Understanding Data Quality: How IQ Scores Work
Learn how DataSetIQ calculates quality scores for 50,000+ datasets. Our methodology explained, with tips for interpreting scores.
Why Data Quality Matters
Bad data leads to bad decisions. A 2023 Gartner study found that poor data quality costs organizations an average of $12.9 million per year. Yet most data platforms give you no quality indicators at all.
DataSetIQ solves this with IQ Scores—a 0-100 rating for every dataset.
The IQ Score Components
Each IQ Score is calculated from four weighted dimensions:
1. Freshness (30% weight)
How recently was the dataset updated relative to its expected frequency?
| Update Status | Score Impact |
|---|---|
| On schedule | +30 |
| 1 period late | +20 |
| 2+ periods late | +10 |
| Discontinued | 0 |
Example: A monthly CPI dataset updated yesterday scores 30. The same dataset last updated 3 months ago scores 10.
2. Completeness (25% weight)
How many gaps or missing values exist in the time series?
| Gap Percentage | Score Impact |
|---|---|
| 0% gaps | +25 |
| <5% gaps | +20 |
| 5-15% gaps | +15 |
| >15% gaps | +10 |
Example: GDP data with no missing observations since 1960 scores 25. A survey with 10% missing quarters scores 15.
3. Reliability (25% weight)
Is the source authoritative? How often is the data revised?
| Source Type | Score Impact |
|---|---|
| Official statistics agency | +25 |
| Central bank / ministry | +22 |
| International organization | +20 |
| Academic/research | +15 |
| Private provider | +10 |
Example: BLS employment data scores 25. A private estimate scores 10.
4. Usability (20% weight)
How clean and accessible is the data format?
| Data Quality | Score Impact |
|---|---|
| Clean, consistent format | +20 |
| Minor format issues | +15 |
| Requires cleaning | +10 |
| Major issues | +5 |
Example: FRED data with clean CSV export scores 20. A PDF-only source scores 5.
Score Interpretation
| IQ Score | Quality Level | Interpretation |
|---|---|---|
| 90-100 | Excellent | Production-ready data |
| 80-89 | Good | Reliable with minor caveats |
| 70-79 | Fair | Usable, verify carefully |
| 60-69 | Limited | Check for specific issues |
| Below 60 | Poor | Use with caution |
Real Examples
FRED CPI-U (Consumer Price Index)
IQ Score: 97
- Freshness: 30 (monthly, on schedule)
- Completeness: 25 (no gaps since 1947)
- Reliability: 25 (BLS official source)
- Usability: 17 (clean but complex structure)
Experimental Survey Data
IQ Score: 62
- Freshness: 15 (irregular updates)
- Completeness: 12 (20% missing values)
- Reliability: 15 (academic source)
- Usability: 20 (clean format)
How Scores Update
IQ Scores are recalculated:
- Daily for freshness checks
- Weekly for completeness audits
- Monthly for full recalculation
When you see a dataset, you're seeing the current quality, not a stale rating.
Using IQ Scores in Research
Best Practices
- Filter by 80+ for production models
- Compare similar datasets before choosing
- Check individual components if overall score is borderline
- Note score trends - declining scores signal problems
Red Flags
- Score dropped 10+ points recently
- Freshness score below 10
- Completeness under 70% on a mature series
Questions?
Our methodology is transparent. If you disagree with a score or spot an issue, contact us with the dataset slug and we'll review.
Related Articles

7 Economic Indicators That Actually Predict Recessions (Backtested Across 50 Years)
Most recession indicators sound convincing but fail real-world testing. These seven indicators actually predicted every U.S. recession since 1970 — with clear lead times and minimal false alarms. Learn the exact framework used by macro hedge funds and central banks.
How to Find Inflation Data Faster than FRED
FRED is slow and clunky. DataSetIQ normalizes inflation data instantly—find CPI, PCE, and breakeven rates in 3 clicks instead of 20+.
GDP Data: 5 Sources Compared (2025 Edition)
World Bank, IMF, BEA, OECD, or Eurostat? We compare accuracy, timeliness, and coverage of the top GDP data sources to help you choose the right one.
