A dataset small enough to check

Start every data task with a set small enough that you can do the sums in your head. Not because the tool can't handle more, but because you can only learn how much to trust it on work you can independently check — and on a hundred rows, you can't.

item,quantity,unit_price
tables,2,3
chairs,4,2
boxes,3,5

Ask what the columns are before asking for a calculation. Is unit_price per item or per pack? Is the currency stated anywhere? It isn't — and a total that arrives with a currency symbol has invented one. Establishing what the data means is a separate step from computing on it, and skipping it is how a confident total ends up measuring the wrong thing.

Then the line totals and the grand total: 2×3 = 6, 4×2 = 8, 3×5 = 15, giving 29. Do that yourself and compare. The habit you're building isn't distrust of arithmetic — it's the reflex of having an independent answer before you look at the provided one, which is the only way to notice a plausible-but-wrong result.

Now break it deliberately. Make a copy with one missing quantity and one duplicated row, and ask for problems to be flagged. What you're testing is whether they come back as reported — a missing quantity is not a zero, and a duplicate row is not something to silently drop. Either fix might be right, and both are your call.

Cleaning rules belong to you too. Add ragged capitalisation and stray spaces, then define the rule explicitly: trim whitespace, title-case the item names, and keep the original column beside the cleaned one so the change is visible. A cleaning step that overwrites its input is a cleaning step nobody can review.

Loading your workspace…