Start with the PDFs you already have.
Already have a PDF? Standard extraction from RMB 30 per document.
ClariRecord extracts data from Hong Kong company filings, either from your existing PDFs or alongside filing retrieval. Specify the document types, volume, required fields and target format. Agree the sample and quotation before proceeding with the batch.
Choose your delivery format
| Format | Workflow | Details to agree |
|---|---|---|
| Excel | Manual review, client files and working papers | Column order, date format and source pages |
| JSON | Import into existing systems and databases | Field names, data types, arrays and nesting |
| XML | Existing data interchange workflows | Element names, nesting, encoding and null handling |
Match each field to its source.
The NAR1 sample covers company names and number, business name, company type, return date, accounting period, address, email, mortgages and charges, member count and share capital. On the sample page, select a field to locate its source, or switch to JSON, XML or custom fields and download an example.
The sample covers the fields in a two-page excerpt of ARCH MARKETING LIMITED’s annual return, made up to 2 August 2022. The legacy company number must not be treated as a Business Registration Number.
View the sample, source and coverage
NAR1 field guide and free JSON Schema →
Keep your existing fields and codes
Company names can map to your internal fields; director roles and share classes can follow your code list. Code meanings and the treatment of missing or unreadable values are agreed before batch delivery. Standard JSON / XML is included in Data and Reviewed; custom fields and nesting are quoted separately.
What needs further verification?
A filing records information at the time of filing. A historical NAR1 does not establish the company’s current directors, shareholders or trading status. For up-to-date work, confirm the search date and document scope. Check illegible text, missing fields and inconsistencies against the source.
For search options and available information, see the Companies Registry’s e-Search FAQs.
Check one filing before processing the batch.
Source starts at RMB 39 per document; Data at RMB 59 includes standard JSON / XML; Reviewed at RMB 99 adds manual checks of agreed fields. All include one standard source document. Data supports up to 30 pages; Reviewed up to 10, including continuation sheets. Supported versions and fields are confirmed before acceptance. Custom and out-of-scope work is quoted separately.
From reading the PDF to using its data
Use our public two-page NAR1 excerpt to separate three tasks: identify the company and information date in the PDF, organise the values needed, then compare the result against the source and intended use. This distinguishes selecting the wrong filing from transcription errors or misunderstood fields.
The excerpt’s legacy company number is 1063959. Retaining the digits is only part of the task: retaining what the number represents matters too. Copying those digits into a field labelled BRN could mislead a recipient despite a correct transcription. That is why it helps to check what the fields mean, not just how many have been extracted.
Agree what you need before we begin
Use a representative document to agree essential fields and acceptable output. For example: “I need the company name, information date and share-capital rows, traceable to source pages; do not guess unavailable information.” Send your actual document with the request so we can confirm which fields are supported and what work is needed.
Match receipt checks to the order. For retrieval, check the PDF’s company, date and type. For extraction, confirm the output refers to that PDF and covers the agreed fields. For review, agree which fields will be checked and how unresolved items will be handed over. Opening the files successfully is only one check.
Only a company name, an existing PDF, or an Excel template?
These are fictional enquiries. A has only a company name and needs a historical filing: agree retrieval first. B already has the complete PDF and wants standard data: supply it with the field requirements for a scope check. C must populate a customer’s Excel workbook: include a blank target and an example, since preparing that workbook is a separate task.
If C’s target includes “customer risk rating”, identify who supplies that assessment and on what basis. A judgement absent from the source is not automatically part of extraction. Separate source transcription, transformations under your agreed rules and decisions still owned by your team.
Found something wrong? Show us where to look
If something looks wrong, you do not need to work out the answer before asking for help. Send the file name, page and value that concerns you. For example: “File A, page 2, first capital row: the PDF says HKD, but the result shows a different currency.” That gives us a clear place to start checking with you.
Keep the file you first received and save any corrections as a new version. That way, you and your colleagues can see what changed. Before ordering, also ask how to report a problem and how corrected files will be returned.
See how the data fits your workflow
Check the source, adjust field names and download the result.
Start with your documents
If you have a company list, tell us which documents and years you need. If you already have PDFs, keep the originals and let us know whether they are complete files or excerpts. Mention any preferred format or need for human review too.
For your first enquiry, choose one document and use the public sample to show which information and formats would help you. We will then confirm what we can process from your file. The sample covers two pages; tell us if you also need directors, shareholders or other details not shown.