The most precise analysis cannot correct poor sampling
A laboratory result can be technically excellent and still fail to represent a delivery correctly.
For silica sand, the control chain therefore does not start with XRF, XRD, sieving or moisture measurement. It starts when we decide which lot must be represented, where to sample, how much material to collect and how to reduce that material to the portion actually analysed.
ISO 14488:2007 specifically addresses obtaining a representative test sample from a defined bulk of particulate material, with particular relevance to properties such as particle size and particle-size distribution. ASTM D75/D75M likewise states that sampling is as important as testing itself.
For an industrial buyer, the practical issue is straightforward:
a certificate of analysis can only be interpreted properly when the tested sample can be linked to an identified lot and a defensible sampling method.
This logic complements our general industrial silica sand guide and our report on XRF and XRD analysis of silica sand.
1. Define the lot before discussing the sample
A sample can only be “representative” in relation to something that has been defined.
Depending on the contract and industrial organisation, a lot may correspond to:
- production over a defined period;
- an identified silo or cell;
- a homogenised stockpile;
- a truck, container or series of bulk bags;
- a quantity produced under the same conditions;
- a delivery combining several logistical units.
The essential point is to establish this definition before interpreting the result.
If an analysis is based on material taken from a single bulk bag, it should not automatically be presented as proof of conformity for several hundred tonnes produced at different times.
Link the lot to the decision
The plan should also define the decision it is intended to support:
| Objective | Question to answer |
|---|---|
| Production control | Is the process remaining stable? |
| Lot release | Does the lot meet the specification? |
| Customer receipt | Does the delivered material match the specification? |
| Investigation | Is a deviation or contamination localised? |
| Supplier qualification | Is the material sufficiently consistent across several lots? |
The control frequency and level do not necessarily need to be identical for these five situations.
2. Why silica sand can be heterogeneous within one lot
Silica sand is a particulate material. Its constituents do not necessarily remain perfectly uniformly distributed during handling.
Segregation may occur during:
- discharge from a conveyor;
- silo filling;
- stockpile formation;
- truck loading;
- filling or emptying a bulk bag;
- vibration during transport;
- reclaiming with a screw, belt or loader.
Coarse grains, fines, particles of different density or certain impurities may therefore not occupy exactly the same zones.
The theory of particulate-material sampling specifically distinguishes errors related to selection from those introduced during preparation. More recent work on grouping and segregation error likewise confirms that this heterogeneity must be considered in a sampling plan.
This is why taking a convenient handful from the surface of a pile is not a representativeness method.
3. Increments and composite samples
An industrial plan often relies on several increments.
Each increment covers a small part of the lot or material stream. Combining them can create a composite sample intended to cover lot variability more effectively.
The purpose is not simply to increase the number of grabs. They must be distributed according to a defined logic: time, mass, packaging units, sections of a stream or accessible positions according to the selected method.
A moving stream can be useful
Where a material stream can be crossed correctly, sampling may provide access to several fractions of the material rather than only to an already segregated surface.
This does not mean that conveyor sampling is automatically compliant. Equipment, frequency, cutter opening, speed and procedure must still suit the material and the applicable method.
Stockpile, silo, truck and bulk bag: do not improvise
These configurations have different constraints.
With a stockpile, access to the interior is limited and segregation may be spatial. In a silo, material available at discharge may change over time. With bulk bags, one container may not represent the full population. In a truck, the top layer is not necessarily equivalent to the rest of the load.
The specification should therefore state where and when sampling is permitted.
4. How many increments should be collected?
It would be tempting to give a simple rule: five samples per truck, ten per lot or one kilogram per tonne.
A universal rule of this kind would be misleading.
The number of increments and required mass depend in particular on:
- the contractual method or standard;
- lot mass;
- maximum particle size;
- expected heterogeneity;
- the property being measured;
- possible rare constituents;
- the risk associated with a wrong decision;
- the minimum mass required by the final test.
ISO 14488 addresses representativeness associated with a defined confidence level. In contractual practice, this means the plan should be sized for the material and decision rather than copied mechanically from another product.
Rare constituents require particular care
ASTM C702 highlights an important issue during reduction: when only a few coarse particles or contaminant fragments are present, their random inclusion or exclusion from a small portion can strongly influence the result.
For contamination, oversize-particle or very minor-fraction control, a small portion may therefore be inadequate even when it is suitable for another test.
5. The gross sample is only half the job
After collection, there is often much more material than the laboratory will actually use.
The gross sample must therefore be converted into one or more test portions.
This is sample reduction.
ASTM C702/C702M-24 covers three methods for reducing aggregate samples and states that the aim is to minimise variations between selected test portions and the large sample. It also warns that poor execution can produce a non-representative portion.
For silica sand, the key principle is simple:
reducing mass must not mean selecting the easiest grains to reach.
6. Splitters, quartering and controlled reduction
Mechanical splitter
A suitable splitter divides a material stream into fractions using repeatable geometry. When correctly sized and operated, it removes much of the manual selection involved in reduction.
The material still has to be fed correctly. A splitter cannot correct an already biased gross sample and can itself be misused.
Quartering
Quartering involves homogenising according to the selected procedure, forming the material, dividing it into quarters and retaining opposite fractions before repeating the operation where necessary.
The term “quartering” should not be used for taking arbitrary shovel loads from different parts of a pile.
Select the method for the condition of the sand
Moisture, cohesion, fines content and maximum particle size can affect which reduction method is suitable.
The method should therefore be stated in the protocol or applicable standard, particularly where results are used to accept or reject a lot.
7. Errors that can appear after sampling
Even a good sample can be degraded before analysis.
Loss of fines
Handling that generates dust can preferentially remove the finest particles. The measured particle-size distribution then no longer represents the original sample.
Moisture change
An open container, long delay or non-sealed packaging can change moisture before measurement.
Cross-contamination
Poorly cleaned shovels, splitters, containers or surfaces can transfer particles from another product or previous lot.
Attrition or breakage
Aggressive handling can fragment some grains and alter particle-size distribution.
Identification error
An excellent analysis with the wrong lot number is still unusable for a contractual decision.
The control system must therefore preserve both physical integrity and documentary identity.
8. What a sampling record should contain
To make a result usable, traceability may include:
- supplier and product;
- lot reference;
- date and time;
- quantity represented;
- packaging or transport mode;
- sampling point;
- method used;
- number of increments;
- approximate gross-sample mass;
- reduction method;
- sampler identity;
- container and seal where applicable;
- analyses requested;
- observed anomalies;
- retained-sample identification.
This record links the field operation, laboratory, certificate of analysis and acceptance decision.
9. Laboratory sample, counter-sample and retained sample
For sensitive lots, several fractions may be prepared from the same controlled reduction:
- one fraction sent to the laboratory;
- one fraction retained as a reference;
- where appropriate, one fraction reserved for counter-analysis.
The contract should define retention time, storage conditions, sealing and the procedure for disputed results.
The aim is not to multiply sample bags unnecessarily. It is to maintain a coherent chain of evidence where lot conformity has an important industrial or financial consequence.
10. Sampling for particle size, chemistry and moisture
The same initial sample may sometimes support several tests, but the precautions are not identical.
Particle-size distribution
Representation of the different grain classes is central. Loss of fines or preferential selection of coarse grains directly distorts the curve.
For a practical application of particle-size parameters, see our report on silica sand for water filtration.
Chemistry
For SiO₂, Fe₂O₃, Al₂O₃ and other constituents, laboratory preparation may include additional drying, grinding or homogenisation steps depending on the analytical method.
Our article on XRF and XRD analysis explains why bulk chemistry and mineral phases answer different questions.
Moisture
Moisture is particularly sensitive to delay and packaging. A sample intended for this measurement should be protected from uncontrolled water exchange before testing.
11. Lot control for high-purity material
For high-purity silica sand, impurity concentrations may become low enough for contamination introduced during sampling or preparation to matter more.
The protocol should then consider:
- tool cleanliness;
- container compatibility;
- cross-contamination;
- representativeness of sampled zones;
- preparation before analysis;
- blanks or quality controls where required by the analytical method.
A low analytical detection limit never compensates for a contaminated sample.
12. Building a supplier–customer control plan
A robust specification can separate six elements.
| Element | Define |
|---|---|
| Lot | Quantity or period represented |
| Sampling | Point, time and responsibility |
| Increments | Number and distribution according to the method |
| Reduction | Procedure used before the laboratory |
| Tests | Parameters, methods and tolerances |
| Decision | Acceptance, counter-analysis and deviation management |
This structure helps avoid a common situation: supplier and customer each hold a different result but cannot determine whether the difference comes from the material, sampling, reduction or analysis.
13. Read a certificate of analysis with the right question
When a certificate reports SiO₂, Fe₂O₃, moisture or particle-size distribution, the buyer should be able to ask:
- which lot does this certificate represent?
- when was the sample collected?
- where was it collected?
- was it a single or composite sample?
- how was the sample reduced?
- which portion was sent to the laboratory?
- is a retained sample available?
- which test method was used?
These questions do not challenge the laboratory. They simply define the real scope of the result.
14. The most common mistakes to avoid
Sampling only where material is accessible
Accessibility is not a representativeness criterion.
Mixing sub-lots without preserving identity
An average can hide a localised deviation.
Reducing with a shovel and no procedure
The gross sample may be correct while the final portion is biased.
Allowing fines to escape
Particle-size distribution and sometimes chemistry can change.
Copying an increment count from another contract
The plan must match the product, lot, test and risk.
Confusing analytical repeatability with representativeness
Two very similar analyses on the same small portion mainly demonstrate repeatability for that portion.
15. Purchasing specification: information to request
For an industrial silica sand purchase, we recommend documenting at least:
Product
- final application;
- required particle size or distribution;
- relevant chemical parameters;
- moisture and other criteria where needed.
Lot
- lot definition;
- quantity;
- packaging;
- traceability.
Sampling
- standard or procedure;
- place and time;
- responsible party;
- increments and composite formation;
- minimum mass according to the tests.
Reduction and laboratory
- reduction method;
- analytical preparation;
- test methods;
- agreed laboratory where necessary.
Decision
- tolerances;
- acceptance rule;
- counter-sample;
- dispute procedure.
Key takeaway
Sampling is not a formality that happens before the “real” analysis.
For silica sand, representativeness depends on an entire chain: lot definition, increments, possible composite formation, reduction, packaging, identification and preparation before testing.
ISO 14488 provides a general framework for particulate materials. ASTM D75/D75M emphasises the importance of sampling, while ASTM C702/C702M shows that reduction itself can preserve or degrade representativeness.
The practical consequence for the buyer is direct: a laboratory number should always be read together with its sampling plan and traceability.