Once you've established that the data is relevant and from a credible source, you must assess its internal quality.
Accuracy: How well does the data reflect the real-world phenomena it is supposed to measure? This can be difficult to assess without external validation, but documentation often provides details on how the data was checked for accuracy.
Completeness: Are there significant amounts of missing data? A high percentage of missing values for a key variable can be a serious problem. You should investigate why data is missing and if there are patterns that could introduce bias.
Consistency: Are the data values consistent? For example, are dates formatted uniformly? Are categorical variables (e.g., "male," "female") coded consistently? Inconsistencies can make analysis difficult or even impossible without extensive cleaning.