Measurement recordTwo sources, 17 selected records
Registry vs database: 15 of 17 status labels differ
Different words can describe the same company status. In 17 selected records, 15 status labels differed in wording but mapped to the same broad class. Compare the labels before treating a mismatch as a problem. The full records were not proved equal.
1. What the comparison rules matched
Seven fields across 17 records gave 119 cells. One branch capital cell was marked not applicable, leaving 118 comparisons. The original design set the seven field rules; the branch exception was added when that record was encountered.
| Field compared | Rule matched | Rule not matched |
|---|---|---|
| Registration status, normalised | 17 of 17 | 0 |
| Company type | 17 of 17 | 0 |
| Establishment year | 17 of 17 | 0 |
| Legal representative present | 17 of 17 | 0 |
| Domicile present | 17 of 17 | 0 |
| Business-scope length | 17 of 17 | 0 |
| Registered capital | 13 of 16 | 3 |
13 of 16 capital entries matched on amount and currency. In the other 3, the saved registry text omitted a currency while the database named one; the numeric amounts matched. A matching number without a currency does not prove monetary equivalence. The branch cell was excluded.
The presence checks cover 34 cells, and the scope-length checks cover 17. Matching presence does not show the names or addresses are the same. Similar length does not show the scope wording is the same.
2. What a buyer would actually see
All 17 saved status pairs mapped to the study’s broad existing class. The wording counts below describe these records. They do not establish a supplier’s current condition.
| Official registry wording | Commercial database wording | Companies |
|---|---|---|
| 在营（开业）企业 | 开业 | 9 of 17 |
| 存续（在营、开业、在册） | 存续 | 6 of 17 |
| 存续 | 存续 | 2 of 17 |
2 of 17 status pairs matched word for word; 15 did not. Those 15 mapped to the same class under our rule. That count measures text differences. It is not a tested error rate.
A wording difference is a reason to inspect the field and its source. It does not by itself establish a change in status or a problem with the supplier.
First match the Chinese legal name and registration code to your documents. Then compare each field using its meaning, source and date. A code-format check alone does not confirm the entity.
3. Method, and what this study cannot say
- Frame: 20 entries were drawn with a fixed seed from the 46 inputs in our registration-field study. Collection stopped at 17; no replacements were added. The upstream filter required one name-search candidate and a non-empty code field. It did not validate identity. That source frame contained 45 strings of 18 characters and one of 8 characters.
- Registry side: the collection notes describe manual searches in the National Enterprise Credit Information Publicity System, with the owner handling CAPTCHA and fields read from rendered pages. This review uses the saved observations; it does not repeat or independently verify those visits.
- Database side: the commercial record captured on 28 August 2026 for the same codes, already published as our registration-field dataset. One database. This says nothing about commercial databases in general.
- Comparison rules: status mapped to five broad classes; type matched after spaces and parentheses were normalised; capital matched on amount and currency; establishment matched on year. Representative and address matched on presence only. Scope lengths could differ by up to five percent. These rules do not compare full names, addresses or scope content.
- Different dates: the saved approval dates were all before 28 August. Measured from 5 September, their median age was 449 days and the minimum was 53 days. Approval dates cannot rule out updates or source lag. Change-history observations were empty for all 17 records.
- Saved evidence: the public file omits company and natural-person names. It retains codes, comparison verdicts, status and capital wording, approval dates and scope lengths. The internal observation URLs are truncated; they cannot independently reproduce the source pages.
- Limits: 17 selected records, one commercial source, seven fields and two dates. This is not a representative accuracy study. Transcription errors are possible. The checks do not establish supplier safety, complete record equality or current status.
What the branch record showed
One saved record was marked as a branch. Both capital entries were empty, and the comparison treated that cell as not applicable. The capital denominator is therefore 16.
The collection notes describe a person-in-charge field and a place-of-business field for that entry. The saved comparison only records whether those fields were present. It does not retain their contents.
Check the entity type and the applicable field before treating an empty value as a data gap. This one entry does not establish a rule for every branch or its liabilities.
Questions about the comparison
Does a commercial database match the official registry?
In these 17 selected records, 115 of 118 checks matched under limited rules. Some checks measured only presence or text length. This does not prove the full records were the same, or show how either source performs for your supplier.
Are different status labels false alarms?
The 15 different label pairs mapped to the same broad class in this study. We did not test real alerts or measure whether they were right or wrong. Read each label with its source and date.
Can approval dates rule out database lag?
No. All saved approval dates predated the database reading, but change-history observations were empty. These dates cannot rule out updates or source lag between the two readings.
Why was one capital field excluded?
One saved branch record had no capital value on either side. It was treated as not applicable, so the capital comparison used 16 records. This observation does not establish a rule for every branch.
Citing this study
Archived independently of this site, with its own DOI: figshare. If this page ever disappears, the measurement does not.
A citation with everything it needs:
Currawong, “Registry vs database: 15 of 17 status labels differ”. Registry observations dated 5 September 2026 and commercial records dated 28 August 2026; interpretation corrected 8 September 2026.
https://currawongweb.com/verify/company-database-vs-china-registry/
BibTeX
@dataset{currawong_verify_registry_versus_database_study_2026,
author = {Bao L. Zhou},
title = {{Official Chinese registry records compared with a commercial database: 17 companies, field by field}},
year = {2026},
publisher = {figshare},
doi = {10.6084/m9.figshare.33438322},
url = {https://doi.org/10.6084/m9.figshare.33438322}
}You may reuse the counts with both dates, the sample size and comparison limits. The historical dataset and 28 August baseline remain unchanged. Their older framing should be read with this correction: limited matches do not prove full factual equality, alert accuracy or the absence of updates.
NEXT CHECK
Compare the status wording and the query date.
Read the Chinese status labels before treating different words as a conflict. Keep the source and date for both records; similar wording does not prove that the full records agree.
For your own comparison, start with the registry-check guide and keep the source and date beside each field.