Chinese Herbal Soup Manufacturer | OEM, ODM and Private Label for Export Markets

A herbal soup that survives a sea container and a herbal soup that tastes as though it came out of a kitchen are usually two different products. The distance between them is not the recipe. It is a set of decisions about format, process and ownership, all of them taken long before the first bowl is filled.

Three variables decide which of the two you end up with. The first is which botanicals go in and at what grade. The second is how solids and liquid are balanced against the retort schedule. The third is how much of the process the buyer intends to own. The third one is the one that gets skipped in the first meeting, and it is the one that decides whether the recipe can be repeated by anybody other than the person who developed it.

Herbal soup used to be sold where the herbs were sold: a counter, a paper packet, and instructions from someone standing behind it. That trade still exists and it is no longer where the volume is. Volume now arrives from four directions at once. A supermarket buyer wants a shelf-stable bowl that photographs well and carries an ingredient list a shopper can read. A foodservice operator wants a case that opens without a tool and reheats without a stove. A gifting buyer wants a carton that survives a courier. A grocery chain serving a diaspora community wants the taste its customers remember, at a price the shelf can carry.

None of those four is buying a tonic. They are buying a soup that happens to be made from botanicals, and that shift redraws the specification. A tonic shop tolerated bitterness; a supermarket will not. A kitchen could finish the dish with salt; a retail bowl cannot. The category did not become easier to make when it moved into the aisle. It became easier to sell, which is a different problem.

Almost every product in this category is one of three formats, and the format settles more about the soup than the recipe does. The same nine ingredients produce three different products depending on which of these they enter.

A bowl carries its own heat source, so it can be eaten where there is no kitchen. The price of that convenience is structural. The shell, the inner container and the heater have to be designed together, the fill weight is capped by what the heater can warm reliably, and the shelf life belongs to the pack as much as to the soup. Our 2021 design patent on a rotating self-heating bowl shell came out of exactly that constraint. A bowl is the right answer for travel, office and gifting channels, and the wrong answer for anything sold on price.

A pouch is the cheapest route to a shelf-stable soup and the least forgiving on texture. Heat moves through a flat pack unevenly, the solids have to be sized for the pack, and a long cycle at sterilising temperature changes the structure of meat unless the raw material has been prepared for it. A 2022 study in Foods of a retorted ginseng chicken soup found that an enzymatic pre-treatment before retorting held water-holding capacity at 59.69 to 62.98 percent. It also reduced shear force from 23.05 N to 11.67 N, the practical distance between meat that shreds under a spoon and meat that has to be sawed.

A dried pack is not a soup. It is the ingredients for one, and it lives or dies on moisture control and on whether the household owns a pot. Its advantages are freight and shelf life; its disadvantage is that the product is only as good as the cooking instruction printed on it. A brand planning to sell all three formats should know in advance that they will not share a line, a specification or a shelf-life claim, and that the third one is a different business with a different buyer.

ZeaGrove was established in 2018 and runs more than thirty production lines in Guangzhou, with 350 to 400 staff and between 80 and 90 percent of output going to export markets. The formula library holds over 5,000 recipes, research and development takes more than ten percent of spend, and the company has held utility model patents since 2020 and design patents since 2021. It was recognised as a food-medicine homology innovation enterprise in 2023 and as a national high-tech enterprise in 2024. More than 2,000 brand clients had been served as of April 2024.

Those figures explain why a buyer can ask for a second flavour without waiting for a new supplier to be qualified. They do not describe capacity. What they say is that the line exists, that the people running it have run it for years, and that a specification written this month has somewhere to land.

A plant selling mostly into its home market can treat shelf life as one number and the label as one language. An export plant cannot. The same soup sold into three markets carries three ingredient lists, three net-content conventions, three sets of permitted claims and often three different storage statements, and one process has to remain valid for all of them at once. That is why the specification for an exported herbal soup is a matrix rather than a sheet: one process, several label files, and a written rule for which change forces a re-validation.

It is also why an export plant tends to be conservative about claims. A statement that is legal in one destination and not in another splits one recipe into two products, and the second one has to be validated from scratch. The cost of that split is not the translation. It is the second thermal validation, the second shelf-life trial and the second label review.

Buyers often open with a question about which certificates the plant holds. The question is reasonable and the usual answer is useless, because a certificate name without an issuer, a scope and a product list says nothing about whether your bowl is covered. A certificate issued for one process does not extend to a second process added later, and one issued for a category does not cover a new format.

What we will do is show the scope document with the product categories named, and say which of our lines it covers and which it does not. What we will not do is print a string of acronyms on a quotation and let the buyer infer the rest. When a certificate matters for a market, the useful question is which scheme that market's retailer accepts, and it is answered before the artwork is signed off.

Testing on this line is organised by the point in the process where a problem can still be corrected. Three sets of checks carry most of the weight.

Every lot of root, flower or seed is released against a written specification: species, grade, cut size and moisture, plus a screen for pesticide residues and for sulphur dioxide where the material is dried. The species name matters more than buyers expect, because two materials traded under one name can behave differently in a retort and carry different legal status in a destination market. A lot that fails any check is rejected at the gate rather than blended into a batch and averaged out.

Once the bowl is sealed, the checks move to the fill and the seal: net content against the declared figure, solids ratio against the approved sample, headspace, seal integrity before the retort, and the retort record itself. The botanical load belongs here too, because how much of a root ends up in the bowl depends on how it was cut and how long it was extracted, not only on the weight in the formula. A 2023 study in the Journal of Ethnopharmacology cooked chicken soup with Panax notoginseng at different ratios and for different times. The extracted saponins and polysaccharides moved with both, reaching 2.04 percent saponins and 44.3 mg per gram of polysaccharides at a ratio of 100 to 5 after 5.5 hours of steaming. The formula sheet alone does not predict what is in the bowl.

Finished goods are checked for label accuracy against the destination file, carton condition and the sensory standard retained for that product. A batch that differs from the retained standard in colour or aroma is a deviation even when every measured number passes. Microbiological release happens by lot against the criteria set for that product and that market, and the result is attached to the batch rather than summarised in an email.

The same soup sold to a restaurant group and to a supermarket is not the same product commercially, even when it is the same recipe. Foodservice buys volume per case, cares about opening and reheating, and will accept a pouch that retail would reject. Retail buys shelf presence: a bowl that reads at two metres, a carton that survives a pallet returned to the depot, and a label that passes the retailer's own review. Gifting buys the outer carton and the story. E-commerce buys a pack that survives a courier and photographs well.

Pricing follows the channel rather than the recipe, and a quotation that ignores the channel is renegotiated at the second order. The expensive mistake is accepting a retail brief and discovering at artwork stage that the unit price was built for foodservice.

Three variables account for most of the gap between a cheap herbal soup and an expensive one.

A retort cycle costs the same to run whether the basket is full or half empty, so a small batch carries a disproportionate share of changeover, cleaning and validation. The smallest sensible batch is set by the retort, not by the purchase order. It is the most common reason a quotation comes back higher than a buyer expected.

Each market needs its own artwork, and each artwork needs review against its own rules. Extra languages on one pack multiply the review work without multiplying the volume. A brand entering three markets with one SKU should budget for something closer to three labels than one.

Ambient, chilled and frozen are three supply chains with three cost curves. A product validated for ambient storage can travel by sea; a chilled product cannot. The choice is made for shelf-life reasons and paid for in freight, so it is worth making deliberately rather than inheriting it from the reference product.

The three terms are used interchangeably in enquiries and they are not interchangeable. What separates them is who owns the formula and who carries the development risk.

In an ODM arrangement the factory develops the product from a brief and owns the base formula, and the customer licenses it under their own brand. The customer gets a shorter route to market and the factory carries the development work. This is the route most first products take, and it is why a brand can list a soup within a season without hiring a food technologist.

In OEM the customer brings the formula and the factory makes it to that specification. In private label the customer selects from a catalogue and the pack carries their name. The distance between those two is smaller than the distance between either and ODM, because in both cases the recipe is settled. Both run on the same line, with the same parameters and the same batch file as an ODM product. What changes is who holds the recipe and who signs for it.

Four things have to be settled before a sample is useful: the destination market, the format and net content, the flavour reference, and the price band the unit has to sit in. With those, development starts against a specification. Without them, the first sample is a guess, and the guess tends to fail in the same direction every time: more salt than the market expects, more solids than the shelf can carry, a darker broth than the reference.

The sample round is a working session rather than an audition. The useful thing to ask for is the specification sheet behind the sample, because a sample that cannot be reproduced is a photograph. A buyer who receives a sample without a sheet has received a promise.

The label is where a supplier's responsibility and a brand's meet, and the handover has to be explicit. We prepare the technical content: ingredient names in the form the destination requires, net content, nutrition panel, allergen statement, storage conditions and preparation instructions. We check that content against the destination rules we have been given. The brand's regulatory contact signs it off, and the signed file becomes the reference for the print run.

When a market changes its rules, the brand tells us. Artwork is not updated by inference. That division sounds bureaucratic until the first label is rejected at a port, at which point it is the only thing anyone remembers about the project.

Some markets require a product registration or an import notification before the first shipment, and the documents differ by country. We support the process with the technical file: process description, ingredient specifications, test reports and label artwork. The registration itself belongs to the importer or the brand, because it attaches to a legal entity in that market rather than to the factory.

The same file structure supports a recall. A batch code connects a bowl to its fill record, its retort chart, its test results and the lots of each botanical that went into it, which is the first thing a regulator or a retailer will ask for. A recall reconstructed from memory takes weeks and produces a version of events nobody can defend. A recall read from a file takes an afternoon.

Release runs through three gates, and a batch has to pass all three before it is stored as finished goods. Each gate exists to catch a different class of failure, and no gate can substitute for another.

The first gate is the only one that catches a process failure, because it reads the retort record rather than the product. The second catches a packaging failure. The third catches a documentation failure, and it is the gate most often skipped when a factory runs late, which is why a retailer's auditor opens it first.

Every batch leaves a retained sample behind, held under the same conditions as the commercial stock and checked at intervals against the sensory standard. The sample is not a souvenir. It settles a complaint: if the retained sample has changed, the product changed before it left; if it has not, the change happened after.

Retained samples also carry the shelf-life trial forward, because behaviour at month nine is easier to observe on a warehouse shelf than on a customer's. The set is kept for the period the destination market requires, and its location is recorded with the batch, not with the person who put it there.

Production scheduling on a multi-SKU line is a constraint problem, and the constraint is the retort. Formats cannot be mixed in one basket, recipes with different schedules cannot share a cycle, and a change of format costs a changeover plus a check that the validated parameters still describe what is in the vessel. The practical consequence is that the production window belongs to the format rather than to the order.

A second flavour ordered late in the cycle may therefore have to wait for the next window rather than the next week. We publish the format calendar to customers with confirmed orders for that reason: a date that is known is easier to plan around than a date that is negotiated twice.

A pallet of finished bowls was moved from a temperature-controlled store into the dispatch hall on a humid morning. The cartons had been specified for dry storage, and the specification was correct. What had not been written down was how long a pallet could stand in a warmer, wetter room before the corrugated board took up moisture. Nothing on the dispatch sheet recorded it.

The problem was found at the pre-shipment packaging check, not by the operator who moved the pallet. The bottom layer had lost stacking strength, and a compression test on a sample showed the board below the figure the pallet pattern assumed. The pallet was held in the warehouse, the affected cartons were replaced, and the order shipped on a later pallet under the same batch code.

What changed afterwards was a record item and a parameter. The dispatch record now carries the temperature and humidity of the loading hall, and the time a pallet spends outside controlled storage is written down and limited. Cartons for humid-season shipments are specified with a higher moisture resistance, and the pallet pattern is checked against the carton's compression figure rather than its dimensions. The check sits on the loading sheet, the one document a warehouse supervisor signs.

Every botanical in a formula has at least two approved suppliers, so a crop failure does not stop a line. The cost is that a substitution is not free. A different origin can change colour, moisture and the way a root behaves in a retort, and a batch made with the second source is checked against the retained standard before release.

Buyers who ask for the same supplier every time are asking for continuity, which is reasonable. They should know that continuity is what the second-source rule exists to protect, because a single-source formula is one bad harvest from a missed season.

We decline three categories. We do not take on medicinal products, meaning anything that has to be registered as a drug in the destination market. We do not take on concepts that only sell if the pack makes a disease claim. And we do not take on formulations built around an ingredient whose identity, origin or legal status cannot be documented, because that material cannot be defended at a border or in a retailer's audit.

Product formats are a different question. A format this line does not currently run is scoped as a project, with its own tooling and validation, and discussed before a quotation rather than ruled out in an email. One answer is a boundary on what can be sold. The other is a schedule.

The fastest route to a useful answer is a short brief and a reference pack on the table. Everything after that is specification work, and every item in it carries a name: the format, the process schedule, the label file, the release checks. None of it is improvised at the purchase order.

Send the reference product even if it belongs to a competitor, and say what you want different. A physical reference settles questions about broth darkness, solids visibility and herb intensity that no brief answers as quickly. It also shortens the sample round, because half of the first round is spent discovering what the buyer did not want.

ZeaGrove (Guangzhou) Health Industry Co., Ltd runs this line in Guangzhou and exports most of its volume. The next step is a specification scoped to your format and your market, not a price list.

The smallest batch is set by the retort and the changeover, not by a published figure. A basket has a working range, and a batch below it pays the same cleaning, heating and validation cost across fewer units. We quote per format and per recipe because those two variables move the answer more than order size. A buyer who gives us the format, the market and an annual volume gets a recommendation that reflects what the line can run.

It can be developed for both, but it cannot be validated for both at once. A pouch and a bowl heat differently, so each format needs its own process schedule, its own packaging trial and its own shelf-life work. The recipe is the starting point, not the product. Brands that plan both should budget for two validations and expect the sensory profile to be adjusted between them.

We will show the scope document with the product categories named and say plainly which lines it covers. We will not print a list of acronyms on a quotation, because a certificate name without an issuer and a scope tells a buyer nothing about their own product. If you tell us the destination market and the retailer, we can say which scheme that buyer is likely to ask for and whether our scope covers it.

The brand signs the label file; we prepare the technical content and check it against the destination rules we have been given. That split is written into the project record rather than agreed verbally. When a market changes a rule, the brand tells us and the artwork is revised. Artwork is never updated by inference, which is the mechanism behind most label failures at a port.

Yes, as a separate product with its own specification. A dried pack is judged on moisture, cut size and the cooking instruction, and it is made on a different line from the bowls. It shares the botanical sourcing and the release standards, and it does not share the retort schedule, the shelf-life claim or the buyer. Treat it as a second project rather than a pack variation.