Analyses are only as good as their data. A few checks catch most problems before they reach decisions.
Before Analysis
- Row counts: do they match expectations and source systems?
- Date ranges: is the period complete, with no missing days?
- Duplicates: are there repeated records?
- Missing values: which columns, and how many?
- Value ranges: negative quantities, impossible dates, outliers?
- Categories: unexpected values or spelling variants?
After Joins
- Did row counts change unexpectedly? Many-to-many joins inflate numbers.
- Did records drop out of inner joins?
Reconciliation
Compare totals with trusted reports — finance figures, official dashboards. Explain differences.
Definitions
Confirm what each field means. "Revenue" might include tax, refunds or cancelled orders.
Document
Record the checks performed and data issues found, so others can trust and reproduce the analysis.
Automate
For recurring analyses, build these checks into queries or pipelines.