “mean error” and “standard deviation.” There are upper limits both on the shift of the mean over consecutive weighings (systematic error) and on the variation (random error). For example, with a factor (x) of 1, for loads over 100 g and up to 200 g, the maximum permissible standard deviation is 0.24% at initial verification and 0.3% at in-service inspection. In the table, bands expressed in % and bands expressed in g alternate, so please pay attention to the units when reading it. In Category X tests, for loads of 1 kg or less, 60 consecutive weighings are made to determine the mean and standard deviation.

Note that this article checked the 2006 edition, and we could not confirm from the OIML catalog whether it is the current latest edition.

OIML R 87 (2016 edition) sets three requirements that prepackages must meet at every stage of distribution.

1. Average requirement: the average actual quantity is at least equal to the declared quantity
2. Packages with a T1 error (shortfall greater than T but not more than 2T) make up no more than 2.5% of a homogeneous group
3. There are no packages at all with a T2 error (shortfall greater than 2T)

For a declared quantity of 200 g, T is 9 g, so applying R 87 means that bags below 191 g but at least 182 g (T1 error) must be no more than 2.5%, and bags below 182 g (T2 error) must be zero. The important point is that, because of the average requirement, it is not the case that “you can lower the average as much as you like as long as you reject the bags at the lower limit.” The lower the average, the closer you come to both the average requirement and the T1 and T2 limits.

The EU example

In the EU, Annex VIII (MI-006) of the Measuring Instruments Directive (MID, 2014/32/EU) covers automatic catchweighing instruments, and here too there are Category X and Y. Category X covers instruments that check prepackages made up in accordance with the requirements of the Prepackages Directive 76/211/EEC. This article checked the UK version of the text published on the UK government website.

That Prepackages Directive 76/211/EEC requires that the average actual content is not less than the nominal quantity, that the proportion of packages with a negative error greater than the tolerable negative error is sufficiently small, and that packages with a negative error greater than twice the tolerable negative error must not bear the “e” mark. The detailed rules for sampling tests are in Annex II, which this article has not checked. If you export to the EU, please check the legal text.

Thailand: the Weights and Measures Act and the 2018 notifications

In Thailand, the Weights and Measures Act (B.E. 2542, 1999) governs the net content of prepackages. The following is based not on the text of the Act itself but on a summary of the key points of the law prepared by the Ministry of Commerce’s Udon Thani Provincial Office (a summary document from a government agency).

• Section 62: The Minister of Commerce sets, by notification, the types of prepackaged goods for which packers (including importers and repackers as well as manufacturers) must declare the quantity, and the criteria and methods for that declaration
• Section 63: The declared quantity must match the actual quantity, and the error must not exceed the tolerance notified by the Minister of Commerce. The criteria and methods for checking the correctness of the declaration are notified by the Director-General of the Department of Internal Trade
• Section 64: Imported prepackages labeled in foreign units must also be labeled in Thai units

The notification under Section 62 is the Ministry of Commerce notification dated March 12, 2018 (which sets the types of prepackaged goods, the criteria and methods for declaring quantity, and the tolerances), and the inspection method under Section 63 is the Department of Internal Trade notification dated June 8, 2018. As penalties, a packer who violates the Section 62 notification faces imprisonment of up to 6 months, a fine of up to THB 20,000, or both (Section 83), and packing while knowing the declared quantity is incorrect in a way likely to cause damage to others is subject to imprisonment of up to 1 year, a fine of up to THB 40,000, or both (Section 84), among other provisions.

This article has not been able to check the tolerance table in the text of the notification. We also do not know whether Thailand’s tolerances are the same as OIML R 87. Before deciding how far to lower the average, please check the value for your product category in the text of the 2018 notification.

Designing your records

To show that you meet legal requirements, you need records. We recommend keeping, as data rather than on paper, the bag count, average, standard deviation, number of rejects and their reasons, and calibration records for each lot and each shift. Designs for linking checkweigher data to quality records are covered in Weighing Scale Data Integration: Stable-to-Quality Design, and systems for tracing lots from raw materials to shipment are explained in Food Factory Traceability System: RFP and 90-Day Deployment.

Note that this article does not take a definitive position on whether net-content control should be treated as a HACCP CCP (critical control point). Net content is mainly a matter of quality and legal metrology control, but whether to make it a CCP depends on each company’s hazard analysis. Minebea Intec positions the checkweigher as a key component for reliably preventing overfilling and underfilling and for meeting legal requirements.

(5) Confirm the numbers in acceptance testing

The last step is to confirm that the accuracy, outlet and records decided so far are actually achieved. This is done in two stages: FAT (factory acceptance test) at the vendor’s plant and SAT (site acceptance test) at your own plant. The details are explained later in the section “What to confirm in FAT/SAT.”

12 items to include in a checkweigher RFP

When asking vendors for quotations, include the following 12 items. Thermo Fisher also recommends giving the supplier all product and application information when requesting a quotation. If you obtain quotations without aligned assumptions, you can compare neither accuracy nor price.

Item 2 in particular is the core of this article. Do not ask for “maximum N ppm” and “accuracy ±N g” separately; ask for them as a combination for your product and your speed. Unless you also specify whether the accuracy figure is a standard deviation or some multiple of it, you cannot compare figures across vendors.

The division of responsibility in item 12 is also important. Feedback control spans both the checkweigher and the filler. Unless you decide which side, the one sending the signal or the one receiving it, is responsible for the result of the correction, you will not be able to isolate the cause if “the average won’t come down” after start-up.

What to confirm in FAT/SAT

The diagram above shows the flow of acceptance testing. It starts with warm-up and static calibration, then checks dynamic repeatability, eccentricity, article separation, reject confirmation and alarms, feedback control, and data output in that order.

A useful template: the ASEAN guideline

A useful reference when planning acceptance test procedures is the “Guideline on Automatic Catchweighing Instrument Version 2.0,” compiled by the legal metrology working group of the ASEAN Consultative Committee for Standards and Quality (ACCSQ). It is a common reference document, based on OIML R 51-1 (2006 edition), for aligning verification procedures among member states.

However, this guideline comes with important caveats.

• It covers instruments of class Y(a) and Y(b) used for trade within ASEAN, and Category X checkweighers that check prepackages are outside its scope
• It assumes the instrument already has type approval
• Every requirement can be adjusted to each country’s national regulations
• This article has not been able to confirm whether Thailand has adopted this guideline as its national verification procedure

In other words, this is not Thailand’s legal procedure but a “useful template” for designing a factory SAT. On that basis, here are the ideas you can use.

• Warm up for about 30 minutes before testing
• Use a control instrument for reference measurements that is at least 3 times more accurate than the instrument under test
• Prepare at least 5 test loads spaced roughly evenly between minimum and maximum
• For the eccentricity test, run a test load of 1/3 of maximum load 3 times each at 2 positions, toward the front and toward the back of the conveyor
• For the weighing performance test, run the same test load a total of 10 times
• For the article separation test, run 2 items with no gap 3 times, then swap their positions and run another 3 times
• In the visual inspection, check that dust, air currents, vibration and the like are not disturbing operation, and that the instrument is installed on a stable, level foundation

FAT/SAT test items

Combining this template with Thermo Fisher’s recommendations for routine management, acceptance test items can be organized as follows. The factory sets the acceptance criteria based on the accuracy and reject requirements decided in the RFP.

How to think about the number of test runs

Thermo Fisher recommends running a single package at least 20 times to obtain statistically valid results in “gray zone verification” (checking that products near the decision boundary are judged correctly), done at each product changeover or before each shift starts. This is the company’s recommendation and is separate from the 60 consecutive weighings that OIML R 51-1 Category X requires for loads of 1 kg or less. In acceptance testing, write down separately what each test is for, so that the legal number of test runs and your own daily-check runs are not confused.

Run the same procedure in FAT and SAT

FAT takes place at the vendor’s plant and SAT at your own, but it is important to run the same product with the same procedure. Even if accuracy is achieved in FAT, it can drop at your plant because of floor vibration, air-conditioning drafts or vibration from neighboring machines. If you run the same tests in both, you can tell whether a difference is “a machine problem” or “an installation environment problem.”

Points specific to factories in Thailand

1. Check the Weights and Measures Act and notifications yourself

As noted earlier, the tolerances for prepackages in Thailand are set by the Ministry of Commerce notification dated March 12, 2018, but this article has not been able to check those values. Please decide how far to lower the average only after checking the values in this notification.

Also, on verification of measuring instruments used in trade, Section 25 of the Weights and Measures Act states that measuring instruments used for buying and selling goods or calculating fees and charges must be inspected and verified by officials (according to the government agency’s summary document). Verification and related work is handled by the Central Bureau of Weights and Measures, located within the Ministry of Commerce’s Department of Internal Trade (Section 6). However, this article has not been able to confirm whether a checkweigher used only for sorting within the process is subject to verification. We cannot say definitively that it is either required or not required, so please check with the Central Bureau of Weights and Measures or a specialist.

2. Unit labeling on imported prepackages

Under Section 64, imported prepackages labeled in foreign units must also be labeled in Thai units (according to the government agency’s summary document). If you import and sell prepackaged products from Japan or other countries, please remember to check the labeling as well.

3. Floor vibration, forklifts and air-conditioning drafts

A checkweigher measures very small differences in weight. Floor vibration, forklifts running nearby, neighboring machines and drafts from air-conditioning outlets directly affect accuracy. Thermo Fisher also points to grounding the frame and leveling the conveyors, not letting other conveyors touch the checkweigher, and not starting and stopping the line frequently. When choosing the installation location, check the positions of air-conditioning outlets, forklift aisles and floor condition on drawings, and confirm accuracy in the actual environment during SAT.

4. Cleaning and materials

Food plants need protection ratings and materials suited to their cleaning methods. WIPOTEC explains that IP69 withstands high-pressure water jets, and IP69K additionally withstands cleaning with hot water and steam jets. Thermo Fisher cites 304 stainless steel, which resists corrosive detergents (316 for harsher environments), and IP65. Confirm how much cleaning the machine can withstand against the manufacturer’s conditions, and write your own cleaning procedure (water temperature, pressure, detergent) in the RFP to get a response.

5. Local support and calibration arrangements

For zero and span calibration, Thermo Fisher recommends doing it daily or at each product changeover. Who manages the test weights and test packs used for calibration, whether annual inspections can be done by engineers in Thailand, and whether spare parts are stocked in Thailand make a big difference to the impact of a breakdown. Whether the operating screens and manuals are available in Thai is also essential for getting the shop floor to follow procedures.

6. Labor savings are not the main driver

Thailand’s minimum wage, as of the revision that took effect on July 1, 2025, is THB 337–400 per day and varies by province (THB 400 applies in Bangkok, Chachoengsao, Chonburi, Phuket, Rayong and others). The wage ceiling used to calculate social security contributions was also raised to THB 17,500 from January 1, 2026. Labor costs are gradually rising.

Even so, the estimate in this article does not count labor savings from sampling inspection as a benefit. Even once every bag is weighed, QC sampling and calibration checks remain. What determines the payback of a checkweigher is, after all, the reduction in product (giveaway).

7. See machines in person at trade shows

In Thailand, the packaging and processing machinery trade show ProPak Asia 2027 is scheduled to be held at IMPACT, Muang Thong Thani, on June 9–12, 2027. If you bring samples of your own product and the average and variation data you measured in step (1), and ask vendors concretely about “the accuracy when running this product at this speed,” comparison becomes easier.

Common mistakes

Choosing a model on speed

You chose a model with a high maximum catalog speed, but accuracy on your product falls short and you cannot lower the average. If you choose without considering the speed-accuracy trade-off, the very premise of the investment collapses. In the RFP, ask for accuracy at your own speed.

Installing it but not lowering the average

Even though every bag can now be weighed, the average fill is left at the original 206 g “just in case.” In that case the product benefit is 0, and only reject handling and maintenance costs increase. In Model Factory Q’s numbers, you would be throwing away a product benefit of THB 2,073,600 a year. In the implementation plan, write down to how many grams, and by when, you will lower the average.

Leaving alarms ringing constantly

Thresholds are set too tight and the alarm rings all day. Thermo Fisher points out that alarms that are always going off end up being ignored by operators. Because the line will no longer stop when it really should, reset the thresholds to suit operations.

Not locking the reject bin

An operator decides a rejected bag is “only slightly light” and puts it back on the line. Unless you have a lockable bin and decide who has authority to open it, underweight bags will reach the market.

Skipping calibration at product changeover

The product is changed over, but the zero and span check and the gray-zone check are skipped. If you run with the previous product’s settings, judgments will be off. Include calibration and checks in the changeover procedure.

Ending with data on paper

Checkweigher summaries are copied onto paper and filed in binders. This hides changes in the average over time and trends by product, so you cannot judge whether the average can be lowered further. Keep it as data and link it to quality records and traceability.

Ordering without deciding who owns feedback

The checkweigher and the filler are ordered from different manufacturers, and nobody has decided who will build the communication and the correction logic. After start-up, it turns into a blame game: “we are sending the signal” versus “the signal content is wrong.” Decide the division of responsibility in item 12 of the RFP.

Frequently asked questions (FAQ)

What is a checkweigher, and what can it do?

It is a machine that weighs products one by one as they move along the line, without stopping them, and automatically rejects anything outside the set weight range. It is also called a checkweighing machine or an automatic weight checker. Besides detecting underfilled and overfilled packs, it can be used to check counts and components by weight, record lot averages and standard deviations, and run feedback control that sends weight data back to the filler to correct the fill amount automatically. In terms of payback, the biggest benefit is using data on every pack to bring the average fill closer to the declared quantity and reduce giveaway.

How much does a checkweigher cost?

This article does not use primary vendor pricing information, so it does not give market prices. In the model estimate, we set placeholder initial costs of THB 1,200,000 for one checkweigher (reject only) and THB 1,800,000 with feedback control, but these values are only there to show the calculation template. When comparing quotations, please request them on aligned terms: accuracy for the same product at the same speed, reject reliability functions, scope of feedback, data output, costs for installation, training and FAT/SAT, and annual maintenance and calibration costs.

Should we choose a combined checkweigher and metal detector?

A combination unit mounts a checkweigher and a metal detector on a single frame. However, weight inspection and foreign-body inspection have different purposes, and depending on the type of foreign body, an X-ray inspection system may be needed. Decide first which foreign bodies you will look for, where on the line, and with which equipment, and then judge whether to use a combination unit or separate machines. The approach is covered in the article on metal detector selection.

What is the difference between load cell and EMFR checkweighers?

A strain-gauge load cell weighs by converting the strain caused by the load into an electrical signal. Electromagnetic force restoration (EMFR) adjusts an electromagnetic force to balance the load and derives the weight from that current. Anritsu’s 2012 technical document shows in-house experimental values of ±2.3 g for the strain-gauge type and ±0.8 g for the electromagnetic type with the same weigh conveyor length, and recommends the electromagnetic type for applications needing high accuracy and the lower-cost strain-gauge type for tasks such as checking missing items. However, the document does not give the statistical definition of accuracy or the test conditions, so please compare on your own product at your own speed.

Does a checkweigher need verification in Thailand?

This article has not been able to confirm this. Thailand’s Weights and Measures Act requires inspection and verification for measuring instruments used in trade, such as buying and selling goods (according to the government agency’s summary document). However, we could not confirm whether a checkweigher used only for sorting within the process is covered. Because we cannot say definitively that it is either required or not required, please check with the Central Bureau of Weights and Measures or a specialist. Note that the quantity declaration and tolerances for prepackages are set by the Ministry of Commerce notification dated March 12, 2018.

How much giveaway can be reduced? Does filler feedback control make a difference?

It varies so much from plant to plant that we cannot give a general reduction figure. In this article’s model estimate (all placeholder values), on a line averaging 206.0 g, we assumed Option A (reject only) lowers the average to 203.0 g (3.0 g less per bag) and Option B (with feedback control) lowers it to 201.5 g (4.5 g less per bag). How far you can go depends on the variation of the filler, the accuracy of the checkweigher, and the tolerances in the Thai notification. Please start by measuring the average and variation of your current fill with continuous samples.

Can we lower the average right down to the declared quantity?

Because of the average requirement, you cannot go right down to the limit. In OIML R 87, an international reference, in addition to the average actual quantity having to be at least the declared quantity, there is also a cap on the proportion of bags with large shortfalls (for a declared quantity of 200 g, the tolerable deficiency T is 9 g). If you lower the average while variation remains large, you come closer to these requirements and rejects also increase. In Thailand, please check the tolerances in the 2018 notification and then decide a safe average based on the variation of your own fill.

Summary

• In checkweigher selection, what determines payback is not how many defective packs are rejected but how far you can safely bring the average fill toward the declared quantity (how much giveaway you can cut)
• The order of decisions is five steps: (1) measure your current fill, (2) decide accuracy and speed, (3) design the outlet, (4) design for legal metrology and records, and (5) confirm in acceptance testing
• In the model estimate (placeholder values), on a line averaging 206.0 g, Option A (reject only) pays back in about 0.7 years and Option B (with feedback) in about 0.6 years. If the current average is already 203.0 g, Option A cannot pay back on product savings alone, and Option B takes about 2.1 years
• Almost all of the benefit is product, and reject handling is actually a cost. What makes the difference is whether you have feedback control
• Accuracy tends to drop as speed rises, so in the RFP ask for “accuracy on your own product at your own speed.” Read vendor figures as conditional, vendor-stated values
• OIML R 51-1 and R 87 are international references. For Thailand’s tolerances, check the 2018 notification, and check with the Central Bureau of Weights and Measures or a specialist whether in-process checkweighers need verification
• For acceptance testing, use the ASEAN guideline as a useful template and confirm static calibration, dynamic repeatability, eccentricity, article separation, reject confirmation, feedback and data output

You are welcome to start at an early stage of your review, such as how to measure the average and variation of your current fill with continuous samples, or how to organize a checkweigher RFP and acceptance criteria. If you are considering net-content control or giveaway reduction at a plant in Thailand, we can help you sort things out from the very first stage. Please feel free to reach out via our contact page.

References

• Processing Magazine, PACK EXPO International 2026 Preview: METTLER TOLEDO Product Inspection Group (September 22, 2026): https://www.processingmagazine.com/home/article/55400237/pack-expo-international-2026-preview-mettler-toledo-product-inspection-group
• Minebea Intec, announcement of Essentus Advanced L (August 19, 2026): https://minebea-intec-com.mynewsdesk.com/pressreleases/focus-on-product-quality-the-new-checkweigher-essentus-r-advanced-l-3461266
• Minebea Intec, exhibiting at Anuga FoodTech India 2026 (August 13, 2026): https://minebea-intec-com.mynewsdesk.com/pressreleases/inspection-reimagined-minebea-intec-at-anuga-foodtech-2026-in-india-3460788
• Innovations Food, Fortress Technology Raptor Flex (April 28, 2026): https://innovationsfood.com/fast-and-flexible-new-food-checkweigher/
• Weighing Review, Product News listing: https://www.weighingreview.com/tag/product%20news
• Ishida Europe, DACS-GN checkweighers (April 22, 2024): https://www.ishidaeurope.com/en/about-us/case-studies-and-events/news/new-checkweigher-range-meets-european-market-needs
• Anritsu, Technical Note Vol.6 (September 2012): https://dl.cdn-anritsu.com/anritsu-infivis/en-gb/technical-note/technicalnote-vol6-en.pdf
• WIPOTEC, Weighing principle (EMFR): https://wipotec.com/en/weighing-principle
• WIPOTEC, Precision at high speed: the EMFR advantage (October 15, 2025): https://www.wipotec.com/en/innovation-magazine/precision-at-high-speed-the-emfr-advantage
• Thermo Fisher Scientific, A Practical Guide to Checkweighing and Checkweighers: https://documents.thermofisher.com/TFS-Assets/CAD/Handbooks/checkweighing-checkweighers-ebook.pdf
• OIML R 51-1 Edition 2006: https://www.oiml.org/en/files/pdf_r/r051-1-e06.pdf
• OIML R 87 Edition 2016: https://www.oiml.org/en/files/pdf_r/r087-e16.pdf
• Directive 2014/32/EU Annex VIII (MI-006) Chapter II (legislation.gov.uk): https://www.legislation.gov.uk/eudr/2014/32/annex/VIII/division/3/chapter/II?timeline=false
• Directive 76/211/EEC Annex I (legislation.gov.uk): https://www.legislation.gov.uk/eudr/1976/211/annex/I
• Directive 76/211/EEC Article 1 (legislation.gov.uk): https://www.legislation.gov.uk/eudr/1976/211/article/1
• Ministry of Commerce, Udon Thani Provincial Office, summary of key points of the Weights and Measures Act (2024): https://udonthani.moc.go.th/th/file/get/file/20241003d553ed022ea3de5be67d658e9759421d134534.pdf
• ASEAN, Guideline on Automatic Catchweighing Instrument Version 2.0: https://asean.org/wp-content/uploads/2026/05/ASEAN-Guideline-on-Automatic-Catchweighing-Instrument-Version-2.0.pdf
• Industrial Production Worldwide, Reducing food waste with intelligent checkweighing solutions (June 24, 2025): https://www.industrial-production-worldwide.com/production/reducing-food-waste-intelligent-checkweighing-solutions
• Thermo Fisher Scientific, Retailer compliance backgrounder: https://documents.thermofisher.com/TFS-Assets/CAD/Application-Notes/retailer-compliance-backgrounder.pdf
• WIPOTEC, Washdown hygienic design (October 25, 2025): https://www.wipotec.com/en/innovation-magazine/washdown-hygienic-design-built-for-the-toughest-cleaning-environments
• WIPOTEC, Wash-down checkweigher: https://www.wipotec.com/en/wash-down-checkweigher
• Thailand minimum wage revision (DLA Piper, July 2025): https://www.dlapiper.com/insights/publications/2025/07/thailand-increases-minimum-wage-across-all-provinces
• Thailand social security wage ceiling increase (Baker McKenzie, December 2025): https://www.bakermckenzie.com/en/insight/publications/alerts/2025/12/thailand-higher-wage-base-for-ssf-contributions
• ProPak Asia 2027 event overview (ThaiPR.net, June 19, 2026): https://www.thaipr.net/en/manufacturing_en/3732314
• Thailand food exports, first half of 2026 (Thansettakij, September 17, 2026): https://www.thansettakij.com/economy/trade-agriculture/669128