A/B Testing Google Ads

A/B Testing Google Ads: Systematically Improving Campaign Performance

Introduction

A/B testing in Google Ads is the practice of running two versions of an ad element simultaneously, directing roughly equal traffic to each, and measuring which produces better results. When done correctly, each test produces a marginal improvement that compounds over time into significantly better campaign performance.

For service businesses running Google Ads, the elements worth testing are: ad copy (headlines and descriptions), landing pages, audience targeting, bidding strategies, and ad extensions. This guide focuses on the elements with the highest potential impact and the most practical testing methodology.

What to Test (In Order of Potential Impact)

1. Ad headlines (highest impact) Your headlines are the most-read part of your ad and have the largest influence on CTR (click-through rate). Google's Responsive Search Ads (RSAs) allow up to 15 headlines and 4 description lines — Google's algorithm tests combinations automatically.

However, truly systematic A/B testing requires more control than RSA auto-rotation provides. For meaningful headline testing, use Google Ads' built-in Experiments feature (Drafts & Experiments) to create controlled ad variant tests.

High-priority headline variables to test:

• Price anchoring ("Quotes from £150") vs. benefit-first ("Fast, Reliable, Gas Safe Registered")

• Location inclusion ("Emergency Plumber Leeds") vs. availability ("Emergency Plumber — Available Now")

• Social proof ("4.9 Stars — 200+ Reviews") vs. USP ("20 Years Experience — Fully Insured")

2. Call to action copy (high impact) "Get a Free Quote" vs. "Call Now" vs. "Book Online Today" — test which CTA drives higher CTR and conversion rate for your specific service and audience.

3. Landing pages (high impact, harder to test) Testing landing page variants requires more infrastructure than ad copy testing — but is often where the most significant conversion rate improvements are found. Use Google Ads Experiments to split-test traffic between two landing page URLs, or use a tool like Google Optimize (being retired — consider VWO or ABTasty as alternatives) for page-level testing.

4. Bidding strategies Testing Smart Bidding strategies (Target CPA, Maximise Conversions, Target ROAS) against manual CPC or Enhanced CPC provides data on whether automated bidding improves or hurts performance for your account. Use the Drafts & Experiments feature to run bidding strategy tests fairly.

5. Ad extensions Test different sitelink combinations, callout text variations, and structured snippet choices. Extension testing is lower complexity and lower risk than headline testing.

Using Google Ads Experiments for Controlled Testing

Google Ads' built-in Experiments tool (Campaigns → Drafts & Experiments → Experiments) is the correct framework for controlled testing:

Creating an experiment:

1. Select the campaign you want to test

2. Create a draft of the campaign (your experimental version)

3. Make the specific change you want to test in the draft

4. Create an Experiment from the draft, defining the traffic split (50/50 recommended for most tests) and duration (minimum 4 weeks)

The Experiment runs the original campaign and the experimental variant simultaneously, with traffic split randomly between them. Results are reported alongside statistical confidence metrics.

Interpreting experiment results: Google Ads shows the performance difference between the control and experimental variant for your chosen metric (conversions, conversion rate, cost per conversion). Look for:

• The direction of change (is the experimental variant better or worse?)

• Statistical confidence (Google shows this as a percentage — aim for 95%+ confidence before drawing conclusions)

• Absolute vs. relative impact (a 10% improvement in conversion rate on 100 conversions/month is more meaningful than 10% on 5 conversions/month)

Responsive Search Ads and Testing

Responsive Search Ads (RSAs) are now the primary ad format in Google Ads. RSAs allow you to provide multiple headline and description options, and Google's algorithm tests combinations automatically — surfacing the combinations that perform best for different searches and audiences.

Effective RSA testing strategy:

Pin your most critical headlines and descriptions to specific positions. Headline 1 and Description 1 are most consistently shown — pin your primary message to these:

• Pin 1 headline as your service + location (always shown first)

• Leave remaining headlines unpinned for Google to test

Add as many distinct headline options as possible — not variations of the same idea, but genuinely different value propositions:

• Brand ("[Business Name]")

• Social proof ("4.9 Stars from 200+ Customers")

• Availability ("Available 24/7 in Leeds")

• Price anchor ("Quotes from £95")

• Urgency ("Same-Day Service Available")

• Benefit ("No Fix No Fee Guarantee")

Review the Asset Strength report and the "Combinations" report (ad level → Combinations) to see which headlines Google is favouring and use these insights to inform your next iteration.

Testing Cadence and Management

One test at a time per campaign: Running multiple simultaneous tests in the same campaign makes it impossible to attribute performance changes to specific variables. Test sequentially, not simultaneously.

Minimum test duration: Run tests for at least 4 weeks to account for day-of-week variation and gather statistically meaningful data. Don't end tests early based on initial results — early data is highly variable.

Document everything: Maintain a testing log (a simple spreadsheet works) recording: what was tested, the hypothesis, the dates, the result, and whether the change was implemented. This historical record prevents re-testing the same hypotheses and builds institutional knowledge about what works for your specific campaigns.

Sample size requirements: For conversion rate testing, you need sufficient conversions in both the control and experimental variant to draw reliable conclusions. A test with 10 conversions in each variant is not statistically reliable. Aim for at least 100 conversions per variant before concluding a test.

What Good Testing Discipline Produces Over Time

A service business campaign that runs one structured test per month for 12 months — iterating on copy, landing pages, bidding, and audience targeting — compounds improvements that dramatically outperform the performance of an untested "set it and forget it" campaign.

Even modest per-test improvements (5–10% conversion rate improvement, 5% CTR improvement, 10% CPC reduction) compound across 12 months into significantly lower cost-per-lead and higher lead volume from the same budget.

Conclusion

A/B testing is the systematic discipline that separates improving Google Ads campaigns from stagnant ones. It requires structure (one test at a time, minimum duration, adequate sample size), patience (meaningful tests take weeks), and documentation (building institutional knowledge about what works). The service businesses that run it consistently achieve materially better advertising economics over time.

Want Google Ads campaigns that are continuously tested and optimised? Zava Build manages performance-driven PPC campaigns for UK service businesses. Book a free strategy session →