Process Flow
Animated overview of the full workflow
TL;DR
Google Ads Experiments let you test changes to campaigns in a controlled way by splitting traffic between your original campaign and a variant. You can test bidding strategies, ad copy, landing pages, keyword match types, and more. Create an experiment from the Experiments tab, set your traffic split (typically 50/50), define a duration (2-4 weeks minimum), and let it run without interference. Google provides statistical significance indicators so you know when results are reliable.
Why Use Experiments
Making changes directly to a live campaign is risky — if the change hurts performance, you have lost time and budget with no clean comparison data. Experiments solve this by running a controlled test where a percentage of traffic sees the original campaign and the rest sees your modified version.
Step-by-Step Guide
Follow these 5 steps to complete this guide
- 1
Types of Experiments
**Custom experiments:** Test any campaign-level change — bidding strategy, budget, ad rotation, targeting, landing pages. This is the most flexible option. **Ad variations:** Test changes to ad text across multiple campaigns simultaneously. Useful for testing headlines, descriptions, or URLs at scale. **Video experiments:** Test different video creatives in YouTube campaigns.
- 2
How to Set Up a Custom Experiment
Go to the Experiments tab in the left menu. Click the blue "+" button. Select the campaign you want to test. Choose "Custom experiment." Name your experiment clearly (e.g., "Test Target CPA vs Maximise Conversions — Plumbing Campaign"). Select the change you want to test. Set your traffic split — 50/50 gives the fastest results. Set start and end dates — run for at least 2-4 weeks to gather statistically significant data. Click Create.
- 3
What to Test
**Bidding strategies:** Target CPA vs Maximise Conversions, Manual CPC vs Smart Bidding, or different target CPA amounts. **Landing pages:** Send half the traffic to page A and half to page B. This tests the full funnel, not just the ad. **Match types:** Test broad match with smart bidding vs phrase/exact match. **Ad copy themes:** Test different value propositions or calls to action.
- 4
Reading Experiment Results
Google shows performance comparisons between your original and experiment campaigns with confidence intervals. Look for statistically significant differences (indicated by blue or green indicators). Key metrics to compare: conversion rate, CPA, ROAS, CTR, and conversion volume.
- 5
Best Practices
Test one variable at a time. Run experiments for at least 2 weeks. Do not adjust either campaign during the test. Use a 50/50 split for fastest results. Have enough budget for both arms to generate meaningful data — each arm needs at least 100 conversions for reliable results.
Frequently asked
Questions & answers
Can I run multiple experiments simultaneously?
Yes, but not on the same campaign. Each campaign can only have one active experiment.
What happens when the experiment ends?
You can apply the winning variant to the original campaign, or simply pause the experiment and make changes manually.
Was this guide helpful?
Your feedback helps us improve our guides