Expert report review
The shipment was rejected on a sample: how does a reviewer check whether the opposing expert's sample speaks for the whole lot?
Nobody tests 10,000 sweatshirts or 1,709 microprocessors one by one. A buyer who rejects a shipment, and a seller who sues for the price, both rely on a sample and on an expert who says what the sample means for the rest. Federal courts have accepted an 8 percent sample selected without bias. A review of the other side's report starts with how the sample was drawn, tested and extended to the lot.

Why does a dispute over a whole shipment turn on a sample?
Because testing every unit costs more than the dispute or destroys the goods. UCC 2-515 gives either party the right, on reasonable notice, to inspect, test and sample the goods to establish the facts and preserve evidence.
UCC 2-515(a) provides that, in furtherance of the adjustment of any claim or dispute, either party on reasonable notification to the other and for the purpose of ascertaining the facts and preserving evidence has the right to inspect, test and sample the goods, including goods in the possession or control of the other party. Under 2-515(b) the parties may agree to a third party inspection or survey to determine the conformity or condition of the goods, and may agree that the findings shall be binding on them in any later litigation or adjustment.
A sample answers one question directly: how many defective units were in the units tested. The claim is about another number: how many defective units were in the shipment. Getting the second number out of the first is an inference, and its strength depends on how the sample was chosen, how large it was, what counted as a defect and how the test was run.
That step is where a review of the opposing report does most of its work. A report can describe each tested unit correctly and still overstate what the results mean for the rest of the lot. The reviewer reads the testing record first and the conclusion second, and checks whether the conclusion stays inside what the sample supports.
When have courts accepted a sample as proof about the whole lot?
In Integrated Circuits v. E.F. Johnson (E.D.N.Y. 1988) an 8 percent sample selected without bias was adequate. In Texpor v. Trust Company Bank (S.D.N.Y. 1989) random audits of more than 30 units per style were found representative.
Integrated Circuits Unlimited, Inc. v. E.F. Johnson Co., 691 F. Supp. 630, was decided by Judge Jack B. Weinstein of the Eastern District of New York on August 10, 1988, after a bench trial. E.F. Johnson, a maker of taxi radios, bought microprocessors from a distributor and rejected two orders after an independent laboratory tested samples. For order WO4746 the laboratory tested 130 of 1,709 units, about 8 percent, and 6 failed, 4.6 percent. The court wrote that while the sampling was not scientifically randomized, selection for testing was made without bias, and that it was reasonable for Johnson to conclude that the remaining untested devices were similarly defective. It held that Johnson tested an adequate sample before deciding to reject.
The same case shows how the test conditions decide the result. For order WO4750 the laboratory tested 115 units, and 113 failed at temperatures above commercial standards. At the agreed standard temperature only 3 failed, 2.6 percent. A report that counted 113 failures would have tested the parts against a specification the contract did not set.
Texpor Traders, Inc. v. Trust Company Bank, 720 F. Supp. 1100 (S.D.N.Y. September 12, 1989), concerned 10,570 sweatshirts in two shipments. The buyer's auditors pulled 841 at random. In the first shipment they sampled 50 shirts per color for six styles and 32 for each of three others; in the second, 80 shirts from each of four styles and 125 from the fifth. The court found that the audits constitute statistically valid random samples and thus are representative of the defects in the shipments as a whole. It noted that the auditors used military standards for appropriate audit sampling, that each style was audited separately and that each sample was greater than thirty. Defect rates in some audits reached 64 percent. The court held that the buyer, Oxford, rightfully rejected the goods and dismissed Texpor's breach of contract claim.
The opinion keeps the audits apart from a smaller examination. Fred Gerson examined 70 garments and found material defects in about 90 percent of them, and a witness testified that those 70 garments were not intended to constitute a representative sample. For representativeness the court relied on the original audits. A review that finds the opposing expert examined only what was convenient can point to the same distinction.
What do AQL sampling plans under ISO 2859-1 require?
ISO 2859-1 defines a system of acceptance sampling plans for inspection by attributes, indexed by acceptance quality limit, for lot-by-lot inspection. The 2026 edition replaced the 1999 edition, so the reviewer checks which edition and plan the contract named.
ISO 2859-1:2026 is titled Sampling procedures for inspection by attributes, Part 1: Sampling schemes indexed by acceptance quality limit (AQL) for lot-by-lot inspection. ISO published it in January 2026 as edition 3. Its page states that the standard defines a system of acceptance sampling plans for inspection by attributes, with single, double or multiple sampling schemes indexed by AQL, and describes the AQL as the maximum average percentage of defective items considered acceptable during random sampling. The 1999 edition is listed as withdrawn and replaced by the 2026 edition.
A supply contract may name a plan: the standard, the inspection level, the AQL for each class of defect and the accept and reject numbers that follow from the lot size. Where the contract names a plan, the first check is whether the opposing expert used it. An expert who tested more units than the plan requires has not broken it, but one who tested fewer, or applied a different AQL, has tested against a standard the parties did not agree on.
Where the contract names no plan, the standard is still a reference for what a sample of a given size can show. The reviewer does not substitute a plan for the contract. He or she states what sample size and acceptance number a recognized plan would give for that lot, and whether the opposing expert's sample is larger, smaller or of a different kind.
How was the sample chosen, and does it cover every part of the lot?
The reviewer checks who pulled the units, from which cartons, pallets, styles and production dates, and whether the selection was random or by convenience. A sample drawn from one carton speaks for that carton.
Random selection gives every unit in the lot the same chance of being tested. In practice the units are pulled from numbered cartons chosen by a random rule, and the record lists the carton numbers and the units taken from each. In Integrated Circuits the court accepted a selection that was not scientifically randomized but was made without bias. The burden of showing that falls on the party that drew the sample, and the record is where it is shown.
Lots are rarely uniform. A shipment can combine several styles, colors, production runs or factories. The Texpor auditors audited each style separately, and the court relied on that. A sample that mixes all styles in one count can hide a defect concentrated in one style, or spread it across styles that were sound.
The reviewer also checks who chose the units. A buyer's warehouse that pulled the visibly damaged cartons for the expert, or a seller that supplied retained samples from its own stock, produced a sample of a particular kind. The report should say so. If it does not, the reviewer writes that the source of the sample is not recorded and that the conclusion about the lot depends on it.
What counted as a defect, and against which specification was it tested?
The defect definition and the test conditions must come from the contract, the order or the referenced standard. Integrated Circuits shows the difference: 113 of 115 chips failed above commercial temperatures, 3 failed at the agreed one.
Attribute inspection counts units as conforming or not. The count means nothing until the report says what made a unit nonconforming. A seam that opens under a specified pull, a chip that fails a function test at a stated temperature, a color outside a stated tolerance: each is a defined attribute. A unit that looks wrong to the inspector is not a defined attribute.
A contract may classify defects as critical, major or minor and set a separate AQL for each class. A report that lumps minor cosmetic marks with functional failures in one percentage overstates the rate of failures that would justify rejection. The reviewer re-sorts the opposing expert's findings by class wherever the raw data allow.
The test conditions must match the specification. In Integrated Circuits, testing at temperatures above commercial standards produced 113 failures out of 115, and testing at the agreed standard produced 3. The reviewer compares each test condition in the report with the contract, the purchase order, the product data sheet and any standard they reference, and lists each difference.
Does the expert's conclusion go further than the sample supports?
A sample gives a defect rate in the units tested and a range for the lot. A report that states the lot's exact rate, or counts every untested unit as defective, goes past its data.
Federal Rule of Evidence 702 requires the proponent of an expert to show the court that it is more likely than not that the testimony is based on sufficient facts or data, that it is the product of reliable principles and methods, and that the opinion applies those methods reliably to the facts of the case. A conclusion about 10,000 units drawn from 50 is tested on each of those points.
A defect rate in a sample is an estimate of the rate in the lot, and an estimate has a range. The smaller the sample, the wider the range. A report that says 6 of 130 failed reports a fact. A report that says 4.6 percent of the shipment is defective presents an estimate as a fact. A report that treats all 1,709 units as defective because 6 failed goes beyond both.
The damages figure often rests on the same step. If the claim is priced on the number of defective units in the lot, the reviewer checks how that number was derived from the sample, whether a range was given and whether the calculation used the low, middle or high end of it.
What does the reviewer list in the written review?
Each point where the sampling, the defect definition, the test conditions or the extension to the lot departs from the contract, the referenced standard or the record, tied to a page of the opposing report.
The written review follows the opposing report page by page. Points that usually appear:
Each point is stated as a fact with its source: the page of the report, the clause of the contract, the section of the standard, the carton list. The review does not argue that the goods were sound. It lists where the opposing conclusion about the lot rests on data that are missing or that point the other way.
- The lot size, the number of units tested and how many came from each style, run or carton.
- The selection method and who pulled the units, with the carton numbers if recorded.
- The sampling plan named in the contract, and whether the expert applied it, including the edition of the standard.
- The defect definitions and classes, and whether minor and functional defects were counted together.
- The test conditions compared with the contract, the purchase order and the data sheet.
- How the defect rate in the sample was extended to the lot, and whether a range was stated.
- Whether the untested units were preserved and whether the other side was notified before testing.
How does the reviewing side get access to the goods and the test data?
Through UCC 2-515 on reasonable notice and, once suit is filed, through Rule 34, which lets a party inspect, test and sample tangible things and data in the other side's possession, custody or control.
Federal Rule of Civil Procedure 34(a)(1) lets a party request that the other side produce and permit it to inspect, copy, test or sample designated documents, electronically stored information and tangible things in its possession, custody or control. The response is due in writing within 30 days after service under Rule 34(b)(2)(A).
The request covers the untested units, the tested units if they still exist, the laboratory's raw data and photographs, the carton lists and the warehouse records of how the sample was pulled. With the remaining stock, the reviewing side can draw its own random sample under the plan the contract names and compare the result with the opposing expert's.
If the goods have been sold, destroyed or returned, the review works from the record alone and says so. The absence of the goods is itself a point: a conclusion about a lot that no one can now re-sample rests entirely on how the original sample was documented.
What does IAE do when a shipment was judged on a sample?
IAE prepares a review of the opposing expert's report or an expert report on the client's questions for warranty and supply disputes in US courts, in up to 5 business days.
The review page for goods and product defect reports describes the work for warranty and supply disputes: the review reads the other side's report against Rule 26(a)(2)(B) and Rule 702, lists the tests the expert did not run and the conclusions the data do not support, and checks the valuation under UCC 2-714(2).
The client sends the contract and purchase orders, the specification or data sheet, the inspection or laboratory reports, the carton and sampling records, photographs, and the other side's report, and describes what to check. The secretariat confirms receipt the same day. Within one business day IAE confirms the format, the cost, the deadline and the specialist from its registry, and the client approves the candidate and the CV before work starts. Cases with tight deadlines have a 48-hour rush.
Every specialist is a member of the public IAE registry, verified by Member ID or QR code, and confirms in the document that he or she has no interest in the outcome. The court weighs a review together with the other evidence, and neither IAE nor the expert guarantees the outcome. IAE is not a law firm and does not advise on case strategy.
Sources
- UCC 2-515, Preserving evidence of goods in dispute (LII, Cornell Law School)
- Integrated Circuits Unlimited, Inc. v. E.F. Johnson Co., 691 F. Supp. 630 (E.D.N.Y. 1988), Justia
- Texpor Traders, Inc. v. Trust Company Bank, 720 F. Supp. 1100 (S.D.N.Y. 1989), Justia
- ISO 2859-1:2026, Sampling procedures for inspection by attributes, Part 1 (ISO)
- ISO 2859-1:1999, withdrawn (ISO)
- Federal Rules of Evidence, Rule 702 (LII, Cornell Law School)
- Federal Rules of Civil Procedure, Rule 34 (LII, Cornell Law School)
Read also
All articles
Warranty denied for liquid damage: how does a reviewer check the finding that water caused the failure?
A red liquid indicator records one fact: water touched the indicator. A warranty denial needs a second fact, that the liquid caused the failure
Defect after the warranty expired: is there still a claim, and what must the expert prove about when it arose?
A defect that first shows up after the warranty period has ended is usually outside the express warranty, even if the cause was there from the start
USDA inspection certificate in a produce dispute: what does it prove, and how is it checked?
A certificate issued by a USDA inspector is prima facie evidence of the facts it states in every federal court proceeding under the PerishableExpert report review
Ready to discuss your case?
Send the materials: the secretariat will confirm the format, cost, deadline and the specialist within one business day.

