Planning Your Data Model with ChatGPT and Claude

Between cleaning your data and building charts, there is a quiet step that separates confusing reports from clear ones: the data model. A data model is simply how your tables relate to each other, for example, linking a Sales table to a Products table so a chart can show revenue by product category. Get the model right and DAX becomes easy. Get it wrong and even simple measures give strange numbers.

You do not need database theory to do this well. You need to describe your tables to an AI assistant and let it recommend a sensible structure. In this lesson you will learn what a data model is, why the "star schema" is the beginner-friendly gold standard, and exactly how to have ChatGPT or Claude design one with you.

What You'll Learn

• What a data model and a relationship are, in plain terms
• Why a star schema of facts and dimensions keeps reports simple
• How to have AI design your model from your table descriptions
• How to create relationships in Power BI and check they are right

What a Data Model Actually Is

When you load more than one table, say Sales and Products, Power BI needs to know how they connect. They connect through a shared column, usually an ID. If both tables have ProductID, Power BI can link them: each sales row points to a product row. That link is a relationship. Once it exists, you can put ProductName from the Products table next to Revenue from the Sales table on the same chart, and Power BI knows how to combine them.

Without a relationship, Power BI treats the tables as strangers and your charts break or show the same total repeated for every row. Relationships are the plumbing that makes everything else work.

The Star Schema, Explained Simply

The most reliable structure for beginners is a star schema. It has two kinds of tables:

• Fact tables hold the events you measure: sales, orders, clicks. They have lots of rows and numeric columns like Quantity and Revenue.
• Dimension tables hold the descriptive things you slice by: products, customers, dates, regions. They have fewer rows and descriptive columns like ProductName, Category, or Country.

You put the fact table in the middle and connect each dimension table to it, forming a star shape. This keeps filters flowing in one clear direction (a dimension filters the facts) and makes DAX predictable. Nearly every good Power BI report is built this way.

A special and important dimension is the Date table. Time calculations, year-over-year, month-to-date, running totals, all rely on a dedicated Date table connected to your fact's date column. Ask AI to help you create one; it is a common first step.

Let AI Design Your Model

Here is where AI shines. You describe your tables and let it propose a model. Paste this into ChatGPT or Claude:

"I'm building a Power BI report and I'm a beginner. Here are my tables and columns:

• Sales: OrderID, OrderDate, ProductID, CustomerID, Region, Quantity, Revenue
• Products: ProductID, ProductName, Category, Cost
• Customers: CustomerID, CustomerName, Country, SignupDate

Design a beginner-friendly star schema for me. Tell me which table is the fact table, which are dimensions, what relationships to create (which column links to which), and whether I need a separate Date table. Explain each choice in one sentence."

You will get a clear plan: Sales is the fact table; Products and Customers are dimensions; create relationships on ProductID and CustomerID; and yes, add a Date table linked to OrderDate. Every choice comes with a reason so you learn, not just copy.

You can go further and ask Claude or ChatGPT to generate the DAX for a Date table:

"Write a DAX calculated table named 'Date' for Power BI covering 2023 to 2026, with columns for Date, Year, Quarter, Month number, Month name, and Day of week. Explain how to add it."

Paste the result into Power BI via Modeling > New table.

Creating Relationships in Power BI

Switch to the Model view (the icon on the left that looks like connected boxes). You will see your tables as boxes with their column lists. To create a relationship, drag the shared column from one table onto the matching column in the other, for example, drag ProductID from Sales onto ProductID in Products. Power BI draws a line between them.

Power BI often auto-detects relationships when you load data, so some lines may already exist. Check them: click a relationship line and confirm it links the correct columns and that the direction makes sense (the "one" side should be the dimension, the "many" side the fact). If you are unsure what a relationship's cardinality means, ask:

"In Power BI, my Sales-to-Products relationship shows 'many-to-one' from Sales to Products. Explain in plain English whether that's correct and what it means for my charts."

Checking Your Model Is Right

A good model quietly produces correct numbers. To test yours, build a quick table visual with ProductName and Revenue. If each product shows its own sensible total, your relationship works. If every product shows the same grand total, the relationship is missing or wrong, a classic symptom. Delete the visual, fix the relationship, and try again.

You can describe symptoms to AI for a diagnosis:

"In Power BI, when I put ProductName next to a Revenue measure, every product shows the same total (the grand total). What's wrong and how do I fix it?"

The likely answer: no active relationship between Sales and Products. AI will walk you through creating it.

Why This Step Pays Off

Skipping the model is the most common reason beginners get "weird numbers" they cannot explain. Spending ten minutes letting AI design a star schema means your DAX in the next lesson will mostly just work. Clean data plus a sound model is 80 percent of a trustworthy dashboard; the visuals are the easy, fun part on top.

Key Takeaways

• A data model connects your tables through relationships built on shared ID columns, so charts can combine fields from different tables.
• The beginner-friendly gold standard is a star schema: one central fact table (events you measure) linked to dimension tables (things you slice by).
• A dedicated Date table is essential for time calculations like year-over-year.
• Describe your tables to ChatGPT or Claude and let it design the star schema, name the relationships, and explain each choice.
• Test the model with a simple table visual; if every row shows the same grand total, a relationship is missing, ask AI to help fix it.