Level 6 · Boss: data quality sweep

Data quality checks are a big part of an analyst’s week, especially before a close or a board report. They are also a great way to practice filtering, because each check is just a precise WHERE clause.

Step 1: active products with no cost

An active product without a cost breaks every margin report. Two conditions, both required:

WHERE is_active
  AND unit_cost IS NULL

Inactive products are excluded on purpose: they are no longer sold, so Kevin does not need them in the clean-up list. Remember from the quiz that unit_cost = NULL would quietly return nothing and make the data look cleaner than it is.

Step 2: impossible quantities

A sold line with a quantity of 0 or less is a data error: maybe a return keyed as a negative sale, maybe a test transaction. WHERE quantity <= 0 finds both zero and negative values. In later chapters, when you compute revenue, you will exclude these rows with quantity > 0 and say so in your notes. Analysts who silently include them report revenue that does not match finance.

Step 3: pending orders in 2025

Three conditions: status, a pair of channels, and a year. Marketplace orders are handled by the marketplace partner, so Kevin only wants web and app.

WHERE status = 'pending'
  AND channel IN ('web', 'app')
  AND order_date >= '2025-01-01'
  AND order_date <  '2026-01-01'

Using IN avoids the AND/OR precedence trap. If you wrote channel = 'web' OR channel = 'app' without parentheses, pending-status and date filters would only apply to the app half, and the list would include every web order ever placed.

Why a sweep before the close

Each of these lists feeds a different fix. Missing costs go to merchandising, who own supplier pricing. Impossible quantities go to engineering, because something in checkout or the import job wrote them. Stale pending orders go to operations, who decide whether to ship or cancel. Splitting the results by owner turns one vague “the data is messy” into three tickets that actually get done.

Reporting back

When you send results like these, include the count and the rule you used, for example: “8 active products have no cost recorded (is_active = true and unit_cost IS NULL)”. Stating the rule lets the reader catch a misunderstanding before anyone spends a day fixing the wrong rows.