Self-Heating Herbal Soup Contract Manufacturing Specialist | Full-Process QC and Testing

A self-heating herbal soup is easy to describe and hard to release. The description takes one sentence: a bowl of soup with a heating pack under it. The release takes a validated retort profile for that specific bowl, a filled weight that survives a buyer's own check, and a heating pack that performs the same way on batch fifty as it did on batch one.

This page is about the release side. It covers what is in the bowl, how the recipe and the process parameters are set, what the label is allowed to say, and what has to be in the file before a batch leaves the plant.

The physical product is a sealed bowl holding a cooked soup, with a heating pack and an outer container around it. The soup is sterilised in the bowl at 121 °C and hermetically sealed, so shelf stability comes from the process and the seal rather than from an added preservative.

Net content runs at 258 g as the standard bowl on this line, with 168 g and 228 g also available. Solid content is set at a floor of 20%, which is a weighable figure rather than a description of how generous the bowl looks.

Three things about the format matter more than the recipe when a buyer is comparing suppliers.

• Net content is a weight, not a volume. The fill is dosed to a weight target and then check-weighed, so the number on the label is the number the buyer's own balance will find.
• The solid content floor is a specification. A bowl that delivers broth and nothing else is a different product at a different cost.
• The pack is opened, not prepared. It is eaten straight from the bowl, so the aroma at the moment of opening is the product experience rather than a starting point.

The three sizes share a recipe and differ in fill weight, and each size carries its own retort validation because the filled geometry changes what the cold point does. A buyer who wants a fourth size is asking for a tooling project, not a line change.

A precaution line is written per recipe rather than per brand. Where a recipe includes Dendrobium officinale or Ganoderma lucidum, the label carries a statement that the product is not recommended for pregnant and lactating women and infants. That statement is part of the artwork, and it is set by the recipe rather than by the market.

The lists are short. A typical bowl reads water, the named meat, the named botanicals, ginger and edible salt, with nothing in it that the label does not name. Herbal components here are used as food ingredients with a long history in Chinese cooking, which is what the term food-medicine homology refers to: ingredients recognised in Chinese tradition as both food and medicine.

That history is a cultural fact about the ingredients. It is not a health claim, and the label language derived from it is limited to what the recipe contains and how the soup has traditionally been eaten.

The plant behind this line was founded in 2018 and runs more than 30 intelligent high-speed production lines, with 350 to 400 employees. Between 80% and 90% of output is exported. The company has served more than 2,000 brand customers as of April 2024 and holds a library of over 5,000 recipes.

R&D spending has stayed above 10% of annual revenue. Six patents in this range are directly relevant to a self-heating bowl. Utility models granted in 2020 cover a packaging bowl, a sterilisation device and a production device. Design patents granted in 2021 cover a rotating self-heating bowl shell and a visible bowl-shaped container.

The figures above are the ones a buyer can check against a company record or a public registry. Three categories are deliberately absent from this page: floor area, daily output and annual revenue. They are absent because the numbers that appear elsewhere for each of them disagree with one another, and a supplier who quotes a conflicting figure in an article and in an audit is not worth the trouble of verifying.

Ask for the site registration, the scope on any certificate, and the batch file from a production run. Each of those is a document with a date and a scope on it, which is worth more than a capacity number quoted in a brochure.

Everything printed on the bowl falls into two buckets: statements about what is in the product, and statements about what the product does. The first bucket is a labelling exercise. The second is a regulatory one, and it is where packaging gets refused at a port.

The line this plant draws is consistent across categories. A label may name the ingredients, describe how the product is prepared and stored, state net content and solid content, and refer to the traditional use of its ingredients as food. A label may not claim to treat, prevent or relieve a condition, and may not carry a disease name at all.

The most common one is a formulation described in terms of an effect. It reads naturally, it sells, and it is the single easiest way to lose a listing. Rewriting it takes ten minutes at the artwork stage and a reprint after the plates are made.

The destination market decides which of the remaining statements survive. A file assembled for one market is not a file for another, and the label language, the precaution statement and the permitted nutrition or content statements all have to be re-checked per destination rather than copied across.

A certificate covers a site and a scope. It says that a management system was audited against a named standard at a named location over a defined period. It does not say that a specific recipe is safe, and it does not travel with a bowl.

That distinction is well documented in the food safety literature. Kök reported in the Journal of Food Protection in 2009 on the application of ISO 22000 and HACCP in the Turkish poultry industry, and found that implementation differed with enterprise size rather than being uniform across the sector. The system is a documented way of working at a particular site, and the audit is a check on the documentation rather than a statement about a product.

A recipe that works in a kitchen is a starting point, not a specification. The retort cycle is the most aggressive step the product will ever see, and it changes colour, texture and aroma in ways that have to be designed around rather than corrected afterwards.

The development question is therefore not how the soup tastes when it is made. It is what it tastes like after it has been filled, sealed, sterilised, cooled and stored for a year.

Three properties move under retort, and all three are scored at sampling rather than assumed.

• Salt and seasoning perception. The same added salt reads differently after the process and after storage, so the seasoning is set against the finished, stored bowl.
• The texture of the solids. A root or a seed that holds its shape at the table can soften or break up in the bowl, and a piece that breaks up changes how the soup pours.
• Aroma at the moment of opening. The pack is opened rather than prepared, so the first impression is the whole impression.

The target is set at the panel, on the finished bowl, against a written threshold. A recipe approved on a kitchen sample and released to production without that step is a recipe that will be reformulated after the first shipment.

Piece size is a process parameter, not a presentation detail. It decides how fast the bowl heats, how the soup pours and whether the solids settle in transit. Reformulating a traditional sauce for a retortable pouch is a development project with quality targets attached, not a transcription exercise, and Cheenkaew and colleagues set that out in PLoS One in 2020 while working on a Pad Thai sauce.

More than 5,000 recipes are held in the library, and they shorten the first round rather than replacing it. A starting point that already behaves under retort saves one or two sampling rounds. The balance of the work is still the part that is specific to your product: the ratio, the piece size, the salt level and the target bowl.

Most buyers arrive with a channel and a customer in mind and no appetite for tooling three sizes before anyone has bought one. That is a reasonable position, and the cost structure supports it.

The fixed costs in this format are the bowl tooling and the retort validation. The variable cost is the material. A first run therefore makes sense when it is sized around one print run and one validation rather than around a volume target.

A range can also be built one product at a time. Adding a second recipe once the first has sold is cheaper than developing three at once, because the bowl, the carton and the validated profile are already in place and only the recipe is new.

They differ in who owns the recipe and who owns the development risk.

• OEM. You bring a locked recipe and a specification; this line runs it and releases it against the documents described further down.
• ODM. You bring the channel, the customer and the price point; the recipe is developed here from the library and the sampling rounds, and the specification is written as the work proceeds.
• Private label. You bring a brand and a market; the product is made to an existing specification and the label carries your identity.

The validation and the tooling are fixed, so a very small first run carries a high cost per bowl. That is a scoping conversation rather than a published minimum, and the useful question at the start is which part of the range has to exist on day one.

Sizes, packaging print and carton counts can be extended without touching the recipe that was validated. A recipe change is a sampling project and a re-validation, which is why the first specification is worth getting right rather than getting fast.

Development on this line runs in stages, and each stage ends with a document rather than with an opinion. That is what makes the project reviewable by someone who was not in the room.

The sequence is fixed even when the content is not: a brief, a sampling round, a scoring round against a target, a locked specification, a bowl and tooling decision, a validation, and a first production run.

The point of the staging is that nothing is decided twice. A locked specification means the sampling rounds stop, and a validated profile means the first production run is a repeat of something already demonstrated rather than a test.

No stage in that table carries a promised date. Lead times depend on tooling, on the market file and on approval rounds, and each of those moves with the project rather than with the calendar.

The line runs one sequence, and the sequence is what the validation is written against. The table below lists the stages with the parameters that are controlled at each one.

Two stages in that table decide more than the others. Filling sets the weight the buyer will measure, and it also sets the packed density that the retort profile was written for. Sealing decides whether the bowl is still hermetic after the cycle, which is the whole basis of a room-temperature shelf life.

Everything after sealing is a check rather than a correction. Once the lid is on, the soup inside it cannot be adjusted.

The retort sits after sealing and before cooling, and it is the step that makes the product shelf-stable without refrigeration. It is a fixed parameter for a validated bowl, not a dial that gets set to a job.

Fill temperature affects the load the retort has to deliver and the headspace that remains after sealing, so it is controlled as a window rather than as a floor. A fill that runs cold changes the cycle rather than only the product.

The headspace left above the soup absorbs the pressure change during the heating and cooling phases of the cycle. It is set by the fill specification and confirmed at sealing, and a bowl with too little of it is a bowl whose lid film is under load before it ever leaves the plant.

A batch is released against three checks and a document set. Two of them are laboratory and panel work on the product, and the third is on the pack.

Release is a decision made against records, not an outcome of production finishing. The batch file is assembled first, and the release is signed against it. A batch that has passed every check and has an incomplete file is not released, and that rule exists because the file is what answers a complaint six months later.

The laboratory measures net content, solid content, salt and foreign matter against the specification. The record names the method as well as the result, so that a disagreement with a buyer's laboratory is settled by method rather than by argument. The panel scores the finished bowl blind against the target, with salt broken out as its own attribute. The pack check covers seal integrity, lid condition, code legibility and carton count, and the seal is sampled per run rather than once per batch.

A retained sample per batch is held beyond the declared shelf life, and the batch file is kept under the lot number. The sample is what allows a late complaint to be tested against the physical bowl rather than discussed.

A certificate of conformity is issued with the shipment, listing the lot and the checks performed. It is not a substitute for the batch file, and it is not a certificate of analysis from a third-party laboratory unless one has been commissioned for that shipment.

The third change is the one that matters. A chart that is filed is a record. A chart that is read against a profile is a control, and the difference between those two is the difference between catching this on the day and catching it in a warehouse.

Three kinds of enquiry get a no, and the third one saves buyers the most money in the long run.

1. Anything positioned as a medicine. No disease names, no treatment claims.
2. Concepts that only work if a disease claim carries them. Take the claim away and there is no reason for the customer to buy.
3. Ingredients whose identity cannot be documented. If the paperwork does not exist, the material does not go in the bowl.

A format this line does not currently run is not on that list. The line is built for a bowl-and-carton pack, so a pouch, a can or a dried mix means different equipment. That is a scoping question and it is routed to a separate project through the same ODM and OEM development route rather than answered with a refusal.

The useful first message is short and specific: what the product is, who buys it, which market it sells into, and what has to be true for the listing to be approved. Everything else on this page follows from those four answers.

Ask for a specification sheet and a sample before anyone discusses price. The specification sheet tells you whether the supplier has written the product down, and the sample tells you whether they can make it. If either one takes an unreasonable amount of time to arrive, that is information about the project rather than about the post.

Once a specification is locked, the questions change from creative to operational: which sizes are in the first run, which market file is being assembled, and when the artwork can be signed. Those three answers decide the timeline, and none of them is decided by the plant alone.

At an ingredient whose identity cannot be documented, and at a concept that only works because of a disease claim. Where a recipe needs a format this line does not currently run, that goes to a separate project rather than being turned down. The line is a bowl-and-carton line, and the company is not limited to it.

The channel the product sells through, the destination market in priority order, the target price point, the net content you want, and your launch window. A written recipe helps but is not required for a first scope; an ODM project starts from the library instead. With those five answers we can write a scope and a sampling plan.

Piece size is set in the filling specification with a maximum, the fill is dosed to a weight target and check-weighed, and the packed density is checked on the filled bowl before sealing. Density matters because the retort profile is written against the cold point of a bowl with a known packing, so a density change is a process change.

Statements of content, preparation, storage, net content, solid content and the traditional food use of the ingredients. Not statements that the product treats, prevents or relieves a condition, and no disease names. The destination market decides which of the permitted statements survive, so the market file is assembled per market rather than copied.

The retort profile validated for that bowl and fill, the cold point the profile was set against, the parameters of the cycle as run, and the acceptance criteria for a production chart. It is the document a batch chart is compared against at release, and it is why a change to the bowl, the fill weight or the piece size triggers a re-validation.