The same name, different meanings
In one spreadsheet, a sale means payment; in another, shipment; in a third, a signed contract. Combining them under one column name conceals three different events. Data-quality checks start by establishing what each record measures.
An example where every number is correct
Imagine two online-shop reports. One counts a sale after payment, the other after shipment. Month-end totals differ although both exports are technically correct. Combining them without defining the event creates a convincing chart with a wrong interpretation. Quality starts by asking what a row represents, its period and the rule for inclusion.
Record the observation unit and key: order, line item or payment. Check that joins do not multiply rows. One order may have several payments and deliveries; a simple join can inflate totals. Tracing concrete objects from source to report reveals this sooner than debating unusual growth in a final chart.
Quality depends on the decision
Daily shipment planning needs current status, assortment analysis needs stable categories, and mailing needs correct contacts and permitted use. A general 95% quality score explains little without checks. Tie requirements to the task: mandatory fields, allowed values and the delay that makes a record unusable.
Checks can cover completeness, consistency, uniqueness and currency. An empty column does not automatically need filling. Missing values may truthfully express the unknown. Substituting something similar tidies the table while hiding the problem. Preserve distinctions between observed facts, calculated values and assumptions, especially before feeding data to AI.
From one-off cleanup to monitoring
After correction, establish the error's origin. If a field is entered manually without a definition, cleaning the next export repeats the work. Improve input rules, reference tables or checks between systems instead. Each important metric needs an owner who can explain its calculation and interpret exceptions. Technical staff alone cannot define that meaning.
