How to Use Google Ads Experiments for A/B Testing

Running Google Ads without testing is a bit like driving through Brisbane's Story Bridge in peak hour with your eyes closed—you might get where you're going, but it's going to be stressful and potentially expensive. Google Ads Experiments allow you to test changes to your campaigns (like new bidding strategies or different headlines) against your current setup in a controlled, scientific way.

By splitting your traffic between an 'Original' and a 'Trial' version, you can see exactly what works before committing your entire Australian marketing budget to a change. This guide will walk you through the process from start to finish.

Prerequisites: What You’ll Need

• An active Search or Display campaign: You can't experiment on a campaign that hasn't started yet.
• Conversion Tracking: If you aren't tracking leads or sales, an experiment is useless because you won't know which version actually made you money.
• A clear hypothesis: For example, "I think switching to Target CPA bidding will get me more leads than Manual CPC."
• Enough data: If your campaign only gets 5 clicks a week, an experiment will take years to give you a result. Ideally, your campaign should have at least 30-50 conversions per month.

---

Step 1: Define Your Goal

Before you click a single button in the Google Ads interface, you need to know what you are testing. One of the biggest mistakes I see Brisbane business owners make is trying to test three things at once.

If you change the bidding strategy, the ad copy, and the landing page all in one experiment, and the performance improves, you won't know which change caused the win. Pick one variable.

• Bidding Strategies: e.g., Maximize Conversions vs. Target CPA.
• Ad Copy: Testing a "Free Quote" call-to-action against "Book Online."
• Landing Pages: Sending half your traffic to your homepage and half to a specific service page.

Step 2: Navigate to the Experiments Tab

Log in to your Google Ads account. On the left-hand sidebar menu, you’ll see a section called "Campaigns". Click it to expand, and then look for "Experiments" near the bottom of that list.

Once you are in the Experiments dashboard, click the blue plus (+) button to start a new one.

Step 3: Choose Your Experiment Type

You will be presented with several options. For most local businesses, you'll want to select "Custom experiment" (usually for Search or Display).

• Video Experiments: Specific for YouTube.
• App Experiments: Only if you have a mobile app.
• Performance Max Experiments: Useful for testing PMax against Standard Shopping.

For this guide, we’ll assume you’re doing a Custom Experiment for a standard Search campaign.

Step 4: Set Up Your Base Campaign

Now, you need to give your experiment a name. Be descriptive! Instead of "Test 1," use something like "[Campaign Name] - Bidding Test - T-CPA vs Max Conv."

Next, select your Base Campaign. This is the "control" or the current campaign you want to test against.

Step 5: Create Your Trial Version

This is where the magic happens. Google will now create a "Trial" version of your campaign. This is essentially a draft that is a carbon copy of your original.

• Click "Save and continue."
• You are now inside the Trial setup. Here, you make the one change you decided on in Step 1.
• If you're testing bidding, go to the campaign settings and change the bid strategy.
• If you're testing ad copy, go to the ads section and edit the headlines.

Step 6: Schedule the Experiment

Once you've made your changes, click "Schedule." This is where people often get a bit nervous, but it's simpler than it looks.

• Experiment Split: Usually, you want a 50% split. This means half your budget and traffic go to the original, and half go to the trial. It’s the fastest way to get a statistically significant result.
• Split Option: Choose "Cookie-based." This ensures that if a customer in Chermside clicks your ad twice, they see the same version both times. Using "Search-based" can confuse the data because the same person might see different ads for the same search.
• Start Date: Set it for tomorrow so it starts at the beginning of a fresh day.
• End Date: I usually recommend at least 30 days. Don't set it for 1 week; Australian search volumes fluctuate too much over a weekend to trust 7 days of data.

Step 7: The Waiting Game (The Hardest Part)

Once you hit "Create," the experiment will go into a "Pending" state while Google reviews the ads. Once live, the most important advice I can give you is: Leave it alone.

It is incredibly tempting to check the stats after two days, see that the trial is performing worse, and panic-cancel it. Google Ads uses machine learning, and there is a "learning phase" that can last 7-14 days. If you touch it during this time, you ruin the experiment.

Step 8: Analysing the Results

After your experiment has been running for a few weeks, head back to the Experiments tab. You'll see a table with your results. Look for the little blue diamonds or icons that indicate "Statistical Significance."

Google will tell you if there is a "95% confidence interval." This is fancy talk for: "This result isn't a fluke; the change actually caused the difference."

• Conversions / Cost per Conversion: This is usually the gold standard for Australian small businesses.
• CTR (Click-Through Rate): Important if you're testing ad copy.
• Conversion Value: Important for e-commerce stores looking for a higher ROAS.

Step 9: Apply or End the Experiment

Once the experiment ends (or you have enough data), you have two choices:

• Apply: If the Trial won, click "Apply." Google will then update your original campaign with the new settings. You can either overwrite the original or create a brand-new campaign.
• End: If the Trial performed worse or there was no real difference, just end the experiment. Your original campaign will continue running as if nothing happened.

---

Common Mistakes to Avoid

• Testing too many things: As mentioned, keep it to one variable. If you change the bid AND the location targeting, you won't know which one worked.
• Ending too early: Unless you are losing thousands of dollars a day, let it run for the full 30 days.
• Ignoring Seasonality: Don't run an experiment in December and compare it to November data. The "split" method handles this well, but be aware that consumer behaviour in Brisbane changes significantly during school holidays or the Ekka week!
• Small Budgets: If your daily budget is only $5, the experiment will take forever to get enough data. You might need to increase the budget temporarily to get a result.

Troubleshooting

Next Steps

Now that you know how to run a scientific test, why not start with something simple? Try testing a different landing page for your highest-spending campaign.

If this all feels a bit overwhelming or you're worried about wasting your ad spend, we're here to help. At Local Marketing Group, we manage Google Ads for businesses across Brisbane and Australia, ensuring every dollar is backed by data.