


Learn how to specify cleaning, stone removal, optical sorting and metal detection for bulk coix seed, and which records to request before approving a supplier.
For a bulk buyer, “cleaned coix seed” is not a specification. It is a claim that needs a definition. Coix seed (Yi Yi Ren, Job’s Tears, Coicis Semen) is a small, starchy kernel that moves through harvesting, drying, hulling, polishing and storage before it ships. Each stage can add stones, shell fragments, broken or moldy kernels, dust, insects or metal. A lot that looks acceptable in a sample photo can still cause a QA hold, a customer complaint or a failed mycotoxin test once it reaches your plant. This article explains what cleaning and sorting should cover, which defects to name in a purchase specification, how to verify a supplier’s controls, and what to prepare before you request a quotation.
Cleaning removes material that is not coix kernel: soil, sand, stones, dust, straw, weed seeds, shell fragments, hair, string and metal. Sorting separates kernels that are coix but do not meet the specification: broken, shriveled, discolored, insect-damaged, moldy or off-size pieces. Grading then assigns the remaining product to a size, color and polish category. Buyers who treat these as one step usually discover the gap at receiving inspection.
The WHO guidelines on good herbal processing practices for herbal medicines describe sorting (garbling) as the first step in ensuring purity and cleanliness. They name dirt such as soil, dust, mud and stones, and impurities such as insects, rotten tissue and extraneous plant parts, as material to be removed. That framing helps with coix seed because it separates the two questions a specification must answer: what is not coix, and what is coix but unacceptable.

A specification should list each defect category separately, with a way to measure it. The table below is a qualitative planning tool, not a published standard. Numerical limits must come from your destination market, your pharmacopoeia or your own approved sample.
| Defect category | Typical examples in coix seed | Common removal approach | How to verify |
|---|---|---|---|
| Stones, sand, soil | Field stones, grit, clumped soil | Screening, aspiration, density (destoner or gravity) separation | Sample sieving and weighing by category; ash testing if your spec includes it |
| Light foreign matter | Dust, chaff, straw, hull fragments | Aspiration, sieving | Visual inspection of a weighed sample; dust check after shaking |
| Other seeds and plant parts | Weed seeds, unhulled grains, stalk pieces | Sieving, gravity table, optical sorting | Hand-sort of a retained sample |
| Defective kernels | Broken, shriveled, discolored, insect-damaged, moldy kernels | Optical sorting, gravity separation, final manual check | Defect count on a weighed sample against the golden sample |
| Metal | Wire from screens, fastener fragments, equipment wear | Magnets, metal detector | Detector challenge logs and rejected-material records |
| Size and polish variation | Mixed kernel size, uneven whiteness | Size grading, controlled polishing | Size distribution and color comparison to the golden sample |
Two practical points follow from this table. First, each category has a different removal method, so a supplier that mentions only “color sorting” has answered one row. Second, a defect that you do not name in the contract cannot be rejected cleanly later.
Keep these as two limits. Foreign matter (anything that is not coix kernel) is typically a pass/fail safety and cleanliness issue. Defective kernels are often a commercial grade issue, where the tolerance depends on whether you sell cereal, tea, powder or decoction material. Combining them into one “impurities” percentage hides which problem the lot actually has.
The sequence below is a general pattern used in grain and seed processing. Treat it as a list of questions, not a description of any one factory. Suppliers differ, so ask for the actual flow chart and the equipment at each stage.
Optical sorting improves appearance and can remove visibly damaged kernels, but it does not certify chemical or toxin safety. The distinction matters because coix seed has a documented mycotoxin profile.
A 2025 study, Mycotoxins Co-Exposure Risk Assessment in Coix Seed, analyzed 50 batches collected from five major production regions in China. It reported mycotoxins in 94% of samples, with zearalenone the most frequent at 82%. It also found two or more mycotoxins together in 74% of samples. The sample is limited to 50 batches, so it does not describe every lot on the market. It does show that visual quality and toxin status are separate questions, and that a test for aflatoxin alone does not cover the whole risk.
The same reasoning applies to physical hazards. The FDA’s compliance policy guide on hard or sharp foreign objects in food records that objects smaller than 7 mm in maximum dimension rarely cause trauma or serious injury except for special-risk groups. That is a U.S. enforcement reference, not a purchasing standard. A buyer supplying infant, elderly or institutional markets, or one facing strict retailer audits, may reasonably set tighter limits than a regulator would act on.
For residues, aflatoxin triggers and COA content, the site’s fruit and seed herb checklist for residues and aflatoxin risk is a practical companion to this article. It stresses that residue panels should match the selling region and that a COA should show numerical results, not only a pass statement.

No. A brighter, more uniform kernel is easier to sell visually. But the tighter you sort and polish, the more material you reject, and the more the product itself changes.
The defensible position is that the right cleaning and sorting level is the one that meets your end-use specification at an acceptable accept rate. It is not the whitest or cleanest lot a supplier can produce. A tea blender, a powder manufacturer and a decoction pharmacy each tolerate different things.
Food and herbal channels can use the same crop but often accept different tolerances and need different documents. The site’s comparison of food-grade and herbal-grade coix seed makes the point that these labels are not universally harmonized grades. They mean something only when tied to a named specification, test method and batch COA.
| End use | What the cleaning spec usually emphasizes | What to confirm with the supplier |
|---|---|---|
| Retail cereal or soup packs | Visible uniformity, low broken-kernel rate, no foreign matter, clean odor | Approved golden sample, defect count method, packaging protection |
| Beverage or powder base | Low dust, consistent size, no metal or stone, stable color | Whether sorting happens before milling, metal detection point |
| Herbal tea or decoction material | Identity, plant part, absence of extraneous parts, controlled processing | Identity records, processing status, pharmacopoeial reference if claimed |
| Extraction material | Low foreign matter and mold, predictable moisture | Moisture result, storage history, contaminant panel |
Do not assume one cleaning level fits all four. State the use and the destination market in the RFQ.
Write the specification before you compare quotes. At minimum, include the fields below and agree on the test method for each.
| Specification field | What to state | Why it affects the order |
|---|---|---|
| Product identity and form | Coix seed / Yi Yi Ren / Coicis Semen, hulled or polished, polish level | Defines what counts as foreign matter versus kernel |
| Intended use and market | Food, herbal or dual use; destination country | Sets which limits and documents apply |
| Foreign-matter limit | Maximum percentage by weight, with categories (stones, plant matter, metal) | Enables rejection if exceeded |
| Defective-kernel limit | Broken, moldy, insect-damaged and discolored tolerances | Separates grade from safety |
| Size and color range | Reference to an approved retained sample | Controls batch-to-batch consistency |
| Metal control | Detector or magnet use, challenge records, action on rejects | Documents prevention of a high-severity defect |
| Moisture | Contract limit and test method | Affects mold risk and storage stability |
| Contaminant testing | Pesticide, heavy metal, mycotoxin and microbiological panels with methods and limits | Visual sorting does not cover these |
| Documents | Batch COA, lot traceability, process flow, sorting and metal-detection records | Allows audit and dispute resolution |
| Failure terms | Which sample and laboratory control, who pays for retesting, replacement or credit | Prevents arguments after arrival |
Set the numerical foreign-matter and defect limits yourself, based on your destination-market rules, pharmacopoeial reference or internal acceptance standard. Do not copy a number from an unrelated herb or an unverified supplier brochure.

Send a written package, not a single line asking for the “best price”. Include the application, destination market, order volume and delivery schedule, form and polish level, foreign-matter and defect limits, size and color reference, contaminant panel, packaging format and the documents you require. Ask the supplier to describe its cleaning and sorting flow, provide a pre-shipment sample, and show a recent batch COA with numerical results and a metal-detection record.
The supplier’s bulk coix seed (Yi Yi Ren) product page states that free samples, COA reports and technical support are available, and that the products are made in a GMP-certified facility under ISO 22000 standards. Treat those statements as starting points. Request the certificate scope and the batch-specific records that show how your lot was cleaned and sorted, and confirm the details in writing before approving production.
Cleaning removes non-coix material such as stones, soil, dust, plant fragments and metal. Sorting removes coix kernels that fail the specification, such as broken, moldy, discolored or insect-damaged pieces. A specification should set separate limits and verification methods for each.
No single limit applies to every order. Acceptable levels depend on end use, destination market and any pharmacopoeial or customer standard. Define the limit by weight and category in your contract, and agree on the test method before production.
No. Optical sorting removes visibly defective kernels but cannot confirm pesticide, heavy-metal or mycotoxin compliance. A 2025 study found mycotoxins in 94% of 50 coix seed batches, so batch-specific laboratory testing remains necessary.
Yes, in most B2B programs. Metal can enter from processing equipment, not only from the field. Ask where the detector or magnet sits in the line, how it is challenged, and what happens to rejected product.
Not necessarily. Polishing changes color, bran content and taste, and it can create dust and broken kernels. A polished kernel can still contain stones, mold or contaminants. Cleanliness depends on the cleaning, sorting and testing controls, not the polish level alone.
Request the process flow chart, batch COA with methods and numerical results, lot traceability, sorting and reject records, metal-detection logs, moisture results and a retained golden sample. For higher-risk orders, ask for an independent laboratory report.