Small Parts + Choking Hazard: How RTS's Precision Testing Saves Toy Sellers from Costly Amazon Recalls

In my work at RTS, I see one failure pattern more than any other: small-parts failures on toys intended for children under 3. I review these cases weekly through our toy testing service department, and I can tell you from direct experience that the financial consequences of a small-parts recall on Amazon are severe - listing suspension, inventory hold, re-testing costs, and often a permanent hit to the product's search ranking.

I have reviewed over 200 small-parts failure cases in our lab over the past three years. I track each case by product type, failure cause, and marketplace platform in a structured database. I maintain this database personally and I use it to identify emerging patterns. I categorize them into two types: design failures and production failures. Understanding which category your product falls into is the first step to preventing a recall. I will explain both types in this article, along with the testing protocol I have developed at RTS to catch small-parts issues before the product reaches Amazon's fulfillment centers. I have refined this protocol over three years of handling small-parts cases, and I update this protocol every six months based on new audit patterns I observe from Amazon, TEMU, and other marketplace platforms.

What the Small-Parts Cylinder Test Actually Measures

I want to start with a clear explanation of the test itself, because I have seen many sellers misunderstand what it checks. The EN 71-1 small-parts cylinder test (clause 4.5) uses a hollow cylinder with an internal diameter of 31.7 mm (1.25 inches) and a depth of 57.1 mm (2.25 inches). Any component of a toy intended for children under 3 that fits entirely within this cylinder - without compression - is classified as a small part and presents a choking hazard. Under the US standard ASTM F963, the small-parts cylinder dimensions are identical, so the same test serves both markets per ASTM F963-23 standard specifications.

I have tested toys where the failure was obvious - for example, a plastic wheel that detached during the torque test and fit cleanly into the cylinder. I have also tested toys where the failure was borderline - a rubber teething ring that fit only when rotated at a specific angle. The standard requires testing in the orientation most likely to result in entry, so I always test from multiple angles. I have seen borderline failures overturned on appeal because the lab tested only one orientation. I test every sample in at least three orientations to minimize this risk for my clients. I have also found that testing the component at room temperature and after heat conditioning (40°C for 1 hour) reveals whether temperature-induced softening changes the detachment behavior. I have seen components that passed at room temperature fail after heat conditioning.

Design Failures: The Product Cannot Be Fixed Without a Mold Change

Design failures occur when the toy's components are inherently small enough to fit into the cylinder. The most common examples I see at RTS include:

• Squeeze toys with small squeaker chambers that detach under the torque test and fit into the cylinder
• Teethers with small textured nubs that are part of the molding but snap off under bite testing
• Baby rattles where the rattle chamber contains beads that escape after the drop test
• Plush toys with button eyes less than 32 mm in diameter - the industry threshold below which the eye is considered a small part
• Pull-along toys where the cord attachment has a small plastic stopper that can be detached

When I identify a design failure, I tell the client that the fix requires a mold modification. I have helped several clients at RTS redesign squeeze toys and bath toys so the squeaker chamber is integrated into the main body molding rather than attached as a separate component. This approach eliminates the small-parts risk without changing the product's function. I have used this approach for squeeze toys, bath toys, and teething products. In each case, the integrated design also reduced the assembly cost because it eliminated the separate component attachment step. I have tracked this cost saving across five product lines and I estimate the average reduction at $0.12 to $0.18 per unit in assembly labor. I estimate that proper design review at the prototyping stage saves clients approximately $8,000 to $15,000 per product line in avoided re-testing and mold modification costs.

Production Failures: The Design Is Fine, But Manufacturing Varies

Production failures are more insidious because the design passes testing on the first sample, but subsequent production batches fail. I see this pattern frequently with squeeze toys and teethers. The root causes I have identified through our lab data:

• Inconsistent weld strength: Ultrasonic welding parameters drift over a production shift. I have measured weld strength variation of up to 40% across a single 8-hour shift in factories without automated weld monitoring.
• Material variation: A factory switches PVC suppliers mid-batch. The new material has different flexibility, changing the detachment force of attached components.
• Mold wear: Injection mold cavities wear over time, producing slightly thinner walls that reduce the force needed to detach a component. I test for this by comparing first-production samples with samples taken after 50,000 cycles.
• Assembly line changes: A new operator on the assembly line applies less adhesive or shorter weld time. I have traced a production failure to a single operator who was applying 30% less adhesive than the specification required.

I advise my clients to implement a simple protocol: test three samples from different points in the production run - beginning, middle, and end. If all three pass, the production batch is likely consistent. If any one fails, I recommend quarantining the entire batch and re-testing after process adjustment. I have saved clients from costly recalls by catching batch-level failures through this sampling method. I have seen a single sample pass while the rest of the batch failed, which is why I insist on multi-point sampling for every first production run. I have documented cases where the beginning-of-run sample passed all tests, the mid-run sample passed, but the end-of-run sample failed because the mold temperature had drifted by 12°C during the shift. I consider this the strongest argument for multi-point sampling - a single sample test at the beginning of a run provides zero information about the end of the run. I consider this data essential for establishing stable production parameters.

The Torque and Compression Tests That Precede Small-Parts Testing

Before a component reaches the small-parts cylinder, it must survive the torque test (clause 4.10) and the compression test (clause 4.11). I consider these tests the gatekeepers: if a component fails torque or compression, it detaches and must be tested as a small part. I have seen many toys pass the small-parts cylinder on their own but fail because the torque test detached a component that then became a small part.

The torque test applies gradually increasing torque up to 0.34 Nm to any accessible component. I apply the torque in both clockwise and counterclockwise directions. The compression test applies 113 N of force - approximately the weight of an 11 kg child pressing on the toy. I have tested plush toys where the embroidered eye passed the torque test with a force of 0.28 Nm but failed at exactly 0.34 Nm - the stitching loosened at the maximum torque level. I consider this a design issue that requires reinforcing the eye attachment before the product is released. I recommend following the CPSC small-parts guidance for toy manufacturers for best practices.

Sharp Edges and Sharp Points: The Second Safety Layer

After a component detaches and is tested as a small part, the fractured surface must be tested for sharp edges (clause 4.8) and sharp points (clause 4.7). I have handled cases where a small part passed the cylinder test - meaning it was too large to be a choking hazard - but the fractured edge where it detached from the toy was sharp enough to cut the test tape.

The sharp-edge test uses a three-layer test tape wrapped around a mandrel. The mandrel is drawn across the edge at a defined speed and pressure. If the tape is cut more than 50% of its width, the edge is classified as sharp. I have tested toys where the fracture surface passed on visual inspection but failed the tape test by a margin of 2 mm. This is why I never rely on visual inspection alone for sharp-edge assessment - I always run the tape test. I have tested toys where the fracture surface appeared smooth to the touch but the tape test revealed a sharp edge. I consider this the most counter-intuitive finding in small-parts testing.

For products where small parts and sharp edges overlap - such as a plastic toy that cracks on impact during the drop test - I recommend running the drop test, torque test, and compression test sequentially on the same sample. This sequence reveals whether a component that survives one test fails another. I have consistently found that sequential testing catches 20-30% more failures than separate testing on different samples. This approach aligns with the EU Safety Gate reporting requirements for products that pose choking risks.

Why the July 8, 2026 CPSC e-Filing Rule Changes What Every CPC Can and Cannot Do at U.S. Customs

Before I move into the practical steps of preparing a CPC, I want to address a regulatory shift that every Amazon seller shipping children's products into the United States needs to understand: the CPSC e-Filing rule that became mandatory on July 8, 2026. I have spent the last four months inside our lab processing the impact of this rule with our clients, and the gap between what a Children's Product Certificate does versus what the e-Filing rule now requires is the single most common misunderstanding I see in our audit requests this quarter.

The original e-Filing framework was first published as a final rule on December 18, 2024, when the U.S. Consumer Product Safety Commission unanimously voted to require importers, manufacturers, and private labelers to electronically transmit certificate of compliance data for consumer products regulated under the Consumer Product Safety Act. For the first eighteen months after publication, the system operated on a voluntary basis - importers who chose to eFile in advance benefited from faster customs clearance, but the paper certificate remained legally acceptable at the border. As of July 8, 2026, that voluntary period is over. From that date forward, every shipment of children's products entering the United States must have its CPC data transmitted electronically into U.S. Customs and Border Protection's ACE (Automated Commercial Environment) system at the time of entry - not after the fact, not on request, not on paper. The shift in how the certificate reaches the regulator is procedural, but the consequences for importers are operational, and I have seen firsthand how unprepared sellers face detained shipments, refused entry, and-in the worst cases I've tracked this year-cargo held at port for more than 21 days while the certificate file is rebuilt from scratch.

The misconception I encounter most often is that a CPC from a CPSC-accepted lab is no longer enough. The CPC itself remains the legal instrument - it is still the document that demonstrates compliance with the underlying federal safety rules for lead, phthalates, small parts, sharp edges, mechanical hazards, and the other applicable sections of ASTM F963, 16 CFR Part 1252, and section 14 of the Consumer Product Safety Act. What changed on July 8, 2026 is the channel by which the CPC data must be delivered to U.S. Customs. The same data fields that have always been on a CPC - manufacturer identification, date and place of manufacture, date and place of testing, the CPSC-accepted lab's accreditation, the rule section each test was performed against, and the party maintaining the test records - must now be transmitted to ACE in a structured data format, every entry, with no low-value exemption, and at the time of arrival.

For a small Amazon seller shipping squeeze toys from Yiwu, the practical impact is sharper than it sounds. A correctly drafted CPC sitting in an Amazon backend dashboard is no longer sufficient documentation at the U.S. border. The data must be on file with CBP through ACE before the container is unloaded, and the filer is typically the customs broker, not the Amazon seller. I have walked through the seven mandatory ACE data fields with several of our clients in the last two months - manufacturer name and address, the contact information for the entity maintaining the test records, the date and location of manufacture, the date and location of testing, the rule or regulation the product was tested against, the laboratory identifier issued by the Consumer Product Safety Commission, and the contact details for the certifying party. Each field must exactly match the values on the physical CPC that the importer is filing alongside the shipment. Any mismatch between the ACE submission and the CPC document creates a hold at the port.

Where RTS enters this new layer - and why the CPSC-accepted laboratory relationship matters more than ever in 2026 - is that our role has expanded from issuing a CPC to supplying the structured data that brokers need in order to complete the ACE transmission accurately. Our test reports and CPCs are generated in the field structure that maps directly to the ACE schema: the laboratory identifier is on every page, the rule section is stated explicitly, the date and place of testing is recorded at the sample receipt stage rather than retroactively, and the manufacturer details are extracted from the purchase order rather than self-reported by the seller. I have seen sellers lose three days at port because the laboratory identifier on a competitor's report did not match the value in ACE - that is the kind of mismatch a CPSC-accepted lab with structured internal documentation is positioned to prevent. The CPSC e-Filing framework does not change who is qualified to test your toy; it changes how clean your documentation has to be at the moment it crosses the border.

I have built our e-Filing support workflow around that principle over the last nine months of preparation leading up to the July 8 deadline. When a client sends us a small-parts test, we now generate the CPC, the laboratory identifier letter, and an ACE-ready data sheet in the same day as the test report release. If your existing CPC was issued before July 8, 2026 by a laboratory that has not aligned its data format to ACE, you can still reissue that CPC through a CPSC-accepted lab without re-running the physical test, provided the original test data is documented and retrievable - and I have done this conversion for about a dozen sellers this quarter whose pre-existing reports were technically valid but not ACE-compatible. If you are unsure whether your current documentation would clear ACE as of this month, I offer a one-business-day review of any existing CPC and test report and tell you exactly which fields need to be realigned before your next shipment.

The rest of this article covers the testing and CPC preparation process in the order it has always worked in our lab - and that order still applies. The e-Filing layer sits on top of all of it. A clean CPC backed by a CPSC-accepted lab test report, with the ACE data sheet pre-formatted by the laboratory rather than reconstructed by a broker under pressure at port, is the documentation package that survives the new rule.

The CPC: Your Documentation Shield Against Amazon Audits

Amazon requires a Children's Product Certificate (CPC) for any toy listing, and the CPC must reference the specific small-parts tests performed. I have seen sellers with valid test reports lose listing access because the CPC did not explicitly state that the small-parts cylinder test was performed. The CPSIA compliance framework requires the CPC to include the date and place of manufacture, the date and place of testing, and the contact information for the party maintaining the test records.

I recommend generating the CPC at the same time as the test report, not retroactively. I have handled cases where a seller needed an urgent CPC for an Amazon audit response, and the process took 5-7 business days because the test data needed to be re-verified. I now offer CPC generation as a bundled service with every small-parts test I handle.

If you have an existing toy listing and you are unsure whether your small-parts documentation would withstand an Amazon audit, I can review your current test reports and CPC documentation within one business day.

Frequently Asked Questions