This page explains where our data comes from, how we process it, and where it falls short. Our articles and datasets rely on it.
Sources
- Procurement: the official procurement and tendering portals of 111 Chinese financial institutions: banks, insurers and securities firms. We collect tender notices, award (result) notices, failed-tender notices, amendments and supplier calls.
- Hiring: the official career sites of 114 financial institutions, including Chinese and global banks.
All sources are publicly accessible web pages. We do not use leaked, purchased or non-public information.
Collection
An automated collector takes periodic snapshots of each source. For every record we keep the title, publication date, source URL, the full text where the portal provides it, and the dates we first and last saw the record.
Processing
- Deduplication: records are keyed by institution and the portal’s own notice ID. Re-crawled records whose content is unchanged are not counted twice.
- Classification: each notice is tagged by spending category and by technology track (AI and large models, compute, domestic substitution, fraud and identity, data governance). We use rule-based keyword matching and check samples by hand.
- Awards: we extract winning suppliers and award values from result notices where the text discloses them. Coverage varies by portal and is stated in each dataset.
- Translation: titles are kept in the original Chinese with an English translation. Names of institutions and companies are given as published.
Known limitations
- Uneven coverage. Some portals publish full notices. Others expose only a title. Where the text is missing, we cannot confirm requirements, winners or values.
- Gaps over time. Some portals stop updating or change structure. We flag institutions with no recent records rather than treating silence as zero activity.
- Notices are not projects. One project can produce several notices: tender, amendment, failure, re-tender, award. Unless stated otherwise, counts refer to notices.
- Linking. Many portals do not link a result notice to its tender. Where we link them, we say how.
- Hiring dates. Career sites often omit posting dates. For job postings, dates usually mean the date we first saw the posting.
Corrections
If you find an error, email info@it4finance.info with the link and the correction. We fix confirmed errors and note material corrections in the article.