Compliance
Grades and buyer specifications
A specification is a promise you can test: moisture, water activity, defect counts, piece size, colour, and who measures what. This page shows the typical structure of a dried-food spec, the numbers each product family usually carries, how sampling works, and how to write one a buyer will accept without a fight.
Anatomy of a specification sheet
Commercial specs — whether from the Codex, a supermarket, or your own sales sheet — all follow the same skeleton:
| Section | Contents | Who tests |
|---|---|---|
| Product identity | Name, variety, style, process (sun/hot-air/freeze), origin | You, on the label |
| Physical | Piece size/diameter, thickness, whole vs pieces %, foreign material | You, at packing |
| Chemical / compositional | Moisture, water activity, added sugar or salt, sulfite residual | You + lab |
| Microbiological | Pathogens absent, counts within limits | Accredited lab |
| Sensory / colour | Colour reference, odour, taste — often as graded descriptions | You + buyer on arrival |
| Packaging & marking | Pack type, net weight, lot coding, pallet/label requirements | You |
| Shelf life & storage | Best-before, storage conditions the claim rests on | You, with evidence |
Typical spec numbers by product family
Indicative industry ranges — negotiate your own, but start from here. "aw" targets assume shelf-stable, no added hurdles.
| Product | Moisture | aw target | SO₂ residual | Typical physical limits |
|---|---|---|---|---|
| Dried apple rings | 16–20 % | ≤ 0.65 | ≤ 10 ppm (or declared) | Ø 55–75 mm; ≤ 5 % broken; no peel-on pieces (if peeled spec) |
| Dried apricots (sulfured) | 22–25 % | ≤ 0.70 | 500–2000 ppm (declared) | Whole, uniform size count (e.g. 140–160/kg); no pits |
| Prunes | 24–30 % | ≤ 0.70 | usually nil | Size count per spec; ≤ 3 % pits/pit fragments |
| Sun-dried tomatoes | 8–12 % | ≤ 0.60 | 0–500 ppm | Halves; ≤ 2 % foreign matter; no black mould spots |
| Dried mushrooms | ≤ 12 % | ≤ 0.65 | nil | Cap size grade; ≤ 1 % vermin-damaged; free of mould |
| Jerky (US-style) | MPR ≤ 0.75:1 | ≤ 0.85 w/ cure | nil | Strip length/width; ≤ 10 % length variation; fibrous, not brittle |
| Herbs (whole leaf) | ≤ 8 % | ≤ 0.55 | nil | ≤ 2 % stem/foreign matter; colour per type reference |
| Fruit powder | 3–5 % | ≤ 0.35 | per recipe | ≥ 95 % through 500 µm; free-flowing, no caking |
| Dried beans (re-packed) | ≤ 16 % | ≤ 0.65 | nil | ≤ 2 % splits, ≤ 1 % damaged, ≤ 0.5 % foreign seeds |
Defect tolerances — how to count them
Buyers sample a defined quantity and count defects against tolerances. Build the same habit into your batch record.
| Defect class | Examples | Why it matters |
|---|---|---|
| Critical (zero tolerance) | Live insects, mould, glass/metal, rodent evidence | Safety — automatic rejection, possible recall |
| Major | Rancid odour, fermentation smell, insect damage ≤ threshold, decayed pieces, unapproved additives | Buyer rejection at their discretion; quality failure |
| Minor | Colour variation, broken pieces, small blemishes, slight size deviation | Counted against the tolerance; drives grade discounts |
Working a defect count (dried apple rings example)
- Sample Take a random 500 g from the batch after mixing (not the top of the tray).
- Count total pieces (e.g. 118 pieces).
- Sort and tally each defect class: broken (9), blemished > 5 mm spot (4), peel-on when spec says peeled (2), off-colour (3).
- Compute percentages broken 7.6 %, blemished 3.4 %, peel 1.7 %, off-colour 2.5 %.
- Compare to spec — spec allows ≤ 5 % broken: FAIL. Decision: re-grade the batch (sort out breaks) or divert to a "pieces/bakery grade" customer, not to the same spec.
Sampling plans without statistics software
- Containers in the lot: sample √N containers (N = number of packs), randomly selected — 100 packs → sample about 10.
- Sample size per container: 50–100 g for pieces; the whole sachet for retail-format checks.
- Mix into a composite for the lab: moisture, aw and micro results are only meaningful on a well-mixed composite.
- Never sample only the top of a pallet — defects and moisture migrate; dig into the stack.
- Keep retain samples from every lot (one per lot, sealed, labelled, stored cool) — they are your defence in any dispute or recall investigation.
Writing a spec sheet a buyer accepts
- Start from your evidence — set limits at what your process demonstrably achieves plus margin, not at Codex maxima you cannot hit.
- Define every measurement method — "moisture by oven LOD 105 °C", "aw by chilled-mirror meter"; buyer and seller must use the same method or disputes follow.
- State the sampling plan — who samples, how many, composite or not.
- Agree the action on failure — reject, re-grade, discount, or replacement; written before the first shipment, not argued after.
- Attach the certificate of analysis format — see the COA template.
- Version and date the spec — a spec that changes silently is the fastest way to lose a customer's trust.
🤝
The honest-margin trick
Set internal targets tighter than the spec (e.g. spec ≤ 5 % broken, target ≤ 3 %). Shipping at the limit is how one bad tray turns into a rejected pallet; shipping inside a margin is how you become the reliable supplier.
Grade names and discounts
| Grade | Meaning | Typical destination |
|---|---|---|
| Extra / Grade A / Fancy | Uniform, whole, blemish-free, true colour | Retail packs, premium private label |
| Standard / Grade B | Minor defects within tolerance, some variation | Own-label retail, foodservice |
| Pieces / Bakery grade | Broken, mixed sizes — same food safety, worse looks | Bakery, cereal, ingredient blenders |
| Powder / Processing | Milled from any sound material | Powder buyers, extracts, pet food (with spec) |
Grade hierarchy is also your quality-release valve: a batch that fails retail spec can often pass bakery spec honestly. What it must never do is silently move down a grade — the buyer's spec is a contract.