Pick a depth. Each prompt opens in your AI pre-loaded with the lesson. Click a row to preview the prompt.
The single most common mistake for beginners is opening the spreadsheet first and 'exploring.' You end up making six charts, feeling busy, and answering nothing - because with no question, every column looks equally interesting and none of them matters. Writing the question first flips this: it tells you which columns you need, which you can ignore, and how you will know when you are done. It also protects you from the seductive trap of the number that is easy to compute but does not settle anything. This habit costs you two minutes and saves you an afternoon, and it is the difference between walking away with an answer versus walking away with a mess of tabs.
A good question has four parts, and you can test any question against them. Take a coffee shop owner's vague worry: 'I feel like weekends are dead.' That is not answerable. Now sharpen it: 'On average, does a Saturday bring in less revenue than a Wednesday, over the last three months?' Test it: (1) It names a specific metric - average revenue per day. (2) It has a comparison - Saturday versus Wednesday. (3) It has a time frame - last three months. (4) The answer would change a decision - if Saturdays are 40 percent lower, maybe cut Saturday staff or run a weekend promo. Compare that to a weak version: 'How is business doing?' - no metric, no comparison, no time frame, no decision attached. Here is a worked before-and-after you can copy. Before: 'Are our customers happy?' After: 'Did our average review rating drop after we changed suppliers in March, comparing the three months before to the three months after?' The 'after' version tells you exactly what to pull: review dates and ratings, and the supplier-change date as the dividing line. Everything else in the dataset - customer names, order IDs - you can ignore. That is the whole point: the question is a filter for your own attention.