How ChinaValidate Tests Company Reports

This page documents how ChinaValidate checks company-report output and where that evidence stops. A report can pass presentation tests and still require a buyer to verify the transaction, source records, contract, product, and payment instructions. The method below is a dated release record, not a certificate that a company is safe.

Method snapshot: Method v1, reviewed 15 July 2026

Interface under test: public sample company report

Current browser evidence: installed stable Chrome controlled by Playwright 1.60

Viewports: 1440 x 1000, 390 x 844, and 320 x 800 CSS pixels

The method starts with meaning, not screenshots

The first test is whether the report keeps one company identity intact. The registered Chinese name, Unified Social Credit Code, registration status, legal representative, report date, and report reference must remain associated with the same entity at the top of the page. A visually polished report fails if a field belongs to another candidate company or loses its source time.

The second test is whether data states remain distinct. A module with returned records is different from a checked module with no matching records, and both differ from a source that was temporarily unavailable. ChinaValidate's report presentation carries those states separately. This matters because “no records returned” must not be rewritten as “no litigation exists”; the negative-result interpretation guide explains that boundary.

The third test is temporal. Report-level generation and completion times identify the package, while public-record modules carry their own capture times. China's current enterprise information disclosure regulations make government departments and enterprises responsible for the truthfulness and timeliness of the information each discloses and provide for corrections. A report snapshot can therefore be reviewed as evidence from a stated time; it should not be described as permanently current.

Current browser evidence

On 15 July 2026, the public sample report was loaded from a clean browser context at each viewport below. The check required HTTP 200, the correct company H1, the same top identity evidence, no document-wide horizontal overflow, no console or page exception, and no failed network response.

Profile	Viewport	Observed result
Desktop	1440 x 1000	Passed; report width 1440, top evidence and section sequence present
Mobile	390 x 844	Passed; report reflowed to 390 without document overflow
Narrow reflow	320 x 800	Passed; all 28 report sections remained present at width 320

The 320-pixel check is tied to the W3C's explanation of reflow, which uses 320 CSS pixels as the vertical-content reference for avoiding two-dimensional scrolling, with exceptions for content that genuinely requires it. Passing this width is one useful layout signal; it is not a claim of complete WCAG conformance.

What the browser assertion actually catches

• Wrong route or missing report: the initial response is not 200 or the expected report title is absent.
• Identity drift: the visible company heading or anchor fields do not match the fixture being tested.
• Broken responsive layout: the document becomes wider than its viewport, usually because of a table, code-like value, or fixed-width module.
• Failed assets or requests: images have no natural dimensions or a page request returns an error response.
• Client/server disagreement: hydration, JavaScript, or page errors appear in the browser console.

Playwright can be configured for Chromium, Firefox, WebKit, branded browsers, and mobile projects. Its device emulation can also control viewport, screen, touch, locale, and timezone. Those are available testing capabilities, not evidence that every configuration has passed ChinaValidate's current report build. The matrix above records only what was actually run.

Browser report and PDF are separate outputs

The browser report supports on-screen review and preserves the full page hierarchy. The downloadable PDF is a print copy of the unlocked company detail page: it uses the same saved snapshot, expands the advanced modules, and includes the same AI analysis, sources, and limitations while omitting navigation and interactive controls. Its generation tests require a completed report and verify that a downloadable file is written; incomplete reports are rejected rather than exported as finished evidence.

The PDF lists module status and capture time and includes a Sources & Limitations section. When a module has more data than the document layout includes, the PDF labels that it contains the first 100 normalized records. A buyer should not treat a truncated PDF as the complete underlying dataset. Save the report reference and use the approval-file guide to preserve related contracts and payment records.

Known limits of Method v1

Browser coverage: the current documented run uses stable Chrome, not Firefox, Safari/WebKit, assistive technology, or every operating system. Device coverage: fixed viewports reproduce layout constraints but do not replace testing on physical phones, font settings, zoom levels, slow networks, or enterprise browser policies.

Data coverage: a render pass does not prove that a public source was complete, that a translation is legally authoritative, or that a company has no unreturned records. The sample report uses fictitious data and tests presentation only. For fields such as annual reports, preserve the reporting year and field date using the annual-report reading method.

Decision coverage: the report is decision support. Testing cannot guarantee identity, solvency, product quality, performance, authority to contract, a bank account, or future conduct. Higher-value decisions may still require original documents, independent confirmation, inspection, audit, contract advice, or local professional review.

Version record and retest triggers

Method v1 — 15 July 2026: recorded the three Chrome viewports, DOM and network assertions, data-state checks, snapshot timestamps, PDF V2 boundary, and known browser limitations. The PDF version and the method version are separate labels.

Retest is required when the report data contract changes, a public-record module is added or renamed, identity fields move, the PDF renderer changes, the supported browser version changes materially, or a layout fix affects shared report components. The result should be recorded as a new dated method entry rather than silently overwriting what an earlier release actually covered.

To inspect the output used in this method, open the sample report walkthrough or run a company lookup in ChinaValidate Company Search.