What if, instead of requesting a new report or a revision to one of your dashboards, you could just type your questions into ChatGPT or Copilot? This is relatively easy to accomplish (See: ), but once done, you need to learn to ask questions that begin with “What,” not “Why.” Let’s explore.
Once your tech stack is fully connected you type, “Why did sales increase last quarter?” expecting a revelation. But, alas… you get a generic answer filled with jargony clichés. This is to be expected. The “data” doesn’t know anything. Neither does the large language model (LLM) you’re using as a natural language interface. When analyzing data in a database, it’s important to understand the distinction between “why” questions and “what” questions.
What vs. Why
“What” questions are typically more appropriate for data analysis because they focus on observable patterns, trends, and relationships within the data. In contrast, “why” questions often seek to understand the underlying reasons or causes behind the observed data, which may not be directly answerable from the data itself.
So, instead of asking the “why” question: “Why did sales increase last quarter?” You should begin with a “what” question such as: “What was the percentage change in sales between the previous two quarters?”
The difference between these two queries is profound. The “why” question seeks to understand the reasons behind the increase in sales, which may not be readily apparent from the data alone. The “what” question, on the other hand, focuses on quantifying the change in sales, which can be directly calculated from the data.
What If I Don’t Know “What” To Ask?
Transforming “why” questions into actionable “what” questions requires a bit of practice. You can start by breaking down your question into these six steps:
1 – Identify the Core Issue: Understand the underlying problem your “why” question is trying to address.
2 – Break Down the Issue: Deconstruct the core issue into smaller, measurable components.
3 – Formulate Specific Questions: For each component, formulate specific “what” questions that can be directly answered with data.
4 – Consider Contributing Factors: Identify potential factors influencing the issue and turn these into “what” questions about their nature, extent, and impact.
5 – Focus on Actionable Insights: Ensure each “what” question is structured to elicit information that can lead to actionable insights.
6 – Review and Refine: Evaluate the “what” questions for clarity, specificity, and relevance to the core issue. Adjust as necessary to ensure they are focused and actionable.
Following these steps will help you translate broad, exploratory “why” questions into targeted “what” questions that drive towards clear, actionable data analysis.
Of course, if you still need more help, you can ask your AI-powered chat app to help you formulate your questions.
Here’s an .xlsx file filled with demo data. .
Now upload it to ChatGPT, Claude, Gemini, Copilot or almost any chat interface into an LLM with the following prompt: “What can I learn from this data?”
If you type this query into ChatGPT for Teams (paid version), you are going to get a response that looks something like this:
Based on the provided data, which includes customer ID, age, gender, annual income, and spending score, you can learn a variety of things related to customer demographics and behaviors. Here are a few insights and analyses that can be derived:
1. Demographics Analysis:
- Age distribution of customers and its relationship to spending habits.
- Gender distribution and if there are notable differences in spending habits or income levels between genders.
2. Income Analysis:
- Distribution of annual income among your customer base.
- Correlation between annual income and spending score to identify if higher income leads to higher spending within your business.
3. Spending Behavior:
- Identification of high-value customers based on their spending score and annual income.
- Spending score distribution to understand how customers are engaging with your products or services.
4. Segmentation:
- Clustering customers into meaningful segments based on their demographics and behaviors. For example, young high earners with high spending scores could be one segment, while older, more conservative spenders could be another.
- Identifying target segments for specific marketing campaigns or product offerings.
5. Predictive Modeling:
- Using the data to predict spending scores or segment membership based on other variables, which can be valuable for personalized marketing or service offerings.
6. Trend Analysis:
- If this data is part of a time series, analyzing trends over time in income, spending habits, or demographic shifts within your customer base.
7. Gender-Based Analysis:
- Examining if there are significant differences in income levels, spending scores, or other behaviors between males and females.
Each of these analyses could provide valuable insights for strategic planning, marketing, product development, and improving customer service. Would you like to explore any specific analysis or insight further?
Then you can ask ChatGPT to do these various tasks in whatever order suits your needs.
The Power of What
LLMs are not mystics, they are not human, they don’t “understand” the way humans do. And, while some LLMs do appear to be able to reason through problems, this is not yet a reliable tool for the kind of tasks we’re talking about here. Learning to harness the “Power of What” in your prompts will not only give you a personal competitive advantage in your day to day decision-making. It will start you on a path to data-driven super powers.
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.
ABOUT SHELLY PALMER
Shelly Palmer is the Professor of Advanced Media in Residence at Syracuse University’s S.I. Newhouse School of Public Communications and CEO of The Palmer Group, a consulting practice that helps Fortune 500 companies with technology, media and marketing. Named he covers tech and business for , is a regular commentator on CNN and writes a popular . He's a , and the creator of the popular, free online course, . Follow or visit .
