LinkedIn and Vendor Background Checks: What a Buyer Verifies Before the First Order

November 21, 2026Industry Reports0 views36 min read

Registration, barcode prefix, certificate scope, customs records, filings, recalls and staff profiles answer different questions. A buyer's checklist in order.

Key Takeaways

  • GS1 describes its network as Member Organisations in over 110 countries serving more than two million organisations 6.
  • Panjiva, part of S&P Global Market Intelligence, describes its supply-chain product as holding over two billion shipment records from 21 countries with 13 million company-to-company relationships 17.
  • The Commission's annual reporting shows 4,137 alerts in 2024, against 3,412 in 2023 and 2,117 in 2022, with cosmetics accounting for 36% of the 2024 total 29.
LinkedIn and Vendor Background Checks: What a Buyer Verifies Before the First Order

The cheapest check is the one that ends the conversation

A buyer who requests a factory audit before checking who the quoting entity is has spent money to answer the second question before the first. The audit will be performed at an address. If the contract is with a company that does not own that address, the audit report describes a site that has no legal obligation to the buyer, and the report’s most important line is the one nobody reads: the name of the audited legal entity.

Supplier evaluation usually gets discussed as a single judgement. A supplier is “credible” or “not credible”, “vetted” or “unvetted”, and the evidence is assembled until the buyer feels sufficient confidence to place an order. The confidence is real. The problem is that it is being assembled from instruments that establish different things, in an order that is usually accidental, and the resulting composite hides which question each piece of paper answered.

Diligence has a fixed sequence because the questions are nested. Whether a company exists is prior to whether it manufactures. Whether it manufactures is prior to whether it manufactures this product. Whether it manufactures this product is prior to whether it has shipped this product to a market that regulates it. Whether it has shipped is prior to whether it can finance the order, and all of that is prior to whether anything has gone wrong before. A negative answer anywhere in the sequence makes the later questions irrelevant rather than merely weaker. That is the property most evaluations miss. They treat a failed check as a deduction from a score, when it is more often a disqualification of the evidence gathered after it.

There is a second asymmetry, and it is the one that costs buyers money. The strongest evidence available in public data is trade records, and trade records are the most susceptible to misattribution. The weakest evidence is a LinkedIn profile, and it is the easiest to read. Buyers gravitate to what they can interpret. A profile is legible; a bill of lading is not, and the part of it that matters is the part that is usually wrong.

Eight questions, and why the order is not negotiable

The table below is the sequence as a checklist. Each row names the instrument a buyer actually uses, what the instrument establishes, and the specific thing it cannot establish. The rows are ordered by dependency, not by importance.

#QuestionInstrumentEstablishesCannot establish
1Does the entity exist, and is it the entity that manufactures?Company register, credit file, filingsLegal existence, status, officers, registered scopeWhether production happens at the stated address
2Does the product’s identity trace to a named owner?GS1 prefix, Verified by GS1Who registered the number and where; brand owner data if uploadedFactory identity, country of manufacture
3Can this site make this product?BRCGS, ISO certificates, audit reports, IAF CertSearchA named site held a named certificate for a named scope on a dateCapability beyond the scope; current status without a database check
4Has it shipped this product before?Panjiva, ImportGenius, ImportYetiA consignment was declared by a shipper to a consigneeThat the named supplier produced the goods
5Can it complete the order?D-U-N-S, filings, accounts, credit reportsScale, longevity, leverage, payment behaviourWhether the next order will be delivered on time
6Has anything gone wrong?FDA, CPSC, Safety Gate, OPSSA regulator or firm recorded a specific failureWhether the failure was systematic
7Who works there, and what are they hiring for?LinkedIn, job postingsFunctions, seniority, product focus, market focusWhether the stated functions are competent
8Is the company visible at all?Trade press, exhibitions, social channelsA public record of activityQuality; a visible company can be a poor supplier

The dependency runs one way. A supplier that fails row one cannot be rescued by a strong row four, because the shipments in row four were made by whoever holds the bill of lading, and if that is a different company the buyer has been researching a stranger. A supplier that passes row one and fails row three has a factory that cannot make the product, whatever its export history shows about the products it used to make.

Two operational consequences follow. The first is that rows one and two cost nothing and take under an hour, so there is no reason to run them late. The second is that rows three and four are the only ones that generate documents a buyer can put in a file and defend to a compliance team. They are also the two that are most often bought as a package from an intermediary who supplies both, and a report commissioned by the supplier is not evidence in the same sense as a record produced by a customs authority.

The rest of this article works through the eight rows in order, with the specific reading each instrument rewards.

Four sibling pieces in this series carry adjacent ground. The platform and database layer that precedes any named-supplier check is covered in Vertical B2B Platforms and Industry Databases: Where Buyers Look When a General Marketplace Fails. The packaging-level forensics that follow from a barcode are in Working Back from the Shelf: Packaging, Barcodes and Certification Numbers as Factory Fingerprints. The customs-data method applied to a competitor rather than a supplier is in Reading a Competitor’s Container: How Customs Data Reveals the Factory Behind a Brand. Why referral outperforms every instrument described here is in Referral: The Route Buyers Trust Most and Suppliers Reach Least.

Step one: whether the entity in the quotation is the entity in the register

The register answers a narrower question than the one buyers ask

A company register establishes that a legal person exists, when it was formed, where its registered office is, who its officers are, whether it is still trading, and what it told the registrar it does. In the UK that information sits in Companies House, which publishes incorporation documents, annual accounts, confirmation statements and officer appointments 1. In the EU, a VAT number can be tested against national databases through VIES, which tells a buyer whether the number is valid for cross-border supply, not whether the company is solvent 2. The regulatory layer sits alongside those registers rather than inside them: the general product safety regulation requires an economic operator established in the EU to be responsible for non-food consumer products placed on the market, which is why a buyer in that market frequently meets an importer of record before meeting a manufacturer 3. In China, the National Enterprise Credit Information Publicity System (国家企业信用信息公示系统) publishes the unified social credit code, legal representative, registered capital, business scope and operating status for registered entities 4. In the United States, there is no single national register of companies; incorporation is a state function, which is why US diligence usually starts from a credit file rather than a register.

None of these systems records what a buyer most wants to know. A registered office is frequently an address the company does not occupy. Registered capital in the Chinese system is a subscribed figure, not paid-in capital, and it is not a statement of the company’s ability to fund a production run. Business scope is a permission list, phrased in the registrar’s vocabulary, and a company can hold a manufacturing scope while owning no plant.

The check still earns its place first, because it is the cheapest way to discover that the counterparty is not who the marketing said. The pattern worth naming: a buyer receives a quotation on letterhead that carries a factory name, an address in an industrial district, and a certification logo. The register entry for the legal name on the quotation shows a registered office in a commercial building, a formation date two years ago, and a business scope that lists wholesale and import-export rather than manufacture. Nothing about that register entry is illegal. It is simply a different company from the one the letterhead describes.

The trading company problem, stated plainly

Most supplier evaluations collapse at this point, and the collapse is not caused by fraud. It is caused by an industry structure in which the entity that sells is often not the entity that makes. A trading company aggregates capacity across several factories, quotes a price that reflects a chosen factory, and manages the export paperwork. For a buyer placing a small first order across several SKUs, that aggregation is a service. For a buyer building a private-label programme with a specification, a QC regime and a claim about the factory on the packaging, it is a layer that has to be understood rather than removed.

The question to ask is not “are you a factory” but “which legal entity will be the seller on the contract, and which entity will be the producer on the documentation”. Those are two answers, and a supplier that cannot separate them is either confused or evasive. Three documents settle it. The business licence or register extract names the seller. The certificate for the production site names the producer, because a BRCGS or ISO certificate is issued to a site and an entity, not to a brand. The customs record names the shipper, which is frequently the trading company rather than the factory.

When the three names differ, the buyer is not looking at a problem to be refused. They are looking at a supply chain with an intermediary in it, and the diligence budget has to cover the intermediary and the producer separately. What the buyer should refuse is a supplier who presents one name in the quotation, a second on the certificate, a third on the bill of lading, and describes all three as “our factory”.

What the address and the scope actually tell you

Two register fields repay more attention than they usually get.

The registered address, read against a map, distinguishes an industrial plot from a serviced office. That is a weak signal taken alone, since a legitimate factory can have its registered office at a group headquarters. It becomes a strong signal when it is the only address the company publishes, when the same address appears on the certificates of four unrelated companies, and when the company’s own website shows a building it does not name in the register.

Business scope, in the Chinese system, is worth reading literally. A scope that includes manufacture of a specific product class tells the buyer that the registrar was satisfied the company is in that business. A scope that lists a wide range of unrelated goods alongside the product in question is characteristic of a trading entity. Neither reading is conclusive; both are worth ten minutes before a sample order is placed.

The third field is status. A company in the process of deregistration, or one whose annual publicity obligation has lapsed, can still quote. Registers publish status for a reason, and a buyer who checks existence without checking status has done half of a free check.

What a registration check cannot do

It cannot tell a buyer whether the company is good for the order. Registration is a fact about legal personality. Buyers who treat a clean register entry as clearance are treating the first step as the whole staircase.

It also cannot be relied on across borders without knowing which register applies. A Hong Kong company selling into the EU, with production in Guangdong, may appear in three registers and none of them will describe the factory. The rule that follows is unglamorous: establish which jurisdiction the contracting entity is registered in, check that register, and check the producer’s site documentation separately, because they are different questions asked of different institutions.

Step two: whether the identity travels with the product

What a GS1 prefix actually records

A barcode is not a registry lookup, and the misconception is common enough to be worth stating in the plainest terms available. GS1 allocates a company prefix to a member, and the prefix is issued by the GS1 Member Organisation in the country where that company is registered 5. GS1 describes its network as Member Organisations in over 110 countries serving more than two million organisations 6. The leading digits of a GTIN therefore point to the issuing Member Organisation, which usually corresponds to the brand owner’s country of registration.

That is the entire claim the prefix supports. It does not point to a factory. A US brand that contracts production in Vietnam and China will hold a US prefix, because the prefix belongs to the brand owner, not to the plant. A European retailer’s own-brand product will carry the retailer’s prefix regardless of where it was made. A buyer who reads a country prefix as an origin label has inverted the meaning of the number: the prefix is evidence of who took responsibility for the identifier, which is a statement about the brand, not about the supply chain.

Where the prefix becomes useful is in narrowing the field of possible owners. If a sample arrives with a prefix issued in a country where the supplier claims no presence, the buyer has found a mismatch worth explaining. If two samples from the same supplier carry prefixes issued in two different countries, the buyer has found either a licensed range or a repackaged product, and those are different facts to establish.

The product record behind the number

Verified by GS1 returns brand owner, product description, net content and image where the brand owner has uploaded them 7. The qualification matters: the service returns what the owner chose to publish. A sparse record is not evidence of a small company, and a complete record is not evidence of a careful one.

What the record cannot return is the factory, the country of manufacture, the importer, the consignee, the shipment history or the price 7. Those absences are structural rather than a gap in the service, because the identifier was designed to move a product through a checkout, not to describe a production chain.

The practical consequence for a first order is that a barcode check belongs in the identity step and nowhere else. It confirms that the product on the table is a product someone registered, under a name the buyer can pursue. The packaging forensics that go further, print method, carton coding, lot marking, are treated in the sibling piece on working back from the shelf, and the reader who needs that level of reconstruction should go there rather than expecting a GTIN to carry it.

Why this step is cheap and still skipped

The check costs a search and a minute of reasoning. It gets skipped because it produces a negative more often than a positive, and negatives are uncomfortable on a first call with a supplier the buyer has already decided to like. A buyer who runs it will occasionally discover that the product they were shown is a registered brand belonging to a third party, which is a different conversation from the one they planned. That conversation is cheaper on day three than in month seven.

Step three: whether the site can make the product

A certificate names a site, a scope and a date

Certification is the instrument buyers lean on hardest and read least precisely. A BRCGS certificate is issued to a specific site against a specific standard, with a scope that names the product categories covered, and an expiry date. BRCGS operates separate standards for food safety, packaging materials and consumer products 8 9, and a company that holds one is not thereby certified against the others. The certificate number can be looked up in the BRCGS directory 10, which is the only place the current status of the certificate can be established.

The scope is where most readings go wrong. A certificate covering “prepared pet food” does not cover a baked biscuit line if the standard’s category definitions separate them. A site certified for packaging does not hold a food-safety certificate for the product inside the packaging. The buyer’s question is not “are you certified”, it is “read me the scope line, the site address and the expiry date on the certificate that covers this product”.

The site address is the second trap. Group certification exists in several schemes, and a corporate certificate can be presented for a plant that is not in scope. When the address on the certificate is not the address that will make the goods, the certificate is a document about a different building.

Verifying that the certificate is real and current

Certificates are documents, and documents are falsifiable in ways that photographs of documents are not. The verification layer is the accreditation chain. A management-system certificate is issued by a certification body that was accredited to issue it by an accreditation body, and the International Accreditation Forum’s CertSearch database indexes certificates reported by those bodies: the site reported 3,794,343 certifications from 2,577 certification bodies across 81 accreditation bodies when checked on 15 November 2026 11. The database states its method plainly, asserting validity only after cross-checking the certificate, the certification body’s accreditation and the accreditation body’s signatory status for that standard 11.

Two features of that service matter more to a buyer than the record count. It reports suspension and withdrawal, and it can alert a subscriber when a certificate changes state 11. A certificate that was valid at the sample stage can be suspended before the second order, and the buyer who filed a PDF has filed a historical document.

The standards themselves are stable reference points rather than endorsements. ISO 9001 is a quality-management standard with a defined edition 12, and ISO/IEC 17021-1 governs the bodies that audit management systems 13. Neither standard audits a product. A supplier with ISO 9001 has had its quality management system assessed against a requirement set; it has not had a single unit of the product tested.

Three errors, in the order they occur

The first error is reading a brand’s certificate as a factory’s. A brand owner that contracts production does not hold a site certificate for its supplier’s plant, and a certificate in the brand’s name is a certificate about an office.

The second is accepting a scope that names a product family the buyer is not buying. This happens most often in categories where the buyer’s product sits between standard categories: a treat that is also a supplement, a bed with an electrical component, a toy with a battery.

The third is accepting a current-looking certificate without checking status. Expiry dates are printed on certificates and are easy to read; suspension is not printed anywhere except the database.

Social audits are a different instrument

The most persistent category error in this field is treating a social audit as a product certificate, or the reverse. A SMETA audit, delivered through Sedex, examines labour practices and produces an audit report with findings and corrective actions; Sedex describes the platform as used by over 100,000 companies and SMETA as the most widely used social audit methodology 14 15. amfori BSCI produces a monitoring result with a rating 16. Neither certifies that a product conforms to a standard, and neither substitutes for a product certification or a factory inspection of the production process.

The distinction has a practical consequence at the first order. A buyer who needs to know whether a plant can hold a specification needs a process audit and product testing. A buyer who needs to know whether the plant can be named in a retailer’s supplier programme needs the social audit the retailer specifies. Buying one when the other is required produces a file that looks complete and fails at the retailer’s compliance desk. The full treatment of the two instrument families is in the sibling piece on global supermarket pet aisles and the private-label entry path, which covers what a retailer’s audit actually asks for.

Step four: whether the supplier has shipped this product before

What the trade-data tools contain

Customs records are the only public dataset that connects a named company to a named consignment. In the United States, bills of lading filed for imports are released in aggregated form and resold by commercial platforms. Panjiva, part of S&P Global Market Intelligence, describes its supply-chain product as holding over two billion shipment records from 21 countries with 13 million company-to-company relationships 17. ImportYeti offers a free search over a counter it displays as 715,151,312 customs shipment records updated daily, covering US import and export bills of lading plus Mexican customs declarations 18. ImportGenius sells a comparable US-anchored dataset at a lower price point and publishes no independently audited record count 19.

Coverage is the first thing to establish, because it is narrower than the marketing implies. Panjiva’s coverage varies by country, and the Wharton dataset note records China coverage in a 2011–2018 window 20. That window matters: a buyer checking whether a Chinese factory shipped to Germany in 2025 is not looking at the same dataset as a buyer checking 2016. ImportYeti’s coverage is explicitly US plus Mexico 18. Neither tool observes domestic-only trade, and neither observes a shipment that was routed through a third country’s trader.

The second thing to establish is what a record contains. A US bill of lading gives a shipper, a consignee, a port of lading, a port of unlading, a date, a description, a weight and usually a container count. Declared value is frequently absent, and where present it is a declared customs value rather than a transaction price. This is why trade data is a relationship dataset and not a value dataset. A buyer who divides a container’s declared value by its unit count to estimate factory cost is computing an artefact.

The attribution problem

Here is the finding that matters most in this article. A shipment record proves that a consignment left a port under a named shipper, consigned to a named buyer, on a date. It does not prove that the named supplier manufactured the goods.

Four mechanisms produce the gap.

The shipper is whoever books the movement, which is often the exporter’s trading arm, a freight forwarder acting as shipper of record, or a consolidator. A record showing a Ningbo-based logistics company as shipper says nothing about the plant.

Consolidated containers mix goods from several producers under one bill of lading. A single line on a house bill can represent three factories’ output, and the description field will not say so.

The description field is free text. “Pet products”, “pet supplies”, “dog toys” and an HS code appear in inconsistent forms, and HS codes in bill-of-lading data are self-declared. Two searches for the same goods using a keyword and a code return different sets, and neither is complete 18.

The name itself is unstable. A factory that exports through a Hong Kong entity appears under the Hong Kong name; the same plant may also appear under a mainland name with a slightly different romanisation, and a group with several plants may export all of them under one trading name. Name matching across these variants is the analyst’s real work and the reason two buyers can run the same query and reach opposite conclusions.

The consequence is that trade data is strongest as evidence of a relationship that existed and weakest as evidence of who produced the goods. It is excellent at answering “does this brand buy from this region, in this category, at this frequency”. It is unreliable at answering “does this supplier make this product”. The gap between those two questions is where most misreadings happen.

How a buyer reads a shipment record without over-reading it

A defensible reading treats each record as one of four grades of inference.

Grade one, the consignee. The record names a buyer. That is a fact about the buyer’s sourcing geography and it is rarely wrong. If a prospective supplier claims a US retail brand as a customer and the brand appears as a consignee on shipments from the supplier’s city in the right HS chapter, the relationship is plausible.

Grade two, the shipper name matching the supplier’s legal name or a name the supplier has disclosed as affiliated. This is stronger. It still does not establish that the goods were produced at the site the buyer intends to audit.

Grade three, the shipper name matching the audited site’s certificate holder. This is the only configuration in which trade data and site documentation corroborate each other, and it is worth the effort to reach.

Grade four, a shipment traceable to the specific production lot. Public data does not support this grade, and a provider who implies otherwise is selling an inference as a fact.

Three practical moves follow from the grading. Ask the supplier for its exporter of record before searching, then check whether the name in the data is the name on the contract. Pull two years of records and look at cadence rather than volume: a supplier shipping six containers a year to one buyer is a different counterparty from one shipping monthly to twelve. And treat the absence of records as ambiguous rather than damning, because a factory that sells through a trading company, or ships to a market outside the dataset’s coverage, leaves no trace in it.

A fourth move is available to buyers with a compliance file to defend. Where the record matters to a decision, buy the underlying document set rather than a screenshot of a database row, and record the query, the date and the coverage window. Trade data is a snapshot of a database that changes as records are added and corrected. A buyer who cannot reproduce their own search has an anecdote.

Step five: whether the company can complete the order

What the identifiers and filings establish

Financial standing is the check buyers most often skip on a first order, and the reason is rational: the order is small, the deposit is small, and the analysis costs more than the exposure. The calculation changes as soon as tooling, moulds or a private-label print run are involved, because those costs are sunk, supplier-specific and unrecoverable. At that point the buyer is underwriting the supplier’s continuity.

The entry point is identity, again. Dun & Bradstreet issues the D-U-N-S number, a nine-digit identifier assigned at site level, and it is required for registration in the US federal procurement system 21 22. A supplier that already holds a D-U-N-S number has been through an identity resolution step, which is useful but not a credit opinion. The credit file attached to that number is the opinion, and it is built from payment behaviour reported by other creditors, public filings and trade references.

For UK-registered counterparties, Companies House publishes annual accounts, so a buyer can read turnover, net assets and, for smaller companies, the exemptions claimed 1. For most other European jurisdictions the accounts are in a national register with varying levels of public access, and for Chinese entities the annual publicity filing is available through the credit information system, though the depth of financial data published there is limited 4. Where a group files consolidated accounts, the buyer should expect the subsidiary to be opaque and read the group figures as context rather than as the counterparty’s own.

Third-party aggregators reduce the labour of this. OpenCorporates indexes company-register data across jurisdictions and is useful for finding which register holds a given entity and whether a director appears in several companies 23. The interpretation still belongs to the buyer, and the standard caveat applies: an aggregator is a copy of a register, and where the two disagree the register governs.

The signals that arrive free with a filing

Three signals are worth extracting because they cost nothing beyond reading.

Longevity: a formation date ten years old, with accounts filed every year, is evidence of administrative continuity. A formation date two years old is not a negative, but it does mean the supplier’s trading history is short and the buyer’s first orders may be among its first.

Leverage: where accounts are public, the ratio of debt to net assets and the trend in trade creditors indicate whether the company funds production from its own balance sheet or from its suppliers’ patience. A supplier whose trade creditors grow faster than turnover is financing itself on extended terms, and a buyer’s prepayment is a contribution to that.

Director overlap: a director appearing across several companies in the same product category is common and not sinister, but it explains why two “different” suppliers quote similar prices, and it tells the buyer which entity actually holds the assets.

What none of this establishes is delivery performance. A solvent company misses dates; an undercapitalised one sometimes outperforms because its survival depends on the order. Financial diligence answers whether the counterparty can absorb a problem, not whether it will create one.

Step six: whether anything has gone wrong before

The recall databases and how to read a hit

Recall data is the most searchable negative evidence available, and it is published by regulators with a specificity that makes it usable. The FDA maintains a recall table for animal and veterinary products, distinguishing firm-initiated recalls, FDA-requested recalls and FDA-ordered recalls, and at the point of access on 15 November 2026 the table held 62 entries, the most recent dated 10 November 2026 and concerning hay cubes with a botulism risk 24. The FDA also runs a complaint channel that generates some of those entries 25.

The Consumer Product Safety Commission maintains the equivalent for consumer products, and pet supplies appear as a distinct category in its recall filter, with the database exportable for offline analysis 26. A 2025 illustration of why the category exists: Petmate’s Pet Zone pet toys were recalled over a button-cell battery that could become accessible to a child if the illuminated sphere broke, a hazard framed under the mandatory standard for consumer products with button-cell batteries rather than under any animal-safety rule 27.

Europe runs a different instrument. Safety Gate is the EU rapid alert system for dangerous non-food products, and alerts are filed by national market surveillance authorities when they take measures 28. The Commission’s annual reporting shows 4,137 alerts in 2024, against 3,412 in 2023 and 2,117 in 2022, with cosmetics accounting for 36% of the 2024 total 29. The annual reports themselves are published by the Commission 30. In the UK, the Product Safety Database performs an analogous function for notifications and measures 31.

Reading a hit requires three distinctions.

The first is between a firm-initiated recall and a regulator-ordered one. Most recalls in the US system are voluntary actions taken by the firm, often before any regulator involvement. A voluntary recall is evidence that a company detected a problem and acted, which is a quality-system signal in the opposite direction from the one buyers usually assign.

The second is between a product failure and a labelling failure. Undeclared allergens, incorrect guaranteed analysis, and a missing caution statement produce recall entries that say nothing about the plant’s process control and everything about its specification review.

The third is between the entity and the site. A recall notice names a firm. The firm may be a brand owner that contracted the product, in which case the notice tells the buyer something about the brand’s supplier management and nothing directly about the factory.

Aggregate counts should be handled with more care than they usually get. A rising alert total measures reporting intensity as well as product failure, and the Commission’s own analysis attributes part of the 2024 increase to a single product category 29. A buyer who reads a rising total as a deteriorating market has confused an enforcement statistic with an incidence statistic.

Disclosed versus concealed

The distinction that should drive a sourcing decision is not clean versus dirty. It is disclosed versus concealed.

A supplier with a recall in its history that it describes accurately, with the cause, the corrective action and the follow-up test, has demonstrated the behaviour a buyer needs when something goes wrong at 2am. A supplier with no recall record is not necessarily better; small exporters with no US or EU distribution cannot appear in these databases at all, because the databases record actions taken in jurisdictions the supplier has never shipped to. Absence of a record is absence of exposure.

A supplier that has had a recall and does not mention it, when the buyer’s own search finds it, has supplied a different kind of information. The failure is not the recall. It is the gap between the recall and the account of the company. Buyers who treat the two as equivalent are choosing suppliers on the wrong variable, and they are selecting for the ones who are better at not mentioning things.

Step seven: the people

What a company page states, and what a headcount band hides

A LinkedIn company page is the only widely used public artefact where a supplier describes itself in a structured way, and it is the instrument buyers read most confidently. What it actually contains is: a description written by the company, a headcount band rather than a number, a headquarters location, a follower count, and the profiles of employees who have chosen to list the company as their employer 32.

The headcount band is the field buyers over-read. A band of 51 to 200 employees is compatible with a factory of 180 staff and with a trading company of 12 people plus a warehouse. It is a self-selected range, unverified by the platform, and it is not a capacity figure. Capacity requires a number of lines, a shift pattern and an output rate, none of which appear on a company page.

What the page is genuinely good for is the function list. A manufacturer with in-house tooling will show toolmakers and mould designers. A manufacturer with a quality organisation will show QA engineers, a QC manager, a regulatory affairs person. A trading company will show sales, merchandising, logistics and sourcing staff, and will show very few people whose job is to make anything. Reading the titles is a five-minute test of whether the company’s self-description matches its staffing.

Job postings as a reading of strategy

Most buyers never open the jobs tab, and it is the most informative page a supplier publishes. A job advertisement is written for a purpose different from marketing: it has to describe the role accurately enough to attract a candidate, which forces the company to name the site, the function, the reporting line and often the customer market.

Three readings follow.

The roles being hired describe the current constraint. A factory advertising for a mould maintenance technician and a die-cutter operator is running tooling and expanding capacity in a moulded product. A factory advertising for a quality engineer with experience of a named retailer’s audit scheme is telling the reader which customer it is trying to keep or win. A factory advertising for an export sales manager with a named language is telling the reader which market it intends to enter.

The site named in the advertisement anchors the company to a real place. A posting that names a street address in an industrial zone corroborates the address on a certificate. A company whose postings are all for office roles in a city centre, with no production roles anywhere, has described its own structure.

The volume and cadence describe trajectory. Continuous hiring in production roles over a year suggests expansion or turnover. A single burst of 20 production postings followed by silence suggests a contract that started and ended.

The limitation is symmetrical with the value: a job posting is a plan, not a capability. It states an intention to hire a quality engineer. It does not state that anyone competent was hired, that the quality system works, or that the advertised audit experience exists on site. Buyers who treat postings as proof of a functioning quality organisation have converted an advertisement into evidence it is not.

Employee profiles carry a similar double edge. A cluster of process engineers with long tenure is a good sign in a manufacturing business. A sales team whose profiles all list the same two years of tenure is a different signal, though which one depends on whether the company is growing or churning. Profiles are self-authored, incomplete and often stale, and the platform’s own privacy settings vary by jurisdiction, so a thin result set may reflect the platform’s reach in that country rather than the company’s opacity.

The step that requires a phone call

The people layer produces names, and names make the last part of this step possible: asking a named individual a question whose answer is checkable. A quality manager will answer a question about the last third-party audit date. A production manager will answer a question about which line runs the buyer’s product. The answers are not evidence on their own, but a refusal to give a checkable answer is information, and it arrives before any money moves.

Step eight: the public footprint and its absence

The final check is the least structured and the most revealing in the aggregate: whether the company appears anywhere at all. A manufacturer with a decade of exports normally leaves traces in trade press, in exhibition catalogues, in a customer’s supplier disclosure, in a distributor’s product page, in a certification directory. Pet trade publishing covers plant investment, certification and recalls at a level of detail that makes a genuine exporter hard to miss 33 34. Association membership adds another marker, though membership lists vary in how much they publish about their members 35.

Two buyers can look at the same silence and reach opposite, defensible conclusions, which is why the step belongs to the buyer’s risk posture rather than to a rule.

The first reading treats absence as risk. A supplier with no exhibition history, no trade-press mention, no certification entry and no employees on a professional network is a supplier whose claims cannot be corroborated by anything except its own documents. For a buyer whose exposure includes a retailer’s compliance audit, that is a fail, because the file will not survive scrutiny by a third party.

The second reading treats absence as no evidence. A competent small factory selling through trading companies, or into one market through one importer, can genuinely have no public footprint: it does not exhibit, does not advertise, does not need a professional-network presence, and its certificates may be held by its customer rather than displayed. Silence in that case is a fact about the supplier’s channel, not about its capability.

Both readings are correct under different exposures. The buyer placing a 500-unit trial order through a trading company can absorb the second reading. The buyer appointing a supplier for a private-label programme with a retailer’s name on the packaging cannot, because the retailer’s audit will ask for the evidence the footprint would have provided. The error in either direction is the same: treating the absence as a verdict rather than as a statement about which channel the supplier actually serves.

What the sequence cannot establish

Six limitations survive every check described above, and a buyer who does not state them will over-claim in the file.

Certification is not inspection. No certificate in this article tests a unit of the buyer’s product. A factory can hold ISO 9001, a BRCGS certificate covering the right scope and a current social audit, and still produce a first order that fails the buyer’s specification, because the specification was never transmitted precisely enough.

Trade data is not provenance. The attribution problem is structural, not a defect that better searching fixes. A buyer who needs to know which plant made a consignment needs either a supplier disclosure or a factory visit, and both are forms of trust rather than of verification.

Recall absence is not safety. The databases record actions in the jurisdictions that publish them. A supplier with no US or EU distribution cannot appear, and a supplier whose customer handled the recall will appear as a name the buyer does not recognise.

Financial data is lagged by construction. Accounts are filed after the year they describe, and a credit score built on payment behaviour trails the events that changed it. A supplier in difficulty usually looks acceptable in the last filed year.

People data is self-authored and platform-mediated. A profile is a claim, a posting is a plan, and a headcount band is a range the company chose.

Public visibility measures channel, not capability. The most visible supplier in a category is often the one with the largest marketing budget, and marketing budgets are funded by margin, which is funded by price.

Two conflicts within the sources are worth recording rather than smoothing. The first concerns scale figures for the trade-data platforms: Panjiva’s own product page cites over two billion records across 21 countries, while an earlier published description cites over one billion records representing roughly 40% of global merchandise trade by value 17. The two are different vintages of the same product and cannot be used interchangeably; the article uses the 21-country framing and treats coverage as country-specific. The second concerns Safety Gate totals, where a single-year figure of 4,137 alerts for 2024 sits alongside multi-year averages of roughly 2,000 to 2,200 alerts per year in secondary analyses of the same system 29. The difference is between a year and an average, and quoting one as the other misstates the trend.

There is a final point about the sequence itself. Because the checks are ordered by dependency, the marginal value of each later check falls as the earlier ones are done well. A buyer who has established the contracting entity, its registration, the certificate scope and the shipper name has answered most of what public data can answer. The remaining instruments refine a picture whose outline is already fixed. Buyers who spend their diligence budget in the reverse order, on a factory audit and a LinkedIn deep read before establishing who the seller is, buy detail about a counterparty they have not yet identified.

Method notes

This article was reported between 10 and 15 November 2026 from primary sources: regulator databases and their published tables, first-party certification and standards documentation, and the product pages of the commercial data platforms described. Figures carry the year and scope of the source as published; live counters on vendor sites are reported as the state of that page on the access date and are marked as unaudited. Fact-base entries for Panjiva, ImportYeti, ImportGenius, GS1, amfori, Sedex, APPA and the European and UK trade press were inherited with their own access dates and are cited to those publications rather than re-derived.

Excluded from scope: pricing and payment terms, which belong to a commercial negotiation rather than a verification step; the discovery layer of platforms and directories, covered elsewhere in this series; and laboratory testing regimes, which vary by product category and market and are treated in the compliance series. No supplier was interviewed for this article and no individual company’s diligence file was used. Where a claim about industry practice could not be sourced to a document, it was written as an analytic judgement rather than attributed to an unnamed authority.

Featured Companies

Companies from the GlobalPetIndex database active in this area of the pet industry.

Browse the full company directory →

Related reading

GlobalPetIndex Editorial Editorial lead at GlobalPetIndex, covering pet industry intelligence, market trends and company research.

Editorial lead at GlobalPetIndex, covering pet industry intelligence, market trends and company research.

Share this article:

Putting this into practice?

Low-MOQ, private label and compliance checks for small pet retailers

Small-pet-business sourcing hub →
GPI

GlobalPetIndex

Open the app for a smoother experience

Open Privacy