Mastering Demand Gen Experiments with Google Ads

In the fast-moving world of digital advertising, one-size-fits-all strategies just don’t cut it anymore. Whether you’re running YouTube ads, leveraging Display & Video 360 (DV360), or aiming for high-impact performance across Google’s network, continuous optimization is the name of the game. That’s where Demand Gen Experiments come in—a game-changing feature in Google Ads designed to help you test, learn, and grow.

In this post, we’ll break down exactly what Demand Gen Experiments are, how to use them, and how they can supercharge your campaigns. No jargon, just clear steps and practical tips to help you level up.

What Are Demand Gen Experiments in Google Ads?

Demand Gen Experiments are a testing framework built into Google Ads (and DV360) that allow advertisers to experiment with different campaign variables—like creative assets or audience targeting—to determine what drives better performance.

Think of it as your personal ad lab. You can create two or more versions (called “arms”) of a campaign and run them side-by-side. Google will randomly split your target audience between these arms, measure performance, and provide clear, data-backed results.

Whether you're wondering which video performs best or if a new audience segment converts better—Demand Gen Experiments help remove the guesswork.

Key Features and Benefits

Demand Gen Experiments are built with accessibility and performance in mind. Here's what makes them powerful, even for non-technical users:

Controlled A/B Testing

Run split tests with a control group and experimental group. Each version gets equal opportunity (impressions) and random audience assignment, ensuring clean, reliable results.

Statistical Significance Built In

Google automatically calculates statistical significance, so you know whether your results are due to chance or truly impactful.

Supports Creative and Audience Testing

You can test video vs. image ads, different call-to-actions (CTAs), or compare audiences (like interest-based vs. lookalikes).

Insight-Driven Optimization

The tool not only shows you what worked but helps you understand why. You’ll build a solid insights library to improve future campaigns.

How to Set Up and Use Demand Gen Experiments

Setting up your first Demand Gen Experiment is easier than it sounds. Here’s a step-by-step guide:

1. Create Two New Demand Gen Campaigns

Start fresh. These campaigns must never have run before. Label one as the control and the other as the test.

2. Define a SMART Hypothesis

Use the SMART framework (Specific, Measurable, Achievable, Relevant, Timed). For example:

“If we use voice-over in our video ad, we’ll increase conversion rates over a 30-day period.”

3. Choose Your Success Metric

Pick from:

• Cost per conversion
  
• Conversion rate
  
• Click-through rate (CTR)
  
• Average cost per click (CPC)
  

If conversions matter most, go with Cost/Conv. or Conversion Rate.

4. Allocate Traffic

Split traffic 50/50 or 80/20 between test arms. This ensures each variation gets fair exposure.

5. Run for 3–4 Weeks

Let your experiment breathe. Google recommends allowing enough time (and budget) to gather at least 100 conversions for reliable results.

Best Practices for Testing Your Assets

Want to test creatives? Here are a few popular experiment designs:

• Single Image vs. Carousel
  
• Video with CTA A vs. CTA B
  
• Portrait vs. Square video
  
• Image-only vs. Image + Video combo
  

Need to test audiences? Try:

• Lookalike vs. Interest-based targeting
  
• Demographic layers (age, gender) vs. optimized targeting
  

Remember: Test only one variable at a time to clearly isolate what’s driving change.

Measuring and Interpreting Results

Once your experiment is live, head to the Experiments tab in Google Ads. You’ll find:

• Top Card: Highlights which campaign version (arm) performed best for your chosen metric.
  
• Traffic and Conversion Metrics: CTR, CPC, conversions, etc.
  
• Confidence Level: Choose 70%, 80%, or the gold standard—95%. Higher confidence = more reliable insights.
  
• Tooltip Details: Click into any data cell to view deeper insights like confidence intervals and relative uplift.
  

Pro Tip: Resist the urge to edit campaigns mid-experiment. Even small changes can skew results.

Conclusion: Test. Learn. Scale.

Demand Gen Experiments put data-driven creative and audience testing within easy reach—even if you're not a seasoned ad pro. By using this tool, you're not just launching campaigns—you’re launching smarter, better-performing campaigns with every iteration.

So go ahead—form a hypothesis, set up your arms, and start experimenting. Your future ad performance will thank you.