A prompt is user-provided input that directs a generative model toward expected content, analysis, format, language, or behavior. In analytics, prompts turn broad requests into specific instructions for descriptive statistics, predictive modeling, code generation, or stakeholder recommendations. section 12.4.1 section 12.4.4
Effective prompts make the requested work and output constraints explicit:
- Name the desired task, such as describing a dataset, calculating statistics, or creating a predictive model. section 12.4.1 section 12.4.4
- Specify the required output, including performance metrics, Python code, recommendations, format, language, and relevant limitations. section 12.4.4
- Refine the instructions iteratively when the initial response is too broad or fails to support stakeholder understanding. section 12.4.4
From broad to specific
“Analyze the attached dataset” leaves the requested analysis unspecified. A more precise prompt asks for the mean and median of login_days and mean login_days by region, producing directly usable metrics.
Analytics exercise
Upload rats_weather.csv, ask for a dataset description and a model predicting daily rat sightings, require documented steps and performance metrics, then compare the result with an existing analysis. section 12.4.4
