Start with the unit of product, not the AQL table

The core of AQL sampling for bulk polyphosphate balls is a decision the tables do not make for you: what counts as one unit. Once you define the inspection unit and separate defects that occur at bag level from defects that occur at ball level, ISO 2859-1 or ANSI/ASQ Z1.4 will give you a sample size and an accept/reject number. Everything before that step — lot definition, defect classification, sample extraction, test methods — is where incoming inspection plans for bulk water-treatment media usually fail.

This article walks through a workable sequence for planning that inspection, using polyphosphate balls (sometimes sold as siliphos crystal, phosphate crystal, or polyphosphate media) as the example. Note that these names are used loosely in the market. Siliphos is a trade name associated with a specific glassy polyphosphate product, and other polyphosphate balls may differ in composition, dissolution behavior, and documentation, so specifications must come from the actual supplier’s data, not from the product name.

Why bulk media does not fit the standard model cleanly

Attribute sampling standards such as ISO 2859-1 and its US equivalent ANSI/ASQ Z1.4 were written around lots of discrete items: a lot of valves, a lot of moulded parts. A shipment of polyphosphate balls sits awkwardly in that framework for three reasons.

• The product exists at several levels at once: bags or supersacks, individual balls, and the material inside each ball.
• Some characteristics are lot-wide (declared net weight, labeling, packaging integrity) while others vary ball to ball (diameter, chipping, colour).
• Chemical behaviour — dissolution rate, phosphate release, pH effect — is not reliably established by counting defective pieces.

The solution is not to abandon AQL sampling. It is to apply it to clearly defined inspection units and to handle the remaining characteristics with separate verification steps.

Step 1 — Define the lot and the inspection unit

Write down three definitions before you open the standard.

1. Lot: normally one production batch, one shipment, or one purchase order line. Keep lots uniform in grade, size, and packaging. Mixing two production batches into one lot weakens traceability and can hide a batch-specific problem.
2. Inspection unit: the item a defect is counted against. For packaging and labeling defects, the unit is the bag or supersack. For appearance and dimensional defects, the unit is the individual ball.
3. Test portion: the quantity taken to the laboratory for a destructive or instrumental test. This is not an inspection unit and is not counted in Ac/Re decisions.

This distinction matters because a plan applied to bags will not control ball-level defects, and a plan applied to individual balls becomes impractical at large volumes. Many buyers run a two-stage plan: inspect packaging on the sampled bags, then draw a defined sub-sample of balls from within those bags for physical checks.

Step 2 — Classify defects before you choose AQLs

Defect classification drives everything downstream. Classify by consequence, not by how easy the defect is to see.

Agree this list with the supplier in the purchase contract. An AQL that is not referenced in the contract is difficult to enforce later.

Step 3 — Choose inspection level, AQL, and the plan

For most industrial incoming inspections, general inspection level II under normal, single sampling is the default starting point. Systematically reduce the AQL as the consequence of the defect increases. A common structure is 0 for critical, something in the 1.0–2.5 range for major, and 4.0–6.5 for minor — but the correct values depend on your process tolerance and must be agreed, not borrowed.

Read the values for your own situation directly from the current edition of the standard you are working to. Acceptance numbers shift with the AQL, the inspection level, and the sampling scheme, and arrow plans apply where the code letter does not provide the required sample size. Where a lot is one-off and supplier history is thin, ISO 2859-2, indexed by limiting quality for isolated lots, is often a better fit than a lot-by-lot AQL scheme. For measured values such as ball mass or moisture, variables plans under ISO 3951 evaluate the mean and standard deviation instead of counting defectives.

One clarification worth stating internally: an AQL is not a guaranteed defect rate. It is an acceptance rule with defined risks, so a lot can pass while still containing some defective units. Do not describe an AQL as a quality target in supplier documents.

Step 4 — Extract the sample so it represents the lot

Sampling error in bulk media is usually dominated by how and where you take increments, not by how many balls you count. Practical rules:

• Select bags or supersacks at random across the consignment — not only from the front of the container or the top of the stack.
• From each selected bag, take increments from the top, middle, and bottom of the fill. Material at the bottom is where fines, dust, and moisture-related caking concentrate.
• Combine increments into a composite, then split down to the test portion using a riffler or a quartering method. Avoid scooping a test portion from one spot.
• Never pour a composite through a screen or crush it before appearance checks — that creates the fines you are trying to measure.
• Seal test portions in moisture-proof containers, label them with lot, date, and sampler, and retain a sealed reference sample for the agreed retention period.
• Record the sampling location layout. A defensible result depends on a documented sample, not just a correct table.

If the material is blended or comes from several production runs, increase the number of increments rather than the number of balls per increment. Heterogeneity between bags is the problem you are trying to capture.

Step 5 — Sequence the checks from fast to slow

Order the inspection so that gross problems stop the process early.

1. Documentation: compare the certificate of analysis, batch number, grade, net weight, and safety data sheet against the purchase order.
2. Packaging: check bag integrity, liner condition, sealing, labeling, and moisture protection.
3. Visual and dimensional: colour uniformity, cracks, chips, deformation, fused clumps, visible foreign matter, ball diameter, sphericity, and unit mass.
4. Fines and moisture: sieve the agreed fraction to quantify fines; determine moisture by the agreed method. Both should be measured under identical conditions each time, because ambient humidity affects results.
5. Chemical and performance testing: confirm the grade and behaviour using methods agreed with the supplier, such as dissolution or phosphate-release testing under defined conditions.

Set every specification limit from real data — the supplier’s technical data sheet, a validated internal specification, or a preliminary qualification lot. Do not invent limits such as a maximum fines percentage or a dissolution time to fill an empty cell in a form.

What AQL sampling will not tell you

It helps to be explicit about the limits of the method, because incoming inspection plans are often asked to do too much.

• It does not verify chemical identity or performance. A lot can pass every visual and dimensional check and still dissolve at the wrong rate.
• It does not confirm regulatory suitability. A product described as food grade is not automatically suitable for drinking water. Suitability depends on the exact product, the application conditions, and the destination market, and it must be verified against the applicable requirements for that market.
• It does not remove the need for document control. A certificate of analysis is a document check, not a substitute for periodic confirmatory testing.
• It does not correct a vague specification. Sampling plans work only against measurable, agreed limits.

Keep the functions separate while you plan: polyphosphate media are used for scale inhibition and corrosion control film formation. They are not water softeners and do not remove hardness by ion exchange, and they should not be confused with filtration media or disinfectants.

Rejected lots, resampling, and switching rules

Decide the response before the first shipment arrives. A rejected lot should be segregated and quarantined, and the decision should go through a documented review rather than a second informal look. Resampling a failed lot to obtain a pass is not sound practice unless the standard you are using explicitly permits it under stated conditions.

If you inspect the same grade from the same supplier repeatedly, ISO 2859-1 switching rules move you between normal, tightened, and reduced inspection based on recent history. Applying them prevents a supplier with a good record from being burdened by the same sampling intensity forever, and it escalates scrutiny automatically after failures. Document the switch decisions so the change in plan is traceable.

Four questions that come up in almost every plan

How many balls should I inspect per bag?

There is no universal number. If the ball is the inspection unit and the lot contains 500,000 balls, level II normal single sampling gives n = 500 balls for the lot — which is then distributed across selected bags rather than all taken from one. If the bag is the inspection unit, the sample is counted in bags. Choose the unit that matches the defect you are trying to detect, and state it in the plan.

Should fines be an AQL characteristic or a measured value?

Usually a measured value. Fines percentage is a continuous quantity, so a variables approach based on a mean and an upper limit controls it more meaningfully than counting defective balls. Treat individual caked clumps or visibly dusty bags as attribute defects on top of that.

Can I use the same AQL for every supplier and every grade?

No. AQLs should reflect the consequence of failure and the supplier’s demonstrated capability. A new supplier, a new grade, or a change in production site is a reason to tighten, not to reuse the previous plan.

Does a passing sample mean the lot is defect-free?

No. It means the lot met the acceptance rule for the defined characteristics, at the defined AQL, under the stated risks. Sampling reduces risk; it does not eliminate it.

Planning an inspection for a specific order?

If you are putting together an incoming inspection plan for bulk polyphosphate balls, the useful information to share is: the application and equipment type (dispenser or cartridge), the grade and ball size you intend to buy, your lot definition in bags or supersacks, the defect list and specification limits, the destination market, and any drinking-water requirements that apply. With those details, ENVO CHEMICAL can discuss the technical data and typical documentation for the relevant grade so your acceptance criteria are built on the product you are actually ordering, not on assumptions carried over from another supplier.