Lubinpla

CoA within spec but different from last batch: when it matters

Summary

A new Certificate of Analysis (CoA) can sit within spec and still differ from the last batch. The difference matters to a customer only when the changed property drives a step in the customer's process and the shift exceeds the test method's precision or trends toward a limit. A CoA certifies conformance, not sameness with the last lot.

  • A precision limit is often approximated as 2.8 times the matching standard deviation, at about 95 % confidence for the difference between two results (Eurachem Guide, 3rd ed., 2025).

  • ASTM D3244-24 uses site precision (R') instead of published reproducibility (R) when several results for one batch come from a single laboratory (ASTM D3244-24, 1.4).

  • ISO 4259-2:2026 was designed for petroleum products, which are normally homogeneous, and does not apply to non-homogeneous products (ISO 4259-2:2026, scope).

  • The Western Electric rules count 6 points in a row trending up or down as a non-random signal on a control chart (NIST/SEMATECH e-Handbook, 6.3.2).

This answer covers numeric CoA properties of homogeneous liquid products tested by methods that publish a precision statement. A change of supplier, manufacturing site or formulation is a requalification question, not a lot-to-lot comparison. Active pharmaceutical ingredients under GMP rules fall outside this answer.

When does a within-spec CoA change matter for a customer?

A within-spec CoA change matters to a customer only when the changed property feeds a step in that customer's process and the shift is either bigger than the test's own scatter or keeps moving toward a limit. Use the table below, property by property, to decide whether a lot-to-lot difference needs action before a reorder ships.

CoA property shift

Verdict

Why

Shift smaller than the method's precision limit

No action; record it

Two results on the same material differ by no more than this about 95 % of the time (Eurachem Guide, 2025)

Shift in a property the customer's process does not use

No customer action; supplier trends it

The customer's result does not depend on it

Shift beyond precision in a property the process is sensitive to, such as viscosity at a metering pump (illustrative)

Check with the customer before shipping

The process was tuned on earlier lots, not on the specification

Values trending toward the customer's own tighter limit

Check with the customer and trend it

6 lots in a row moving one way is a control chart signal (NIST/SEMATECH e-Handbook, 6.3.2)

Changed test method or specification revision on the CoA

Re-check before comparing

Results under different methods or limits are not directly comparable

Work down from the top and stop at the first row that fits. A difference inside the method's precision is noise; it needs only a record.

Most real problems land in the third row. A viscosity change well inside the supplier's range can still alter flow through a customer's volumetric metering pump (illustrative), because someone set the pump on earlier lots. From the third row down, the verdict turns on the customer's process, which no supplier specification describes. The person who owns the account has to bring that knowledge to the review.

How big a lot-to-lot CoA difference is real, given test reproducibility?

A lot-to-lot difference on a CoA is real only when it is larger than the precision limit that matches how the two results were produced. That limit is the difference that test scatter alone can produce between two results on the same material. Reproducibility is one kind: the spread between results on the same material from different laboratories. Each kind of spread has its own standard deviation, and a precision limit is often approximated as 2.8 times the matching standard deviation, covering the difference between two results at about 95 % confidence (Eurachem Guide, 3rd ed., 2025).

If the supplier's own laboratory tested both lots, the matching figure is a within-laboratory one. For that case, the Eurachem Guide computes the intermediate precision limit the same way, from the intermediate precision standard deviation (Eurachem Guide, 3rd ed., 2025). ASTM D3244-24 likewise uses site precision (R') instead of R when several results for one batch come from a single laboratory (ASTM D3244-24, 1.4).

Take the figure from the precision statement of the method named on the CoA. Where a method states precision as a function of the value, read it at the level of the result you are comparing.

For petroleum products, fuels and lubricants, ASTM D6300-26 and ISO 4259-1:2026 set how precision is derived from an interlaboratory study, and ISO 4259-2:2026 sets how it is applied to specification limits. ISO 4259-2:2026 does not apply to non-homogeneous products. Other chemicals take the precision statement of their own test method.

Which CoA properties is a customer's process most sensitive to?

The properties that matter most set a dose, a flow, a reaction or a finish at a specific step. Only the customer's process record shows which ones do that, so map each CoA line to the step that uses it before you judge any change.

In typical plants (illustrative), viscosity counts where a pump meters by volume or a coater lays down a set film. Density counts where a product is dosed by volume but formulated by mass, and active content where a recipe fixes the addition rate. Water content, acid number or pH matter at a moisture-sensitive or reactive step; color matters when the product shows in the finished part. These links differ from plant to plant and line to line.

Even a property that sits mid-range on every lot can matter if the customer tuned settings around recent lots rather than the specification. Ask which settings were adjusted at startup and which incoming checks the plant runs. Those checks tell you what the plant already watches and where a shift would show up first.

When should a CoA shift be trended and when should it be checked with the customer?

Trend a CoA shift when it sits inside the method's precision or touches a property the customer does not use. Check with the customer when it exceeds precision in a sensitive property or keeps moving toward the customer's limit.

For each sensitive property, keep a simple control chart that plots each lot's value in order. The NIST/SEMATECH e-Handbook lists "6 in a row trending up or down" among the Western Electric trend rules (NIST/SEMATECH e-Handbook, 6.3.2), and it warns that adding these rules increases false alarms. So treat a signal as a reason to ask, not as proof of a problem.

When you call, send the property, both lots' values, the test method with its precision, and the process step you think it touches. Then ask whether the plant has noticed any change. If the answer points to a process effect, the reorder may need the steps in "When a chemical product reorder needs requalification". Lubinpla's reorder review reads the CoA in the data room and the project's measured-condition history, then judges whether a changed value reaches the customer's operating conditions.

Frequently asked questions

What is the difference between in-spec and equivalent to the last lot?

In-spec means each CoA result falls inside the supplier's acceptance limits. Equivalent means the lot will behave in the customer's process the way the last one did. ICH Q7, the GMP guide for active pharmaceutical ingredients, says the certificate should list each test performed, the acceptance limits and the numerical results (ICH Q7, 11.42). A certificate of that kind supports in-spec, not equivalence.

Does a different test method on the new CoA change the comparison?

Yes. Compare two lots only when both certificates name the same test method and edition. A different method has its own precision and may read systematically higher or lower, so an apparent shift can come from the method rather than the product. Ask the supplier for a result by the earlier method on a retained sample.

How do a supplier and a customer settle a dispute over specification conformance?

ASTM D3244-24 gives two parties a way to combine their test results into an assigned test value and compare it with an acceptance limit agreed before testing (ASTM D3244-24, 1.1 and 4.3). It is designed for contracts covering petroleum products and lubricants, and it recommends a qualified statistician's guidance.

Related posts