7 Costly Mistakes to Avoid in Intermediate Specifications

Abstract: Every acceptance criterion on a specification is a promise a supplier makes about a batch. This buyer’s guide explains how specification limits are set, how to judge whether they are reasonable, and the seven costly mistakes buyers make when they accept a spec without question.

The Number Nobody Verifies

A buyer spends a week negotiating the unit price of a pharmaceutical intermediate, another week on lead time and MOQ, and then signs off on the technical package in an afternoon. The document that will be used to judge every future batch — the specification with its acceptance criteria — gets a glance, not a review.

That is backwards. The price is a number you pay once. The specification is the standard every batch you ever receive will be measured against. A loose spec lets marginal batches pass and costs you downstream. A spec so tight it cannot be met honestly invites the supplier to make the data pass instead. Either way, the money you save on the unit price is dwarfed by what a badly set specification costs you in rejected API batches, failed filings, or a supply relationship built on doctored COAs.

This guide is written from the supplier side of the same table. It walks through what a specification actually is, where acceptance limits come from, why both loose and overly strict specs are dangerous, and the seven mistakes buyers make when they treat the spec as boilerplate.

What a Specification Actually Is

A specification is the set of tests, analytical methods, and acceptance criteria that defines a conforming batch. It is the contract that sits behind every certificate of analysis: the COA reports what was measured on one batch, while the specification defines the limits those measurements must fall inside. No specification, no standard of conformity — the COA becomes a list of numbers with nothing to judge them against.

This is exactly how the ICH framework treats it. ICH Q6A — Specifications: Test Procedures and Acceptance Criteria for New Drug Substances and New Drug Products — defines a specification as the quality standard a material must comply with, and it is the closest thing the industry has to a shared language for how those standards should be built.

The framework matters to intermediate buyers for one practical reason: the specification your supplier sends you is not an arbitrary list. It should be a data-derived document, and once you know how it should be built, you can tell whether the one in front of you was built at all or just assembled.

Two common confusions are worth clearing up before going further. First, a specification is not the same as a quality standard in the general sense — the general standard describes what good practice looks like; the specification describes what this product must be.

Second, a specification for an intermediate is usually a commercial document agreed between buyer and seller, not a pharmacopoeial monograph. Pharmacopoeias set minimum expectations for finished drugs; intermediate specs are negotiated, which means you have a seat at the table — if you use it.

Where Acceptance Limits Come From

Acceptance criteria are not invented, and they are not copied from a competitor’s COA. Under ICH Q6A, every limit should be justified by data from development, stability studies, and manufacturing variability. For an intermediate supplier, that translates into three concrete inputs you can verify:

The statistical logic behind the limits is more accessible than it sounds. For an impurity, Q6A’s decision-tree approach starts from the mean of measured impurity levels on real batches, adds a statistical allowance for variability (in practice, an upper confidence bound of roughly three standard deviations above the mean), then compares that figure with the growth predicted by stability data over the retest period.

The acceptance limit is set at the more protective of the two — but critically, not so tight that it tightly wraps the batch data itself. Q6A explicitly warns against establishing acceptance criteria that merely encompass the batch data at the time of filing, because early data is rarely enough to judge process consistency.

For the buyer, this has one practical implication: the supplier’s ability to show you the derivation matters more than the numbers themselves. A spec with limits that can be traced to batch and stability data is a document you can trust. A spec whose limits match a pharmacopoeia’s or a competitor’s with no derivation is a number with no anchor.

Universal Tests vs Specific Tests: What Must Be On the Spec

ICH Q6A separates tests into two groups, and the distinction is a ready-made checklist for reviewing an intermediate spec.

Universal tests belong on every specification:

• Description — physical state and color, with acceptable limits stated as a description, not a color photograph.
• Identification — a method specific enough to discriminate the compound from closely related structures. A single HPLC retention time is not, by itself, a defensible identity test.
• Assay — quantitative content against a reference standard, reported as a range (for example 98.0–102.0%).
• Impurities — organic impurities, inorganic impurities, and residual solvents, each with individual and total limits.

Specific tests are added when the molecule’s chemistry demands them. For intermediates, the four that matter most in practice:

• Particle size, when downstream formulation or dissolution depends on it — Q6A Decision Tree #3.
• Polymorphic form, when more than one crystalline form exists and the forms behave differently — Q6A Decision Tree #4. The risks here are covered in detail in our guide to polymorphism and solid-form control.
• Chiral purity, when the molecule has chiral centers — a standard achiral HPLC run can hide 50% of the wrong enantiomer. Q6A Decision Tree #5 and our chiral purity guide cover the specifics.
• Water content, when the material is hygroscopic, forms hydrates, or is moisture-sensitive in storage.

Two more Q6A concepts are directly useful to buyers. Release vs shelf-life criteria: many suppliers release with tighter limits than they guarantee at the retest date, because impurities grow in storage.

If the COA you receive shows release limits, ask what the limits are at the retest date — the answer tells you how much shelf life you are really buying. And skip testing: some tests are not run on every batch (a Q6A concept applied after justification). If a COA shows “periodic testing” instead of a result, confirm what is tested every batch versus what is tested occasionally — a gap you should know about before you rely on it.

Every spec will not carry every specific test — that is the point. But the buyer’s rule is simple: if a test is missing and the material’s chemistry says it should be there, that absence is a red flag, not a convenience.

The Loose Spec Trap: What a Wide Limit Really Costs

The most common way buyers lose money on specifications is not paying for a bad batch — it is accepting a spec so wide that the batch cannot fail, and discovering the consequences two steps downstream.

An intermediate impurity that passes a loose limit does not disappear. It travels into the next synthesis step, and if it reacts or carries through, it accumulates. The result is an API batch that fails its own tighter limits, or a finished drug with an impurity profile that a regulator will not accept. This is the mechanism behind the quiet cost of loose specs: the money is not lost at the intermediate purchase; it is lost when the API batch is rejected or the ANDA submission comes back with an impurity deficiency.

Loose specs are also a cultural signal. A specification is the supplier’s statement of what it commits to prove about every batch. A spec widened to the point where nothing can fail says the supplier would rather not have that conversation. Compare that with the behavior of suppliers who treat the spec as a living, data-driven document — and see our guide on impurity control in intermediates for what a serious impurity program looks like.

The Overly Tight Spec Trap: When Strict Becomes Dangerous

There is a mirror-image failure that buyers often fall into after being burned once: demanding limits tighter than any process can honestly deliver.

A specification that sits inside the process’s natural variability creates an impossible choice for the supplier. Either batches fail release regularly — which means your order is late, or the price rises to cover reworks — or the supplier finds a way to make batches pass that were never going to.

In real supply chains, the second option is far too common: re-integrating and re-reporting until a peak falls below the limit, testing sample from the best corner of the drum, or quietly repeating a test until the “right” result appears. These are exactly the data-integrity failures that produced the warning-letter patterns regulators have been chasing for a decade.

The uncomfortable truth for buyers: demanding a spec tighter than capability is a direct invitation to falsified COAs. The supplier who cannot hit 99.5% assay with their process will either tell you so honestly, or make the number appear. You will not know which until a downstream failure — or an audit — exposes it.

The right target is a specification that is tighter than what you need for downstream success but wider than process capability. The first part is about your product requirements; the second part is about keeping the supplier honest. Both are decisions you can make deliberately once you have the data in front of you.

Loose, Tight, or Right: A Comparison

Worked Example: An Impurity Limit That Travels

Numbers make the loose-spec argument concrete. Consider a process impurity in an intermediate that is controlled to 1.0% in the supplier’s specification. A batch measures 0.8% — comfortably inside the limit, released, shipped.

Now trace what happens next. In the following reaction step, roughly half of this impurity carries through unchanged instead of being purged (the fate of an impurity through a synthesis — carry-through, reaction-out, or purge — is exactly what impurity fate mapping quantifies). With a simplified 1:1 mass mapping from intermediate to API, the impurity’s contribution at the API stage is about 0.8% × 0.5 = 0.4%. The API specification for this impurity is set at 0.2%, in line with typical single-impurity control for a new drug substance.

The result: a batch that passed the intermediate spec — 0.8% against a 1.0% limit — is guaranteed to produce an out-of-spec API. The intermediate was compliant; the order was delivered; the failure happens at your next step, at your cost, and usually weeks later. Now set the intermediate limit to 0.3% instead. The same 0.8% batch is rejected at the intermediate stage, where the fix is cheap and fast, rather than at the API stage, where it is expensive and slow.

Five Documents to Demand Before You Trust a Spec

A specification is only as good as the data behind it. These five documents turn a one-page list of limits into something you can actually verify.

1. Batch history data. The last 10–20 commercial batches, full results per parameter. You are checking that the limits sit outside the real distribution of results — and that the batch you are buying looks like the rest of the population. Batch-to-batch consistency is the behavior you are probing; see our guide to batch consistency in intermediate manufacturing.
2. Stability data covering the retest interval. Confirm that impurity growth between release and retest stays inside the stated limits — and understand what the limits are at the retest date, not just at release.
3. Method validation reports. Each test on the COA should trace to a validated method — accuracy, precision, specificity, and the detection/quantitation limits for impurities. A method that cannot measure below the limit is a method that cannot enforce the limit. Our ICH Q2 buyer’s guide explains how to read one.
4. Specification change history. Has the spec been revised? When, and why? A spec that changed after a process change is normal; a spec that changed after a customer complaint or a failed audit is information.
5. Capability context. Not always a formal document, but ask the question directly: “Can this process hit this limit every time, and how far inside the limit do normal batches actually land?” The answer, and the willingness to answer, tells you more than the number.

Seven Costly Mistakes Buyers Make

1. Never asking where the limits came from. Treating a specification as boilerplate is the root of most other mistakes. Every limit should have a data story; if the supplier cannot tell it, that is the answer.
2. Judging a spec by the purity number alone. A 99.5% assay with a missing impurity test tells you nothing about the impurities that matter. The test list is as important as the limits — see the universal vs specific test checklist above.
3. Accepting a loose spec to close the deal faster. A wide spec makes the negotiation easy and the downstream failure expensive. The supplier who offers it is pricing in their own convenience, not your risk.
4. Demanding a spec tighter than capability. The fastest way to manufacture fake COAs. If the supplier’s process cannot honestly hit your number, either widen the limit with a documented justification or walk away — do not create the incentive to fake it.
5. Ignoring the retest-date limits. Release limits and shelf-life limits are different. Buying on release limits while your storage time eats into the gap is buying quality you do not get.
6. Not checking that methods can measure what they enforce. An impurity limit below the method’s quantitation limit is unenforceable — the supplier literally cannot tell you whether the batch meets it. Check the LOQ against the limit.
7. Treating the spec as fixed forever. Specifications should change as data accumulates and processes improve. A spec frozen for years, on a product that has seen process changes, is a spec that has silently become fiction.

The 5-Minute Spec Review

Before you accept any intermediate specification, run this checklist. It takes five minutes and it costs nothing.

1. Is every universal test present? Description, identification, assay, impurities. Any missing item — why?
2. Are the chemistry-required specific tests there? Particle size, polymorph, chiral purity, water content where the molecule demands them.
3. Do the limits make sense against the chemistry? An assay limit wider than 98.0–102.0%, or an impurity total above ~1–2% without justification, deserves a question.
4. Can you see the data? Batch history, stability, and method validation. If they are not offered, ask.
5. What happens at the retest date? Compare release limits with shelf-life limits, and check the retest date against your own inventory plan.
6. Have the methods been validated for these levels? The LOQ of every impurity method should sit below the limit it enforces.

The suppliers who answer these questions fluently, with documents, are the ones who have built their specification the way Q6A intends — from data. The suppliers who deflect, or who cannot produce the batch history behind their own limits, are telling you what the relationship will look like when a batch fails.

Specification review is one stage of the broader qualification picture. For the full procurement workflow, start with our supplier audit checklist and our guide to qualifying a second supplier. When you are ready to review a specification on a specific intermediate, contact our team.