Data Credibility Framework
Policy Documents
Data Methodology
Comprehensive methodology and quality control standards for the end-to-end data pipeline: collection, cleaning, validation, and analysis.
Last updated: 2024-11-18Version: v2.1
1Data Collection
1.1 Data Source
- Official channels:Government open data, regulatory agency announcements, public company financial reports, patent databases
- Industry Channels:Industry association reports, research institution publications, and conference presentation materials
- Media Channels:Authoritative financial media, technology media, and industry-specific vertical media
- Owned Channels:Jingzhi Guangnian Industrial Monitoring Network & Partner Data Sharing
1.2 Collection Standards
Timeliness requirements:
- Funding Event: Collected within 24 hours of occurrence
- Policy Documents: Collected within 48 hours of publication
- Industry Report: Collected within 72 hours of publication
- Corporate News: Collected within 7 days of occurrence
Completeness requirement:
- Each record must include: Title, Source, Publish Date, and Core Content.
- Financing data must include: company name, funding amount, round, and investors.
- Company data must include: founding date, headquarters location, core business, and team background.
2Data Cleaning
2.1 Deduplicate and Merge
- Remove Duplicate TitlesIdentify duplicate records based on title similarity > 85%
- Remove duplicatesDuplicate content detection via fingerprinting
- Merge Entities:Map different names of the same enterprise to a unified entity ID.
2.2 Quality Check
- Integrity Check:Required field missing rate < 5%
- Consistency Check:Number range, date format, and enum validation
- Anomaly Detection:Outlier Detection and Manual Review
3Data validation
3.1 Multi-source Cross-Validation
A-Level Data (High Confidence):
- Cross-verify from at least 3 independent sources
- Manual Review Confirmation
- Traceable raw data
- No Conflict of Interest Statement
B-level data (medium confidence):
- Verified by at least 2 independent sources
- Manual Review Confirmation
- Verified source (official or authoritative media)
C-level Data (Reference):
- Single Source Validation
- Auto-collect + Basic Quality Check
- Further verification required
D-tier data (pending verification):
- Unverified
- May vary
- For reference only, not for official reports
3.2 Manual Review Process
- Researcher Initial Review: Verify data integrity and plausibility
- Senior Researcher Review: Validate data sources and logical consistency
- Quality Supervisor Spot Check: Randomly sample 10% of data for in-depth verification.
- Flagged Data Marking: Data with issues is marked as "Pending Review" and enters the correction workflow.
4Data Analysis
4.1 Analysis Method
- Quantitative Analysis:Statistical Analysis, Trend Analysis, Comparative Analysis, Regression Analysis
- Qualitative Analysis:Case studies, expert interviews, policy interpretation, and competitive landscape analysis
- Hybrid approach:Combining quantitative and qualitative methods for mutual validation and complementarity.
4.2 Quality Control
- Methodology Transparency:Disclose all methods, assumptions, and limitations
- Reproducible results:Provide sufficient methodological details to enable replication of the analysis results.
- Uncertainty Note:Clearly state the confidence interval and applicable boundaries of the conclusion.
- Peer Review:Undergoes internal peer review before the release of important reports.
5Publish Data
5.1 Publishing Standards
- CompletenessAll required fields are complete; no critical information is missing.
- Accuracy:Data validated; error rate < 1%
- Timeliness:Data is valid; expired data is clearly marked.
- Traceability:All data is traceable to its original source.
6Data Correction
6.1 Correction Workflow
- Feedback Received:Receive data correction feedback via website, email, or phone
- Initial Assessment:Evaluate the reasonableness and urgency of feedback within 24 hours
- Investigate and verify:Data verification completed within 3 business days.
- Update Release:Correct and publish a correction notice within 24 hours after confirming the error.
- Feedback Loop:Notify the feedbacker of the resolution.
Contact Us:If you find any data errors, please email data@joyhiai.com or submit feedback through the website's "Data Correction" channel.
