Data Prompts

Welcome to “Data Prompts.” Today, we’re looking at a simple idea with surprisingly wide reach: the questions and instructions we use to work with data. Whether you’re asking an AI tool to summarize a spreadsheet, inviting a team to explore a trend, or deciding what to measure in the first place, the prompt shapes what you notice—and what you might miss.

First, a useful data prompt starts with a clear goal. “Tell me about these sales” leaves a lot open: which sales, over what period, and what counts as useful? A stronger prompt might ask, “Compare monthly sales by region for the past year, identify the three largest changes, and note any missing months.” The added details give the analysis direction without pretending there’s only one possible answer. Before you ask a question, try finishing this sentence: “I need this information so I can…” That purpose can help you choose the right prompt.

Second, provide context and define the terms that matter. A column called “active” could mean a customer who logged in this month, made a purchase this quarter, or simply has an open account. If the meaning is unclear, different people—or different tools—may reach different conclusions. Share what the data represents, how key fields are defined, and any relevant time period or audience. If you’re using an AI assistant, include the data itself only when appropriate, and remove sensitive details unless you’re working in a setting approved for them.

Third, ask for a process you can check, not just a polished result. Instead of requesting a confident-sounding conclusion, ask the tool to show its assumptions, explain how it grouped the data, and flag gaps or anomalies. You might say, “Summarize the trend, distinguish observations from possible explanations, and list what additional information would help confirm the interpretation.” That wording encourages useful caution. It also reminds us that a pattern in the numbers isn’t automatically proof of what caused it.

And fourth, make room for follow-up. Data work is rarely a one-question exercise. A first prompt can reveal an unexpected spike; the next can ask whether it appears across all regions or only one. Then you might check the source, compare the same period last year, or look at a different measure. Treat each answer as a starting point. When the result seems surprising, verify it against the original data and use a second method where practical.

Good data prompts don’t need to sound technical. They need to be specific, honest about uncertainty, and connected to a real decision or curiosity. State what you want to learn, provide the context that gives the numbers meaning, and ask for evidence you can inspect. Then refine the question as you learn more. The better the prompt, the more useful the conversation with your data—and the more likely you are to come away with insight rather than just a tidy answer.