Build1 distinct publisher3 min readUpdated
A developer field guide maps FDA registration, GUDID, NANDO and EUDAMED onto the checks a private-label brand owner actually needs. Its own sample code shows where the comparison, not the lookup, breaks.
The Engineer · Build desk
Compiled by The EngineerSomething wrong?How this is made
The lookup is the easy half. In the field guide's own sample function, the OpenFDA registrationlisting endpoint is queried by registration number [4], the product codes are pulled from the first returned listing, and a match is declared only when the expected product code appears in that set and the expected address appears as a lowercased substring of `address_1` [c6b]. Set membership on a product code is a real test. Substring matching on a street address is not. Factories write their own address differently on a certificate than the registration filing does, and two units in one industrial park can share enough of a string to pass. Since the stated goal is resolving a registration number to the exact legal name, owner/operator and site you are auditing [13], address normalisation is where the engineering hours actually go.
The same example has a second seam. The guide describes the minimal flow as three assertions, one of them being that the registration is active [6], while the sample code reports `status` as an output field rather than testing it [c6b]. A factory whose registration exists but has lapsed clears that function.
Throughput is not the binding constraint. Without an API key the device API permits 40 requests a minute [5], which is 2,400 an hour [1], and a key lifts the ceiling six-fold [2]. At that rate, how often you re-verify a supplier base is a scheduling choice rather than a capacity one, which is the real difference from document review: a PDF cannot tell you that a registration lapsed last quarter, and a repeat query can.
EUDAMED is the part that a query cannot yet close. Article 33 of EU MDR 2017/745 requires economic operators to register once the database is fully activated, and actor, device and certificate records all sit there, but the modules come online progressively, so the guide's advice is to confirm an actor registration now and re-check when the certificate module is active [11]. A supplier file therefore cannot store a verdict on its own. It stores a verdict plus the date a new module changes what the verdict is worth.
What makes any of this enforceable rather than merely tidy is that both regimes require the underlying records: 21 CFR Part 807 Subpart C for US establishment registration and device listing, Part 820 for quality systems [2], Articles 52 and 31 of EU MDR 2017/745 for conformity assessment and operator registration [3], and Part 830 for the UDI data that populates GUDID [7]. The one artefact with no public registry behind it is the ISO 13485:2016 certificate, where the guide's test is that the certificate number resolves on the issuing certification body's own search tool [12]. That pushes the question upstream: which certification bodies you are willing to accept becomes a procurement decision, because for that document there is no government database to fall back on.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
The guide addresses tooling for private label medical devices, using microneedling pens as a category sitting at the border of cosmetics and regulated devices, and states that the hardest part is not manufacturing but verifying the paperwork, which arrives as PDFs, scans and screenshots.
The guide's framing: a certificate PDF is a claim, and the database record is the evidence.
In the United States, establishment registration and device listing are required under 21 CFR Part 807 (Subpart C), and quality systems under 21 CFR Part 820.
In the European Union, EU MDR 2017/745 Article 52 governs conformity assessment and Article 31 requires economic operators to be registered.
The FDA publishes establishment registration and device listing data through the OpenFDA device API at https://api.fda.gov/device/registrationlisting.json, queryable on the registration number, device listing status, owner/operator and the product codes a facility lists.
The OpenFDA device API rate limits are 240 requests per minute with an API key and 40 per minute without one.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Specific and checkable, but single-source and undemonstrated
The article supplies unusually concrete, externally checkable artefacts: a named API endpoint, a query field, explicit rate limits, and citations to 21 CFR Parts 807/820/830 and MDR Articles 31/33/52. That lifts it above assertion-only commentary. It is pulled down by having exactly one source, no independent corroboration of any regulatory or API detail, no worked example output or verification run, and a reference implementation whose logic does not match the prose it illustrates.
Only a vendor self-disclosure
The cluster contains no release, deployment, benchmark or usage metric for the described verification approach or for any tooling built on these databases. The single adoption-adjacent signal is the authoring vendor asserting that its own compliance team queries these sources and maintains a checklist, with no counts, customers or timeframe. That is not enough to score adoption in either direction.
Modestly overstated by its own reference code
The thesis that vendor verification is 'a database query, not a document review' is directionally sound and backed by real public data sources, so this is not inflated marketing. But the framing overshoots the delivered artefact: the sample function would return match=True for a lapsed registration, trusts the first result without disambiguation, and substring-matches addresses; NANDO scope adjudication and progressively activating EUDAMED modules are explicitly judgement-and-wait problems, not queries; and the 'ten minutes per document' cost claim is asserted with no measurement. Positive but small.
Vendor content marketing with an on-page conversion path
The post is published under a supplier-affiliated dev.to account and its closing section directs readers to TBPHP's OEM documentation checklist, sample technical file structure and the tbphp.shop OEM Guide, positioning TBPHP's compliance team as the party that performs these verifications for private-label buyers. The technical content is genuinely useful and does not depend on the vendor, which limits distortion, but the commercial interest in framing supplier verification as hard and outsourceable is direct and undisclosed as a conflict.
Moderate: verifiable claims, one interested source
Confidence is held mid-range because the substantive claims are precise and independently checkable against public regulators, and the internal contradiction between prose and code is directly observable in the text. It is capped by a single publisher, an author with a commercial stake, no corroboration of the cited rate limits or regulatory articles, and no adoption evidence at all.
science
3.2 million replies, 88 candidates, and the gender gap that prevalence metrics erase1 distinct publisher
build
Force the tool call, then hand Lightsail a long-lived key1 distinct publisher
science
The self-driving lab is out. Whether AI shows up in your filing is still open.1 distinct publisher
science
Text watermarks land on 2 December. The detection they imply does not.1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
dev.to
1 article · August 24, 2026