How to Use What-If Analysis in Tableau with AI
Predicting the future is impossible, but building an informed strategy requires us to try. This is where what-if analysis comes in, allowing you to model potential outcomes based on changing variables so you can make smarter, more proactive decisions. This article will walk you through how to perform what-if analysis in Tableau and explore how adding an AI layer can revolutionize the entire process.
What is What-If Analysis, Really?
At its core, what-if analysis is the process of changing the values in cells to see how those changes affect the outcome of formulas on a worksheet. It’s a way to substitute different inputs into a model to see the potential outputs. Think of it as a financial or strategic playground where you can test theories without real-world consequences.
For example, a marketing manager might ask:
What if we increase our ad spend on Facebook by 15%? How will that impact our overall lead generation and cost per acquisition?
What if our website conversion rate drops by 0.5% next quarter? How much revenue would we lose?
What if our sales team closes 10% more deals? What would our total revenue look like at the end of the year?
This process transforms data from a rearview mirror (what happened) into a map of potential futures (what could happen).
The Traditional Way: Building a What-If Analysis in Tableau
Tableau is a powerhouse for data visualization and offers robust features for creating interactive what-if scenarios. The traditional process generally involves combining two key features: Parameters and Calculated Fields.
While powerful, setting this up requires a degree of technical comfort with Tableau’s interface. Anyone who has spent time learning a serious BI tool knows there’s a significant learning curve to becoming proficient. Let's walk through the high-level steps.
Step 1: Create Your Parameters
A parameter in Tableau is a dynamic placeholder for a value. For what-if analysis, you'll create parameters that let users input their own values to see how the data changes. These become the interactive toggles or input boxes on your dashboard.
For our marketing spend example, you might create a parameter called "Ad Spend Increase" with the following properties:
Data Type: Float (to allow for decimals)
Display Format: Percentage
Allowable values: Range
Range of values: A minimum of -0.5 (-50%) and a maximum of 1.0 (100%), with a step size of 0.05 (5%).
This creates a slider that lets a user adjust the hypothetical ad spend increase from -50% to +100%.
Step 2: Create Calculated Fields to Use the Parameters
A parameter doesn't do anything on its own, it needs to be connected to your data through a calculated field. This is where you define the logic of your what-if scenario.
Continuing the example, you would create a new calculated field called "Projected Ad Spend" with a formula like this:
[Current Ad Spend] * (1 + [Ad Spend Increase])
Likewise, if you assume that a dollar of ad spend generates a certain amount of revenue, you could create a "Projected Revenue" field:
[Projected Ad Spend] * [Revenue Per Ad Dollar]
You can see how this can quickly become complex. You might need multiple parameters (for ad spend, conversion rate, team size) and a web of interconnected calculated fields to create a comprehensive model.
Step 3: Build the Visualization
With your parameter controls and calculated fields ready, you can build your visualization. You would drag the calculated fields (like "Projected Revenue" and "Current Revenue") onto your view, perhaps in a bar chart for easy comparison. Then, you'd show the parameter control on the dashboard, making it an interactive slider or text box for the end-user.
This method works, but it has its challenges:
It requires technical proficiency. You need to understand how to build and link parameters and calculated fields correctly.
It can be slow. Creating each scenario requires setup time. Modeling five different variables means creating five parameters and five or more associated calculated fields.
It’s not truly conversational. The analysis is limited to the scenarios you thought to build ahead of time. Spontaneous questions that arise during a meeting can't be answered without going back and building a new parameter.
Bridging the Gap: Where AI Enhances What-If Analysis
This is where an AI layer change the entire dynamic. Instead of manually building the architecture for every potential question, you can use natural language to explore scenarios on the fly. The traditional reporting process is slow — downloading data on Monday for a report on Tuesday, then spending Wednesday answering follow-up questions. AI allows you to get those answers in real-time.
AI-powered analytics tools can directly connect to your live data sources (think Google Analytics, Salesforce, Shopify, etc.). This allows them to understand the semantics of your data and perform complex calculations instantly based on a simple prompt.
A New Workflow: Conversational What-If Analysis with AI
Imagine a workflow where you can simply ask your questions instead of building the mechanics to answer them. This is what AI brings to the table, and it makes complex analysis accessible to anyone on your team, regardless of their technical skill.
Step 1: Get a Baseline by Asking a Simple Question
You start with a simple prompt to understand the current state.“Show me the total revenue from our top 5 marketing channels for last quarter, displayed as a bar chart.”
The AI tool instantly generates a visualization pulling live data from all your connected sources.
Step 2: Introduce a “What-If” Variable Conversationally
Now, instead of navigating menus to create a parameter, you ask another question.“Okay, what would the total revenue look like if our Instagram Ad spend had been 25% higher, assuming the same ROI?”
The AI handles the calculation behind the scenes. It understands what "Instagram Ad spend" and "ROI" mean in the context of your data, creates the projection, and updates the chart to show a side-by-side comparison of actual vs. projected revenue.
Step 3: Layer on More Complexity and Drill Down
This is where the process becomes transformative. A good visualization often sparks more questions. With AI, you can keep digging.“That’s interesting. For that projected scenario, show the impact on profit, not just revenue.”
Maybe you get a surprising result, and another question comes to mind.“Now, let's go back to the original projection. What if our conversion rate on paid ads also increased by 10% on top of the increased spend? Model that new outcome.”
Creating this multi-variable scenario in Tableau would require multiple parameters, complex nested calculations, and significant setup time. With a conversational AI tool, it’s just one more question. It empowers even the most junior team members to explore data and uncover insights that might have otherwise been hidden behind a technical barrier.
AI as Your Brainstorming Partner, Not Just a Calculator
The true power of AI in analytics extends beyond just being an efficient "order taker." Great AI tools act as a collaborative partner, helping you ask smarter questions you hadn't considered. Instead of only testing your hypotheses, you can ask the AI to generate its own.
You could ask something like:
“Analyze last year's sales data. What are the top three variables I could influence that would have the biggest impact on our revenue next quarter?”
The AI might come back with an analysis suggesting that:
Increasing the ad budget for a specific, high-converting Google Ads campaign offers the highest potential ROI.
Improving the conversion rate on your top landing page by just 0.5% would result in more revenue than a 10% increase in overall traffic.
Leads from your HubSpot webinar series have a 15% higher close rate than any other source, suggesting a budget reallocation.
This proactive guidance turns what-if analysis on its head. It’s no longer just a method for validating your gut feelings, it becomes an engine for discovery, showing you the most effective levers to pull to grow your business.
Final Thoughts
What-if analysis is an invaluable tool for any business that wants to make data-driven decisions. Building these scenarios in powerful BI tools like Tableau is entirely possible, but traditionally requires technical expertise and setup time. Layering AI on top of your data transforms the process from manual and structured to conversational and dynamic, inviting everyone to explore possibilities and uncover insights.
Instead of wrestling with data prep, calculated fields, and dashboard configurations, what if you could just ask for what you need? At Graphed, we built an AI data analyst to remove this friction entirely. You can connect your marketing and sales data sources one time, then simply ask "what if?" questions in plain English — no dashboards to build, no code to write. We instantly create the visualizations and models for you, allowing you to have a live conversation with your data and go from curiosity to clarity in seconds.