IQF Mango Specification: What to Put in the Purchase Order

Every IQF frozen mango supplier Europe buys from quotes the same offer: 10×10 mm dice, -18 °C, 10 kg carton, 24 months. Those four lines decide nothing. No Codex standard exists for quick-frozen tropical fruit, so the clauses that decide acceptance are yours to write: variety, free-flow, drip loss and the sampling plan.

Why do three IQF mango offers read the same?

Because there is no published standard behind them for them to differ from. Codex Alimentarius has a quick-frozen standard for vegetables and four for temperate fruits. It has none for mango. There is no USDA grade standard for frozen mango either, and EU marketing standards apply to fresh fruit only.

That absence is not a technicality. It means the words Grade A and conforms to international specification carry no external content on this product. When a lot arrives and the dice are paler than the sample, a seller who answers that no standard says otherwise is factually right, and being right is what wins that argument. The only thing that answers it is a signed specification with a signed defect photo set behind it.

A GFSI scheme does not close the gap. The certification held through the partner facility tells you that a food-safety system was audited against a scheme. It does not set a tolerance for clumping, and no auditor will reject your container on colour.

Which mango variety is in the dice, and what does that clause control?

It controls colour, texture and fibre, which is most of what you will argue about later. Vietnam does not freeze one mango. Dice comes from Keo and from processing-grade Cát Chu, bought on firm flesh, short fibre, Brix at or above 14 and absence of jelly seed. Large cheek and chunk comes from green-skin xoài tượng, Tứ Quý and Taiwanese varieties, bought on fruit above 600 g with a thick cheek and a flat stone.

Cát Hoà Lộc, the variety a marketing page names because it is the famous one, is kept out of routine frozen lines on flesh softness and raw-material cost. Variety is also a market gate rather than a preference: Japan accepts only Cát Chu among Vietnamese mangoes, and has done since 2015.

The two dice varieties are not even harvested at the same point. HungHau Foods records Keo taken at 75 to 80 percent maturity, with 30 to 50 percent of the peel yellow, and Cát Chu at 80 to 85 percent, with about 70 percent yellow. Keo eats mildly sweet and low in acid; Cát Chu is sharply sweeter and can carry acid if it went in short of ripe. Keo comes off alluvial land in Đồng Tháp and Tiền Giang, Cát Chu from Đồng Tháp and Ninh Thuận.

That difference is what a colour tolerance is measuring. The fullest published specification for this product, the commercial document GG-2616 for IQF Mango Chunks 3/4 inch, permits at most 10 percent pale yellow pieces and at most 5 percent deep orange. Those two percentages are a variety-and-maturity clause wearing a colour label, and neither is enforceable in a purchase order whose product description reads Vietnamese mango.

Brix shows the same problem from the other side. CBI’s buyer sample specification for IQF mango chunks sets 12 to 13 °Bx. Published Vietnamese supplier specifications for the same product carry 14 to 20 °Bx. A cross-commodity specification frame puts it at 12 to 16 depending on variety. All three are accurate and none is interchangeable with another: the buyer figure is an acceptance window narrow enough to enforce on one variety at one maturity, the supply figure is a seasonal and varietal range across Cát Chu and Keo. Fix Brix per variety and per season in the annex, name IFU No. 8 as the method, and record the measurement temperature, or the clause cannot be enforced in either direction.

Two lines exist for mango and for nothing else on a frozen fruit list: fibre and percentage of peel. Write them, one variety per lot, no mixing. A specification that never names the variety contains no clause a substitution can breach; the substitution surfaces months later as colour spread and texture spread, inside tolerances that were never written to catch it. Naming variety, form and size rather than Vietnamese mango also keeps the export documentation set consistent, because a vague goods description is what invites a classification query at the border.

One correction worth having, because it circulates in Vietnamese trade press as a flat statement: green-skin and Taiwanese mango is reported as unsuitable for processing. Read strictly, that means unsuitable for soft-dried product and for purée, on low Brix and pale colour. Large frozen cheek lines and green-fruit lines still use it.

How should the cut, size and fines clause read?

Name the cut, the tolerance, the counting basis and the sample size, in that order. Dice at 10×10 mm, chunks at 20 to 25 mm and whole cheeks are the conventions Vietnamese plants run. The tolerance is the part buyers leave out. GG-2616 allows a maximum 15 percent of pieces outside its 19 to 20 mm range; CBI’s buyer sample asks for at least 85 percent within 12 to 18 mm. Those are the same idea stated from opposite ends.

Fines have no published limit anywhere. Around 5 percent by weight appears as a starting position and industrial buyers commonly push it to 3 percent, but that is a negotiated number, not a standard, and it is worthless without a named sieve aperture. Two rules travel with it. Sieve while frozen and count after thawing: sieving thawed product breaks pieces, so every breakage percentage comes out biased against the seller, who will reject the result and be right to. And state whether the percentage is by weight or by count, because on thin-cut product the two answers differ several-fold.

Write down who owns the broken and the fines. They run on the order of 10 to 20 percent of output, and if the contract is silent they belong to the processor by default.

Form also moves the tariff line, not just the texture: whole fresh or dried mango sits under HS 0804.50 while frozen dice, chunk and cheek move to 0811.90. The same mechanism decides how the tariff heading shifts between dried jackfruit and jackfruit chips, and it is worth settling before the invoice is raised, not after.

How do you prove IQF on the bench?

Pour 1 kg of product, still frozen, onto a tray. Separate it with light hand pressure. Weigh the portion that still adheres. That is the whole test, it takes four minutes, and it is the only thing that distinguishes IQF from block-frozen fruit sold under an IQF label.

The clause becomes enforceable only when it defines the unit of adhesion. GG-2616 splits it: clumps are groups of up to three pieces, held below 3 percent; agglomerates are groups of five or more, held at 3 percent. The CBI buyer sample specification for the same product sets clumping at a maximum of 2 percent, with no split by piece count at all.

Neither is wrong. They are different instruments doing different jobs, and the boundary is worth knowing before you copy a number off one of them: GG-2616 is a supplier’s product specification that defines its own unit, and the CBI figure is a European buyer’s acceptance ceiling that does not. A purchase order saying clumping ≤3% without defining a clump has bought the argument rather than avoided it.

CBI records the general wording European buyers use for this product:

good, reasonably uniform colour · free from foreign matter · for IQF, the pieces must not stick to one another

The third phrase is a specification only once somebody attaches a tray, a mass and a piece-count definition to it. Until then it is a sentence both parties can read in their favour.

How is drip loss measured, and what does it tell you?

Drip loss on IQF mango is measured like this. Take 200 g of product still frozen and record the weight as m1. Transfer it to a 2 mm sieve set over a tared bowl. Thaw at 4 °C for 18 hours. Weigh the solids remaining on the sieve as m2. Drip loss is m1 minus m2, divided by m1, as a percentage. The number is an indirect measurement of how fast the fruit was frozen, and it is the cheapest one a buyer sitting in Europe can run without visiting the plant: slow freezing grows large ice crystals, the crystals rupture cell walls, and the cell contents leave as liquid on the bench. The thaw regime has to sit inside the clause, because 4 °C for 18 hours and 20 °C for one hour do not produce the same figure on the same product.

VinFruits Global asks the partner facility for the thaw protocol behind a drip-loss figure before a specification is agreed, because a percentage quoted without its regime is two different numbers.

For a reference band, the B2B platform Nutrada publishes 3 to 8 percent drip loss for IQF against 10 to 20 percent for block-frozen fruit. Treat that as a supplier-published range and never as a standard; it is the most useful competitor datum in this trade and it arrives with no thaw protocol attached.

What produces the number is upstream of the tunnel. Dewatering before freezing, by vibration or air knife, is the step that decides drip loss, and incomplete dewatering is also what causes pieces to weld together in store. Mango is typically run through a fluidised-bed or spiral tunnel at -35 to -40 °C for 8 to 15 minutes, with the free-flow check taken at tunnel exit. None of that is auditable from Europe, which is precisely why the bench tests belong in the contract instead.

Which temperature and sampling clauses belong in the order?

Three: a core temperature with a measurement method, a sampling plan with a name and a number, and an acceptance ladder that separates safety from cosmetics. Codex CXS 320-2015 sets the product core at -18 °C or colder, held there through the chain, and excludes glaze from declared net weight. Codex CXC 8-1976 adds the distribution rule that the warmest package must not exceed -12 °C.

Measurement detail decides who wins a dispute. Under CXC 8-1976, destructive measurement drills a hole at least 2.5 cm deep; non-destructive measurement taken between cartons carries a 2.8 °C tolerance; the probe must read to ±0.5 °C with a 0.1 °C display. Put a single-use data logger in every container in three positions, near the door, mid-load and at the end, and make a temperature breach an independent ground for rejection, separate from any sensory finding. Add a dedicated cold-store clause: not stored alongside seafood, onion or garlic.

Sampling comes from Codex CXS 233-1969, AQL 6.5, Level I, for pack units above 4.5 kg, with Level II on dispute.

One AQL cannot cover everything in the lot. Split by severity. Critical findings carry an AQL of zero and reject the lot without negotiation: metal, glass, hard plastic, a Salmonella or Listeria positive, an untraceable lot code, a core warmer than -12 °C. Major defects run at AQL 2.5, minor at 6.5. Then write the consequence in advance: critical means the lot is rejected and the seller carries storage, destruction or re-export; major over AQL means 100 percent re-sorting at the seller’s cost, then a downgrade against a published price scale, then return, in that order; minor over AQL means the price scale, logged against the supplier, with three consecutive lots triggering a specification review. A 40 ft reefer of 10 kg cartons is weight-limited at roughly 20 tonnes, so a rejected container is an expensive thing to have no procedure for.

What changes when the pack is declared ready-to-eat?

The microbiological regime, and the shelf life it runs across. Ready-to-eat is not a physical property of frozen mango; it is a consequence of how the seller declares intended use, and that declaration binds your downstream use too. Under Regulation (EC) 2073/2005, entry 1.19, pre-cut ready-to-eat fruit and vegetables carry Salmonella n=5, c=0, not detected in 25 g, applicable throughout shelf life.

The amending Commission Regulation adopted on 20 November 2024 applies from 1 July 2026. For ready-to-eat foods able to support L. monocytogenes growth it sets two tiers: 100 cfu/g where the operator can demonstrate to the competent authority that the limit will not be exceeded through shelf life, tested by EN/ISO 11290-2, and otherwise not detected in 25 g by EN/ISO 11290-1. Both apply to product on the market across its whole shelf life, which on IQF mango is 24 months.

There is a documented way out, and a price for it. Suppliers label frozen fruit as an ingredient that must be heated to at least 75 °C for 30 seconds before consumption, which takes the pack out of the ready-to-eat class. The smoothie, ice-cream, yoghurt and dairy customers who pay best will not buy a bag that says heat before eating. Whether frozen fruit sold into a thaw-and-blend channel supports growth for the purposes of the 100 cfu/g tier is genuinely unsettled; put that to a food lawyer or the competent authority in the importing member state rather than deciding it internally.

Underneath all of it: an IQF line has no kill step. There is no blanching and no pasteurisation. Freezing arrests micro-organisms, it does not destroy them, and everything that enters the line leaves with the product. Recalls on frozen fruit between 2020 and 2026 have been dominated by Listeria and by hepatitis A and norovirus rather than by foreign bodies, including the SunOpta and Sunrise Growers recall of roughly 162,000 cases in June 2023, a frozen strawberry hepatitis A outbreak across 32 US states in 2023, and the Greenyard and Baja case with 53 illnesses and 10 deaths.

The sample mass is where the money is. The law asks for 25 g; large EU and Japanese buyers contract for Salmonella absent in 375 g and Listeria absent in 125 g. Japan’s Food Sanitation Act sets frozen food not intended to be heated at TPC of 100,000/g or less with coliforms negative. Ask for the plant’s environmental monitoring results from the last three to six months alongside the certificate of analysis: BRCGS Food Safety Issue 9, clause 4.11.8, requires a documented environmental sampling procedure with action limits and trending, and a plant that runs none has not measured the risk its certificate of analysis is silent about.

What do you need to know about residues, foreign bodies and metal detection?

That the enforcement arithmetic runs the opposite way to how most buyers read it, and that the regulatory buffer is not yours. Under SANTE/11312/2021 the default expanded measurement uncertainty for enforcement is 50 percent, k=2, at 95 percent confidence, and a result counts as exceeding an MRL only where x minus U is greater than the MRL. A reading of 1.4 mg/kg against a 1.0 MRL is not an official exceedance.

Read it the other way and the risk appears: a lot measuring 0.9 may truly be 1.35. Your commercial contract contains no uncertainty rule unless you write one, so set the internal threshold below the MRL rather than on it, and state in the arbitration clause which comparison applies, x against the threshold or x minus U against it.

MRLs are set on the commodity as harvested, in most markets the whole fruit including peel: the EU per the Annex I definitions, the US per 40 CFR 180.1(h) as the raw agricultural commodity as a whole, Japan under the Positive List with the stem removed and the peel on. Regulation (EC) 396/2005, Article 20, with guidance SANTE/10704/2021 Rev 1 in force from 19 March 2025, makes the processed-product MRL the processing factor multiplied by the unprocessed MRL. The importing authority decides whether a factor applies and the evidential burden is entirely the operator’s. Peeling generally drives the factor down, and mango is among the commodity lines where peeling changes the arithmetic most, but do not assume the direction: on banana in the US, azoxystrobin sits at 2.0 ppm on whole fruit and 0.1 ppm on pulp under 40 CFR 180.507, twenty times stricter on the part actually eaten.

Heavy metals behave the other way. They enter through the root rather than the skin, so peeling barely helps, as the cadmium findings in frozen pineapple from Costa Rica showed. Regulation (EU) 2023/915, which replaced 1881/2006, sets lead in fruit generally at 0.10 mg/kg; purée and juice carry their own Annex I entries that have to be read directly. Two items save money rather than cost it: patulin does not apply to tropical fruit, so paying for it on a default certificate of analysis is waste, and sulphite above 10 mg/kg must be declared under EU labelling, which is why many buyers simply specify sulphite free.

For foreign material, the USDA AMS Foreign Material Manual gives a usable class structure: Class 1 at 2 mm or less requiring magnification, Class 2 at 2 to 7 mm, Class 3 at 7 mm and above and objectionable, Class 4 visible and hazardous, and GNC covering animal filth, sharps, petroleum products and glass, which is not acceptable at any level. On metal detection, the Fe 2.0 / non-ferrous 2.5 / stainless 3.0 mm trio is quoted across the supplier layer as though it were a standard. BRCGS Issue 9 requires the site to validate detector sensitivity on the product itself and specifies no figure at all, and frozen product degrades real sensitivity. Ask for the validation report on this product rather than for the three numbers.

Which documents, tariff line and origin questions travel with the lot?

Frozen mango leaves HS 0804.50 and goes to 0811.90, and a phytosanitary certificate is a fresh-goods document. The documents that actually accompany a frozen lot are shorter than most buyers expect:

• commercial invoice and packing list
• bill of lading
• Vietnamese export declaration
• certificate of origin under the applicable free trade agreement
• certificate of analysis, with the test date inside the delivery window

CBI states that mango purée does not require a phytosanitary certificate for the EU. The same logic reads across to IQF, but it has not been confirmed line by line, so confirm it with the importing country rather than generalising from purée. Processed lines also need no planting-area code, no irradiation and no vapour heat treatment. Planting-area codes and packhouse codes are issued by Vietnam’s Cục Trồng trọt và Bảo vệ thực vật and published at ppd.gov.vn; they attach to the named grower or packer and cannot be held by a trading company. They are also suspended in practice: on 22 July 2023 GACC notified phytosanitary breaches on Vietnamese banana, jackfruit, mango, lychee, dragon fruit and durian, and the Vietnamese ministry announced suspension of the codes involved.

Two things worth knowing that rarely reach a European desk. Mango does not appear in Annex II of Regulation (EU) 2019/1793, the enhanced official controls list that catches Vietnamese chilli and okra at 50 percent, dragon fruit at 30 percent and durian at 20 percent. And Vietnam charges no export duty on mango, with VAT at zero and no separate export licence, so the requirements that matter all sit on the importing side. Against that, from 29 January 2026 the EU moved from a risk-based to a hazard-based approach and applies a default MRL of 0.01 mg/kg to high-concern active substances, which tightens the residue side rather than the border side.

Watch the origin question if the raw material crossed a border. Cambodian output runs around 2.25 million tonnes a year, roughly 80 percent Keo Romeat, with the season running February to June and overlapping the Mekong Delta crop of March to July almost exactly; about 86 percent of Cambodia’s recorded mango exports in a four-month period went to Vietnam. Keo imported from Cambodia and processed in Vietnam is no longer wholly obtained. Cambodia is a party to ACFTA, RCEP and ATIGA, where cumulation is possible, and not to EVFTA, CPTPP or VKFTA. The move from 0804.50 to 0811.90 is a change of heading, but whether that satisfies the rule depends on each agreement’s product-specific rules annex, and that question is unresolved. Get a written view from VCCI or the Ministry of Industry and Trade before you claim preference.

On price context rather than price: HungHau Foods published an average export price of 2.28 USD/kg for Vietnamese frozen mango on 8 August 2025, against about 3.3 USD/kg for Thailand, with Vietnam’s total HS 0811 frozen fruit exports at roughly 240 million USD in 2024. That is a year-old published average, not a current quotation. For competitive framing, FreshFruitPortal reported on 25 November 2025 that Peru expected more than 67,000 tonnes of IQF mango in 2024-25, with frozen making up 38 percent of Peruvian mango export value in 2025. Note also that 0804.50 groups guavas, mangoes and mangosteens in one subheading, which is why one platform reported 2.22 USD/kg for Vietnam in 2023 against about 0.69 USD/kg in Chinese customs data for actual mango. Reconcile an invoice against a trade database only after you know which of those two numbers you are looking at.

What do you write when nobody publishes the tolerance?

You write it yourself, and you make it a signed, version-numbered annex to the contract rather than a PDF attached to an email. Whoever writes the specification first sets the rules, and a factory’s own specification is written to protect the factory. Keep three kinds of threshold apart, because they cannot share one AQL: legal limits such as MRLs, heavy metals, Salmonella and core temperature, where no price reduction exists; an internal safety margin; and commercial thresholds such as breakage, colour and clumping, which reduce to money.

Two supports make the table work. Without a signed defect photo set, every percentage is disputable, because two competent people count differently and both are right. Test the set cheaply before signing it: give the same five samples to three assessors independently, and if they agree within 20 percent relative, the set works. And a certificate of analysis is not a quality certificate. It states which parameters were measured, on which sample, with which result, and anything absent from it is something nobody measured. The fields most often missing are the sampling date and plan, the method reference for each parameter, the LOQ and LOD, and the measurement uncertainty. ISO/IEC 17025 accredits individual tests on individual matrices, not laboratories, so ask for the scope document and point at the line covering the test you need on frozen fruit.

Where nothing is published at all, buy the number. There is no ΔE colour tolerance and no post-thaw firmness figure in newtons for IQF mango anywhere in public sources. Buy three to five commercial samples in your own channel, measure them, and set the tolerance at mean plus or minus two standard deviations. One last design rule for anyone buying frozen mango wholesale across several grades: grades A, B and C may differ on cosmetic defects only. They never differ on safety, and a supplier who offers looser microbiology on a lower grade is a supplier to drop.

Common questions

What does IQF commit a supplier to that plain frozen does not?

IQF is a claim that the pieces stay separate. It is testable: pour 1 kg still frozen onto a tray, separate with light hand pressure, weigh what still adheres. The clause is only enforceable if it defines the unit of adhesion, since clumps of up to three pieces and agglomerates of five or more are counted differently. Block-frozen product cannot pass that test at all.

Is there an international standard for frozen mango grades?

No. Codex has a quick-frozen standard for vegetables (CXS 320-2015) and four for temperate fruit: strawberries CXS 52-1981, raspberries CXS 69-1981, peaches CXS 75-1981 and blueberries CXS 103-1981. There is no Codex standard for mango, no USDA grade standard for frozen mango, and EU marketing standards cover fresh fruit only. Every tolerance in an IQF mango purchase order is a number the two parties set themselves.

How is drip loss measured, and what is a normal figure?

Weigh 200 g still frozen, transfer to a 2 mm sieve, thaw at 4 °C for 18 hours, weigh the solids and take the loss as a percentage of the starting weight. The B2B platform Nutrada publishes 3 to 8 percent for IQF against 10 to 20 percent for block-frozen fruit. That is a supplier-published band, not a standard, and it means nothing without the thaw regime beside it.

Does frozen mango need a phytosanitary certificate for the EU?

A phytosanitary certificate is a document for fresh plant material, and processed lines generally fall outside it. CBI states plainly that mango purée does not need one for the EU. For IQF mango the same logic applies but has not been confirmed line by line, so ask the competent authority in the importing country before you write the document list into the contract.

What changes if the pack is declared ready-to-eat instead of an ingredient?

Ready-to-eat is a declaration, not a property of the goods, and it decides which microbiological regime applies. Under Regulation (EC) 2073/2005 entry 1.19, pre-cut ready-to-eat fruit carries a Salmonella criterion of not detected in 25 g across the whole shelf life, which on frozen mango is 24 months. Declaring the pack an ingredient to be heated to 75 °C for 30 seconds removes that, and removes most smoothie and ice-cream buyers with it.