How to Forecast in Tableau

Predicting future trends is a crucial part of any business strategy, but it doesn't require a crystal ball. If you have historical data, Tableau’s built-in forecasting feature can give you a powerful, data-driven glimpse into what’s likely to happen next. This article will guide you through how to create, customize, and interpret forecasts in Tableau, turning your historical performance into actionable intelligence.

What is Forecasting in Tableau?

Tableau’s forecasting feature uses a statistical technique called exponential smoothing. In simple terms, this method analyzes your historical time-series data (like daily sales, monthly website traffic, or quarterly revenue) to identify trends and seasonality. It then projects these patterns into the future to create a forecast.

The "smoothing" part is key. The model gives more weight to recent data points than to older ones, assuming that what happened last month is likely a better predictor of next month than what happened five years ago. This makes it particularly effective for business analytics where recent performance is often most relevant.

When you enable forecasting on a line chart showing a measure over time, Tableau automatically does the following:

• Extends the time axis into the future.
• Plots the predicted future values as a line.
• Shows a shaded area around the forecast line, known as a prediction interval, which represents the likely range for future values.

When Should You Use Tableau Forecasting?

Forecasting is best suited for time-series data - any dataset that has a continuous date or time component. If your question starts with "What will our numbers look like next...", Tableau forecasting can likely help.

Common Use Cases:

• Sales Forecasting: Predict next quarter’s revenue based on the last three years of sales data to set realistic targets.
• Inventory Management: Forecast demand for specific products to optimize stock levels and avoid running out of popular items.
• Website Analytics: Project user traffic for the next few months to plan for new marketing campaigns or ensure server capacity.
• Budget Planning: Estimate future operational expenses based on past spending habits to create a more accurate budget.
• Resource Allocation: Forecast the number of customer support tickets to better staff your team during peak seasons.

The primary requirements are a consistent history of data and at least one measure you want to predict. The more historical data you have with clear patterns, the more reliable your forecast will be.

Step-by-Step Guide: Creating Your First Forecast

Let's walk through building a forecast from scratch. For this example, we’ll use a sample dataset to forecast monthly sales.

Step 1: Get Your Data Ready

Before you start, make sure your data is structured properly. You need at least two essential fields:

• A Date Dimension: This field must be recognized by Tableau as a date or date-time data type (e.g., Order Date).
• A Measure: This is the numerical value you want to forecast (e.g., Sales, Sessions, Units Sold).

Clean data is crucial. Missing months or major data gaps can weaken your forecast, so try to use a complete and consistent dataset.

Step 2: Build a Time-Series View

The foundation of any forecast is a basic line chart that shows your measure over time.

1. Drag your date dimension onto the Columns shelf. Right-click it and make sure it’s set to a continuous date value, like “Month” (the green option). Continuous dates give you a proper axis, which is necessary for forecasting.
2. Drag the measure you want to forecast (e.g., Sales) onto the Rows shelf.

You should now see a line chart displaying your historical sales month over month.

Step 3: Enable Forecasting

This is where the magic happens. Turning on the forecast is a simple drag-and-drop action.

1. Go to the Analytics pane (next to the Data pane in the top left).
2. Under the “Model” section, find Forecast.
3. Click and drag "Forecast" onto your chart. You’ll see a prompt to add a forecast. Drop it right onto the view.

Tableau will instantly extend the date axis and add the forecast line along with a prediction interval. That's it! You've created your first forecast.

Customizing Your Tableau Forecast

The default forecast is a great starting point, but you can refine it for greater accuracy and relevance. To access the customization options, right-click anywhere in the forecasted area of your chart and select Forecast > Forecast Options...

This opens a dialogue box with several tabs. Let’s look at the most important ones.

Forecast Length

Here, you can control how far into the future Tableau predicts.

• Automatic: Tableau decides the length based on the cyclical patterns in your data (e.g., 5 quarters for several years of quarterly data).
• Exactly: Lets you define a specific number of time units (e.g., 12 months, 4 quarters).
• Until: Lets you pick a specific future date to forecast up to.

Source Data

This section lets you modify the data used to build the forecast model.

• Aggregate by: You can change the time granularity directly in this menu (e.g., switch from Monthly to Quarterly). This will update your entire view.
• Ignore last: This is a powerful feature. If the most recent period is incomplete (for example, if it's the middle of the current month), you can tell Tableau to ignore it so it doesn’t skew the forecast.

Forecast Model

The "Model" tab gives you the most control. Tableau defaults to "Automatic," where it analyzes your data to pick the best model for trend and seasonality.

• Automatic: Tableau makes the choice for you. This is usually the best option unless you have a deep statistical understanding of your data.
• Automatic without seasonality: If you know your data has a trend but no seasonal pattern, you can select this to prevent Tableau from looking for one.
• Custom: This allows you to manually define the Trend and Seasonality characteristics. For both, you can choose between "None," "Additive," and "Multiplicative."

Understanding Forecast Quality and Prediction Intervals

How do you know if your forecast is any good? Tableau provides a wealth of information to help you assess its quality. Right-click your forecast and select Forecast > Describe Forecast...

This opens a summary box with two key tabs: "Summary" and "Models." The Models tab gives you detailed quality metrics.

Key Quality Metrics

These metrics are calculated by testing the model against the last part of your historical data. Lower values indicate a more accurate model.

• MAE (Mean Absolute Error): The average size of the forecast error in absolute terms. An MAE of 150 means the forecast was, on average, off by $150.
• MSE (Mean Squared Error): Similar to MAE but gives a heavier penalty to larger errors.
• MAPE (Mean Absolute Percent Error): This is often the most intuitive metric. It shows the average percentage difference between the forecast and the actual values. A MAPE of 5% means your forecast model has been, on average, within 5% of the actual results.
• MASE (Mean Absolute Scaled Error): Compares the forecast’s accuracy against a simple, naive forecast (e.g., predicting next month will be the same as last month). A MASE under 1 indicates your model is better than that basic benchmark.

Reviewing these numbers gives you a quantifiable way to judge your forecast's reliability. If the error metrics are very high, you may need more data or a different approach.

Prediction Intervals

The shaded area on your forecast is the prediction interval. It provides a range within which the actual future value will fall with a certain level of confidence. By default, this is set at 95%. But what does that mean? It means that, based on the model, there's a 95% probability that the actual sales figure will fall within that shaded range. You can change this in Forecast Options > Prediction Interval to be wider (99%) or narrower (90%).

Common Pitfalls and Best Practices

Avoid these common mistakes to create more reliable forecasts:

• Don't Forecast with Too Little Data: Exponential smoothing models need a sufficient amount of historical data to identify trends and seasonality. While Tableau can generate a forecast with just five data points, you'll get far more trustworthy results with several years of monthly data or multiple cycles of seasonality.
• Ensure Your Date Field is Continuous: If your date is set to "Discrete" (a blue pill), you'll just get labels for each period, not a continuous time axis. The Forecast option won't even be available. Always use a continuous date (a green pill).
• Account for Gaps and Outliers: A three-month period of zero sales due to a business disruption will confuse the model. Consider filtering out these anomalous periods or using visualizations to understand their impact before forecasting.
• Remember It’s a Guide, Not a Guarantee: Forecasts are based on the past. They can't predict unexpected market crashes, a new competitor launching, or a viral marketing campaign. Always supplement statistical forecasts with your own business knowledge and context.

Final Thoughts

Creating forecasts in Tableau is an excellent way to move from reactive analysis to proactive, data-driven planning. By following the steps to build, customize, and evaluate your forecasts, you can transform your historical data into a strategic asset that helps you prepare for what's ahead.

While powerful tools like Tableau are essential for in-depth data visualization, we know the learning curve can be steep and time-consuming. We created Graphed because we believe getting valuable insights from your data should be faster and more conversational. Instead of hunting through menus, you can connect your marketing and sales data and simply ask for what you need - whether it’s a sales forecast, a campaign performance dashboard, or an answer to a specific question - and get it in seconds.