How to Do Predictive Analysis in Google Analytics with AI

Tired of only looking at last month’s data? Google Analytics 4 gives you the power to look forward, using machine learning to predict what your users will do next. This article breaks down how to use GA4's built-in predictive features to forecast user behavior and get ahead of the curve. We’ll cover what these metrics are, how to qualify for them, and how you can use AI to go even further.

Graphed

Still Building Reports Manually?

Watch how growth teams are getting answers in seconds — not days.

Watch the Demo

What Exactly is Predictive Analysis?

For years, analytics tools mostly helped us look backward. We could see how many people visited our site last week, which pages they viewed, and where they came from. This is called descriptive analytics - it describes what already happened.

Predictive analysis, on the other hand, uses that historical data and applies machine learning to forecast what’s likely to happen in the future. Instead of just knowing which users bought something last month, you can predict which current users are most likely to buy something in the next week. It’s like having a crystal ball for your marketing efforts, helping you make smarter, proactive decisions instead of always reacting to past performance.

Universal Analytics (the old GA) didn’t really have this capability. But it’s a core feature of GA4, giving you access to powerful insights without needing a data science degree.

The 3 Predictive Metrics Built Into Google Analytics 4

GA4 models user behavior to forecast three specific outcomes. Understanding each one is the first step to putting them into action.

1. Purchase Probability

This metric predicts the likelihood that a user who has been active on your site or app in the last 28 days will make a purchase within the next 7 days. It looks at their recent behavior - like pages viewed, items added to cart, and time spent on site - and compares it to the behavior patterns of past purchasers.

Example: Imagine an online store. A user who has visited a product page three times, added the item to their cart, and started the checkout process will have a much higher Purchase Probability than a user who just landed on the homepage for the first time.

Graphed

Still Building Reports Manually?

Watch how growth teams are getting answers in seconds — not days.

Watch the Demo

2. Churn Probability

Churn Probability is the opposite of engagement. It predicts the likelihood that a recently active user will not visit your site or app within the next 7 days. This is incredibly valuable for identifying users who are losing interest so you can try to win them back before they’re gone for good. You can think of it as an early warning system for customer retention.

Example: A user who used to visit your blog every day but hasn’t visited in the last 10 days will have a higher Churn Probability. This gives you a chance to re-engage them with a targeted email or ad campaign.

3. Predicted Revenue

This metric forecasts the total revenue you can expect from all purchase events made by a user over the next 28 days. It helps you identify not just potential buyers, but potential high-value customers. This is perfect for prioritizing your marketing budget and focusing on users who are likely to contribute the most to your bottom line.

Example: Two users might have the same high Purchase Probability, but one of them has historically bought premium products. The Predicted Revenue for that user will be much higher, signaling that they are a more valuable prospect to target with special offers.

How to Qualify for GA4’s Predictive Metrics

You might log in to your GA4 account and find these metrics aren't available yet. That's because Google has specific data thresholds you need to meet before its machine learning models have enough information to make accurate predictions. For your GA4 property to be eligible, it needs to:

• Meet a minimum number of positive and negative examples of users. For example, for Purchase Probability, you need at least 1,000 users who triggered a purchase event and at least 1,000 users who did not, all within a 28-day period.
• Sustain these threshold levels for at least 28 consecutive days for the models to be properly trained. If your data volume drops below the minimum, the predictive features will be paused until the thresholds are met again.
• Have the purchase event (for websites) and/or the in_app_purchase event (for apps) properly configured to track conversions.

This can be a challenge for smaller businesses or new websites. If you don't qualify yet, your main goal should be on two things:

1. Driving more qualified traffic and conversions to increase your data volume.
2. Ensuring your event tracking is correctly implemented so GA4 can recognize your key user actions.

Once your property is eligible, GA4 will start generating predictions for your users automatically. You won't have to turn anything on.

Graphed

Still Building Reports Manually?

Watch how growth teams are getting answers in seconds — not days.

Watch the Demo

Step-by-Step: Creating and Using Predictive Audiences

Seeing the data is interesting, but the real power comes from turning these predictions into action. The best way to do this is by creating Predictive Audiences, which you can then use for ad targeting and website personalization.

GA4 provides suggested audience templates to make this easy.

Step 1: Navigate to the Audience Builder

In your GA4 property, go to Admin > Audiences (under the "Data display" section) > click New audience.

Step 2: Start with a Predictive Template

You’ll see a prompt to create a custom audience. Near the top, click on the Predictive tab to see the available templates. These will correspond to the core metrics:

• Likely 7-day purchasers
• Likely 7-day churning users
• Predicted 28-day top spenders
• Likely first-time 7-day purchasers
• Likely 7-day churning purchasers

Step 3: Configure Your Audience

Let’s say you choose 'Likely 7-day purchasers'. You can configure this audience based on how confident you want to be. The slider lets you select users who fall within a certain percentile range. For example, you can target:

• Users with the highest likelihood to buy (e.g., 90th to 100th percentile) for a super-targeted, high-intent campaign.
• A broader group of potentials (e.g., 70th to 100th percentile) to fill your top-of-funnel with promising users.

After you configure it, give your audience a clear name (like "High-Intent Shoppers - USA") and a description, then save it.

Step 4: Activate Your Audience in Google Ads

Once saved, this audience will be automatically available in your linked Google Ads account. Here are a few ways to use it:

• Remarketing for High-Intent Users: Create a campaign targeting your "Likely 7-day purchasers" with a compelling offer or a direct call-to-action to close the deal.
• Churn Prevention: Build a campaign for your "Likely 7-day churning users" audience. You can offer them a special "we miss you" discount or showcase new features to bring them back.
• Lookalike Audiences: Use your "Predicted 28-day top spenders" as a seed audience to find new users on the Google Display Network or YouTube who share similar characteristics to your most valuable predicted customers.

This transforms your data from a static report into an active, revenue-generating tool.

Graphed

Still Building Reports Manually?

Watch how growth teams are getting answers in seconds — not days.

Watch the Demo

Beyond GA4: Answering Deeper Questions with AI

GA4's built-in tools are a fantastic starting point, but they do have limitations. You’re restricted to three specific metrics, and the eligibility requirements can be tough to meet. More importantly, GA4 can't easily answer the question "Why?" Why are these specific users likely to churn? What campaigns are driving our most predictable future purchasers?

Answering those deeper, strategic questions is where dedicated AI data analysis tools come in. These platforms connect not just to GA4, but to all your other data sources - your CRM (like HubSpot or Salesforce), ad platforms (Facebook, Google Ads), and e-commerce platform (Shopify). By unifying your data, they get a complete picture of your business.

The biggest difference is how you interact with them. Instead of clicking through menus, you ask questions in plain English, just like you would with a human analyst.

For example, you could ask:

• "Forecast our Shopify revenue for the next 90 days based on our current traffic growth from Google Ads."
• "Show me a list of common behaviors among users with a high churn probability in both GA4 and HubSpot."
• "Which marketing channels from last quarter are predicted to bring in the most valuable new customers next quarter?"

These are wildly complex questions that would typically require hours of manual data wrangling and exporting CSVs into spreadsheets. With an AI analyst, you get answers and visualizations in seconds.

This allows you to move from simply identifying a trend (e.g., churn is an issue) to understanding the root cause (e.g., users who read blog posts but never visit the pricing page are most likely to churn) and build smarter, more effective strategies.

Final Thoughts

Moving from looking backward at past data to looking forward with predictive analysis is a massive step up for any marketer. Google Analytics 4 provides a powerful, free entry point with its built-in Purchase Probability, Churn Probability, and Predicted Revenue metrics, which you can use to create highly effective audiences for your ad campaigns.

While GA4 is a great start, the process can feel limited if you want to answer deeper business questions or connect insights from across all your tools. We built Graphed to be your AI data analyst for this exact reason. You can connect sources like Google Analytics, Shopify, and Facebook Ads in just a few clicks, then simply ask questions in plain English to get real-time forecasts, dashboards, and actionable insights. It makes predictive analysis accessible to your entire team - no data science required.