How to run Google Ads experiments without wasting your budget
Google Ads Experiments (formerly Campaign Drafts and Experiments) lets you test a change on a portion of your budget before applying it to the whole campaign. For B2B SaaS founders, it's the right way to test bidding strategy changes, landing page variations, and keyword match type changes without risking your full spend.
Google Ads Experiments is a feature that lets you test a campaign change — a different bidding strategy, a new landing page, a keyword match type adjustment — on a defined percentage of your traffic, running in parallel against your original campaign. The control and experiment run simultaneously, audiences are split randomly, and you get a direct comparison of performance before deciding whether to apply the change to your full budget. Most solo founders have never used it because it's buried in the interface.
Where do you find Google Ads Experiments?
In Google Ads, go to Campaigns in the left navigation → click the three-dot menu or look for "Experiments" in the navigation panel (the exact location varies by interface version, but searching "Experiments" in the search bar at the top of Google Ads finds it directly). Create a Campaign Draft first — a draft is a copy of your campaign where you make the change you want to test. Then create an Experiment from the draft, setting the traffic split (typically 50/50 or 70/30 control to experiment).
What should B2B SaaS founders test first?
Bidding strategy: test moving from Manual CPC to Maximize Conversions (or vice versa) with a 50/50 split. This is the highest-risk change to make without an experiment — the bidding strategy affects every auction. Landing page: if you're building a new landing page, run an experiment splitting traffic between old and new. You get real conversion data before committing. Keyword match type: test phrase match vs broad match on your top keywords by putting the broad match version in the experiment. Target CPA bid: test a lower or higher target CPA to see how it affects volume vs cost.
How long should you run an experiment?
Long enough to reach statistical significance — typically 2–4 weeks with enough conversions. Google Ads shows a significance indicator in the Experiments report. "Significant" means the performance difference is unlikely to be random. Don't call an experiment early because it looks good at day 5 — small sample sizes produce misleading results. A minimum of 100 conversions per variant (200 total) is a reasonable threshold for meaningful data. For lower-volume campaigns, run experiments for a full 30 days regardless of conversion count.
What do you do after the experiment concludes?
If the experiment wins: click "Apply" to apply the experiment changes to the original campaign — this replaces the original settings with the winning variant. If the original wins: end the experiment and your original campaign settings are unchanged — the experiment ran on a split and didn't affect your main campaign. If there's no significant difference: you've confirmed the change doesn't help or hurt, which is useful information. Experiments let you make data-informed decisions instead of guessing what will improve performance.