Why Our Robot Directory Dropped 46,000 Records
Published: 2026-10-20 10:30 (UTC+8) | Data as of: 2026-10-20 | Source: Zhilianhai robot company directory
What changed
The directory now holds 587,277 records, down from 633,635. That is 46,358 duplicates removed, or 7.3 percent of the old total. The public counter reads 587,737 including 460 reviewed profiles.
Where duplicates came from
None of them were different companies. They differed only in characters that look identical on screen:
- Full-width versus half-width brackets: one record uses Chinese brackets, the other ASCII brackets.
- Full-width spaces: a single invisible space in the middle of a company name.
- Mixed bracket styles: square brackets from different source exports.
How records are matched
Names are normalised first: brackets and spaces are converted to one form, whitespace is removed, letters are lowercased. Records inside a group are then reduced to one survivor, chosen by completeness score, then field richness, then the lower id so the result is reproducible.
What this means for the numbers
Category and region counts were recalculated on the deduplicated set. For example industrial robots read 137,543 instead of 150,645, and Guangdong reads 88,523. Nothing disappeared; the same company simply stops being counted twice.
What we did not merge
Branch offices, subsidiaries, short names versus legal names, and separate legal entities inside a group remain separate records. Those differences carry business meaning and should not be collapsed automatically.
How to read the directory
Use it to narrow a candidate pool, then verify each shortlisted company on its own website and public project records. Inclusion is an index, not a certification.
Browse the full directory or read how records are collected.