How to Forecast Sales in Looker with AI

Building an accurate sales forecast used to mean wrestling with complex spreadsheets and making educated guesses. Today, you can create surprisingly accurate, data-driven predictions directly within Looker using its built-in AI capabilities. This article will walk you through how to use Looker's forecasting tools to predict future sales and make smarter business decisions.

Graphed

Build AI Agents for Marketing

Build virtual employees that run your go to market. Connect your data sources, deploy autonomous agents, and grow your company.

Build Your First Agent

Rethinking Sales Forecasting

A good sales forecast is more than just a number, it’s a strategic tool. It helps you manage inventory, allocate your marketing budget, set realistic team goals, and plan for growth. Without it, you're essentially flying blind.

The traditional method - pulling sales history into Excel and manually extending a trendline - is both time-consuming and prone to errors. It often fails to account for crucial factors like:

• Seasonality: Predictable peaks and troughs, like the holiday shopping rush or a summer slowdown.
• Trends: The overall upward or downward trajectory of your sales over time.
• Anomalies: Random spikes or dips that can throw off a simple linear prediction.

This is where modern BI tools with AI features make a massive difference. By using machine learning models, Looker can analyze your historical data, identify these complex patterns, and generate a much more nuanced and reliable forecast - all within a few clicks.

AI Agents for Marketing

Crash Course

Graphed

Free PDF · the crash course

AI Agents for Marketing Crash Course

Learn how to deploy AI marketing agents across your go-to-market — the best tools, prompts, and workflows to turn your data into autonomous execution without writing code.

Download the PDF

A Quick Primer on Looker

Before diving into the forecasting "how-to," let's quickly review what Looker is. At its core, Looker (now part of Google Cloud) is a business intelligence platform that connects directly to your company's databases (like BigQuery, Redshift, or Snowflake). It doesn't store your data, it queries it live, ensuring you're always working with the most up-to-date information.

The magic of Looker lies in LookML, its modeling language. Your data team uses LookML to define business metrics and relationships (e.g., what counts as "revenue," how "customers" relate to "orders"). This creates a single source of truth, allowing business users to explore data, build visualizations, and create dashboards without needing to write a single line of SQL.

How to Create a Forecast with Looker's Built-in Tools

Looker’s simplest forecasting feature is built directly into its charting tools. It uses historical data patterns to project future trends. Here’s how to set it up step-by-step.

Step 1: Choose Your Data

First, you need to have clean, organized historical data. The quality of your forecast is entirely dependent on the quality of the data you feed it. At a minimum, your dataset should contain:

• A Time Dimension: A field with dates or timestamps, such as an order creation date.
• A Measure: A numeric value you want to forecast, like sales revenue, units sold, or new user sign-ups.

For the best results, you need a decent amount of historical data. The more data points the model has to learn from, the more accurately it can identify trends and seasonality. Aim for at least a year's worth of data, especially if your business has strong seasonal patterns.

Graphed

Build AI Agents for Marketing

Build virtual employees that run your go to market. Connect your data sources, deploy autonomous agents, and grow your company.

Build Your First Agent

Step 2: Build a Time-Series Visualization

Navigate to an Explore in Looker that contains your sales data. You'll build a basic line chart to visualize your sales over time.

1. From the field picker on the left, select your time dimension (e.g., "Order Date"). It's best to group this by week or month to smooth out daily noise.
2. Select the measure you want to forecast (e.g., "Total Sales Amount").
3. Click "Run" to generate your chart. You should see a line chart showing your sales performance over your selected time period.

This chart is your baseline - it shows what has already happened and is the foundation for your forecast.

Step 3: Activate the Forecasting Feature

Now, let's add the AI-powered forecast to your chart.

1. Click the gear icon in the upper-right corner of the visualization pane to open the Edit Chart menu.
2. Navigate to the Forecast tab.
3. Check the box to Enable Forecast.

Once enabled, Looker begins to analyze your data and project future values. You'll see a dotted line extend from the end of your solid "actuals" line, along with a shaded area around it.

Step 4: Configure Your Forecast Settings

You can customize the forecast to better suit your needs. Here are the key settings:

• Length: This determines how far into the future you want to predict. For example, if your chart shows monthly data, a length of "6" will predict sales for the next six months.
• Prediction Interval: This represents the confidence level of the forecast. The shaded area on your chart visualizes this interval. A 95% prediction interval means Looker is 95% confident the actual future value will fall within that shaded range. A wider interval indicates more uncertainty, while a narrower one suggests greater confidence.
• Seasonality: This setting tells the forecasting algorithm to look for repeating patterns over a specific period. You can leave it on "Automatic," and Looker will attempt to detect the seasonal cycle on its own. If you know your business has a clear weekly, monthly, or yearly cycle, you can set it manually for better results. For instance, an e-commerce store with a big holiday rush would have a yearly seasonality.

Once you've configured these options, you'll have a visualization showing your historical sales trend and a statistically-backed forecast of future performance.

Going Deeper with BigQuery ML and Looker

The built-in visualization forecasting is great for quick analysis and high-level trends. But what if you need more power and control? If your data lives in Google BigQuery, you can use BigQuery ML (BQML) to create even more sophisticated forecasting models directly inside Looker.

BQML allows you to train, evaluate, and deploy machine learning models using familiar SQL commands. It takes forecasting out of the hands of data scientists only and makes it accessible to proficient data analysts. The most common model for this task is ARIMA+, which is great at handling complex time-series data with trends and seasonality.

The process looks something like this:

1. Train a Model in BigQuery: You write a CREATE MODEL SQL query in BigQuery, specifying ARIMA+ as the model type and using your historical sales data as the training input.
2. Generate Predictions: Once the model is trained, you use the ML.FORECAST function to generate future predictions. This query outputs a table with your predicted future sales.
3. Connect to Looker: You can expose this BQML forecast back into Looker, usually as a "derived table" in your LookML project.
4. Visualize in a Dashboard: With the BQML forecast available as a data source, you can build dashboards in Looker that show your historical sales side-by-side with your highly accurate, machine-learning-powered forecast.

This approach gives you more granular control over the model's parameters and allows you to incorporate additional features (or "regressors") that might influence sales, like ad spend or website traffic.

AI Agents for Marketing

Crash Course

Graphed

Free PDF · the crash course

AI Agents for Marketing Crash Course

Learn how to deploy AI marketing agents across your go-to-market — the best tools, prompts, and workflows to turn your data into autonomous execution without writing code.

Download the PDF

Best Practices for Better Forecasts

No matter which method you use, keep these tips in mind to get the most out of your AI-powered forecasts:

• Data Quality is Everything: The "garbage in, garbage out" principle applies perfectly here. Make sure your historical sales data is clean, accurate, and free of gaps.
• Monitor and Refine: A forecast is a living document, not a one-and-done report. Regularly compare your forecasted numbers against actual sales to see how accurate the model is. Over time, you can refine your settings or retrain your model to improve its performance.
• Layer in Human Insight: An AI model can't know about the huge marketing campaign you're launching next month or that a new competitor just entered the market. Use the AI forecast as your quantitative baseline, then layer your qualitative, human knowledge on top to make final adjustments.
• Communicate with Context: When sharing your forecast with your team, always present it with context. Explain that it’s a data-driven prediction, not a guarantee. Use the prediction interval to set realistic expectations around the range of possible outcomes.

Final Thoughts

Looker completely changes the game for sales forecasting. By combining its data exploration capabilities with powerful, easy-to-use AI and ML features, it empowers teams to move beyond manual guesswork and build reliable, automated forecasts. This helps you plan more effectively and drive your business forward based on data, not just intuition.

While Looker and BQML offer incredible depth, we know that many marketing and sales teams don't have the time or technical background to master LookML or build SQL-based machine learning models. We built Graphed to remove that barrier. After a one-click connection to your data sources like Shopify or Salesforce, you can just ask in plain English: "Forecast our sales for the next 90 days." Graphed instantly builds the dashboard for you, giving you an AI-powered forecast without a single line of code so you can focus on insights, not setup.