Google Ads Drafts and Experiments Explained: How to Test Campaign Changes Without Risking Your Budget

Google Ads experiments let you test a change against your existing campaign by splitting traffic and budget between the original and a modified copy, so both run at the same time. Drafts were the older way to stage those changes before testing, and the term still shows up in guides, old accounts, and plenty of PPC conversations.

This article covers how drafts and experiments relate today, how an experiment works under the hood, what is worth testing, how to read the results, and the mistakes that quietly ruin a test. Interface details here reflect Google Ads as of October 2026, and Google changes menu names often, so check Google Ads Help if something looks different in your account.

Drafts vs. Experiments: What Each One Does and How They Relate Now

A draft was a staging copy of a campaign. You edited settings in the copy (bidding, targeting, ad copy) without touching the live version, then either applied the draft to the original or turned it into an experiment. An experiment is the test itself: a changed copy (the trial) runs alongside the original (the base campaign) at the same time, and Google compares how each performs.

Here is where the confusion comes from. The old "Drafts & experiments" label has largely been folded into a dedicated Experiments page, reached through Campaigns > Experiments in the left menu as of October 2026. Instead of building a draft first, you typically start a "custom experiment" and set up the trial version in that flow. The staging step still exists in spirit, but it is no longer a separate feature you have to manage. If an older tutorial tells you to open "Drafts & experiments," you will likely land in the Experiments page and follow a slightly different set of steps. Google's naming and menu paths do shift, so confirm against Google Ads Help before you walk someone else through it. For a hands-on walkthrough, see our guide on how to use Google Ads experiments.

So the practical answer: you do not need to "use drafts" as a standalone tool. You need to run experiments, and the setup flow handles the staging.

Why not just edit the live campaign?

Because a before/after comparison is not a test. If you change your bidding strategy on the 1st and compare the next two weeks against the previous two, the result mixes your change with seasonality, competitor activity, news cycles, and any budget shifts. You cannot tell what caused the difference.

A split test avoids that problem. Both versions face the same auctions, the same week, the same market conditions. The only meaningful difference is the thing you changed.

How a Google Ads Experiment Actually Works

The mechanics are simple. You pick a base campaign, create the trial version with your changes, set a traffic split, and choose a start and end date. A 50/50 split is the common default because it gives both sides equal data, but you can often choose other splits. A smaller trial share lowers risk and also slows down how fast you learn.

Split types and why they matter

A traffic split is the percentage of eligible traffic and budget each version receives. How that split is applied matters just as much as the ratio:

• Search-based split: each search is randomly assigned to one version. The same person could see both versions on different searches.
• Cookie-based split: users are randomly assigned to one version and keep seeing it, so each person is exposed to only one arm of the test.

Which options you get depends on the campaign type and experiment type, so verify current availability in Google Ads Help. The reason it matters: if the same user can bounce between versions, or if the groups differ systematically, the comparison gets muddy. A clean, random split is what lets you attribute a gap in results to your change rather than to who happened to see which version.

Budget behavior

The experiment does not come with new money. The trial takes its share of the base campaign's budget according to your split, and the base keeps running with the remainder. Only the trial carries your change. Total spend should stay in line with what the campaign would normally spend.

An illustrative walkthrough

Suppose you run a Search campaign for a local HVAC company on Maximize conversions, and you wonder whether a Target CPA (cost per acquisition) bid strategy would bring leads at a lower cost. You create a custom experiment, keep everything identical except the bid strategy, set a 50/50 split, and schedule it for several weeks. The base keeps Maximize conversions. The trial uses Target CPA with a target set near your recent average. At the end, you compare conversions and cost per conversion between the two. This is an example of the setup, not a result you should expect.

What Is Worth Testing (and What Isn't)

Good candidates are changes where the outcome is genuinely uncertain and the downside of being wrong in the live campaign is real:

• Bidding strategy changes, such as manual CPC to Target CPA or Maximize conversions to Target ROAS.
• Match type changes, such as moving keywords from phrase match to broad match alongside smart bidding.
• Ad copy and assets, including new headlines, descriptions, or creative.
• Landing pages, where the destination changes but the ads stay the same.
• Audience settings, where supported.

Which of these are available depends on the experiment types Google offers for your campaign type, so check the options in the Experiments page rather than assuming. If you are weighing keyword-level changes, our post on managing keyword experiments in Google Ads goes deeper, and if you are testing destination pages, start with landing page optimization for Google Ads.

What to skip

Poor candidates fall into two groups. The first is changes that touch many variables at once: new bidding, new ads, and a new landing page in one trial. If the trial wins, you will not know why, and if it loses you will not know which piece hurt. The second is tiny tweaks, like swapping one word in a description, on a campaign with low traffic. The difference will be too small to ever reach a clear answer.

The rule: one variable per experiment. It is slower, but you finish with a conclusion you can reuse.

Keywords, search terms, and maintenance

Keep a distinction in mind. A keyword-level change, like loosening match types, is a test of how you match queries to your ads, so it deserves an experiment. Search term cleanup is different. Adding negative keywords for irrelevant queries is routine maintenance, and you would usually apply it directly rather than test it. For that kind of upkeep inside the Search Terms Report, Keywordme lets you remove junk terms and apply negatives and match types in a few clicks without leaving Google Ads. It does not run experiments for you. It just keeps the maintenance side quick. If you are not sure how the two differ, read search terms vs. keywords in Google Ads.

Reading Experiment Results and Deciding What to Do

Open the experiment from the Experiments page and look at its results, often called the scorecard. It lays the trial next to the base on metrics such as conversions, cost per conversion, and conversion value. As of October 2026 the layout lets you pick which metrics to compare, but verify the current report in Google Ads Help since Google updates it.

Statistical significance in plain language

Google marks whether a difference between the two versions is statistically significant, usually alongside a confidence level. Put simply, a significant result means the gap is unlikely to be random noise. A higher confidence level means you can be more sure the difference is real.

The flip side matters more. A trial that is ahead but not significant is not a win. It is a lead you cannot trust yet. Early in a test, numbers swing widely, and the leader on day four often is not the leader on day twenty.

How long to run it

Do not rely on a fixed number of weeks from a blog post, including this one. Google publishes guidance on recommended duration, and it varies with campaign type, traffic, and conversion volume, so check Google Ads Help. Two principles hold broadly. Bidding tests need extra time because smart bidding goes through a learning period after a change. And you need enough conversions on both sides for the comparison to mean anything, which makes low-volume campaigns slow to resolve.

Deciding what to do

When the test finishes, the options as of October 2026 are generally:

• Apply the trial to the base campaign, so the winning changes replace the original settings.
• Convert the trial into a new campaign, if you want to keep both running separately.
• End the experiment, leaving the base untouched, if the trial lost or the result was inconclusive.

Confirm which options appear for your experiment type. An inconclusive result is still information: the change probably is not a big lever for this campaign. If the campaign is still underperforming, why your Google Ads campaign is not converting can help you decide what to test next.

Common Experiment Mistakes That Ruin the Test

Most failed experiments fail because of how they were run, not because the idea was bad.

• Changing the base campaign mid-test. If you adjust the base budget, bids, or negatives while the experiment runs, the two arms stop being comparable. You are then measuring your edits as well as your test.
• Ending too early. One side looks ahead after a few days, and you declare victory. Early gaps are often noise, and bidding changes need time to settle.
• Too little conversion volume. With only a handful of conversions per arm, a couple of lucky leads can swing the whole result. Trusting those numbers is worse than having no test.
• Ignoring conversion tracking quality. An experiment confidently measures whatever your conversion tracking reports. If tags double-count, miss form submissions, or count low-value actions, the experiment will pick a winner based on the wrong signal.

More broadly, these errors are part of a larger set of common mistakes to avoid in Google Ads optimization.

A practical habit

Keep a simple change log for the base campaign: date, what changed, why. If something needs to change mid-test, such as a budget cap or an urgent negative keyword, you will at least know when it happened and can interpret the results accordingly. Google also offers an option to apply certain base campaign changes to the trial so the two stay aligned. Check current behavior in Google Ads Help, because which changes sync and how can vary. If you use it, do it deliberately and write it down.

Agencies running tests for clients have one more job: tell the client before the test starts that the base campaign will be frozen on the tested variables, so nobody makes well-meaning edits halfway through.

Drafts and Experiments FAQ

Do I still see "drafts" in Google Ads?

Much of the old "Drafts & experiments" workflow has moved to the Experiments page, with a custom experiment flow replacing the separate draft step. Older accounts, help articles, and third-party guides may still use the old wording. Check Google Ads Help for the current terminology in your account.

Can I run experiments on Performance Max or Shopping campaigns?

It depends on the campaign type and the experiment type. Google has added and changed experiment options over time, including for campaign types beyond standard Search. Look at the experiment options in your Experiments page and the supported-types list in Google Ads Help rather than relying on an older guide.

Does an experiment cost extra?

No separate fee. The trial's spend comes from the split of the base campaign's budget, and you pay for clicks as usual. What can cost you is a trial that performs worse, so a smaller trial share limits exposure.

How many experiments can run at once on one campaign?

Google sets limits, and they can change, so check Google Ads Help for the current number. Even when you can run more than one, avoid overlapping tests on the same campaign. You will not be able to separate their effects.

Test One Change, Wait for Significance, Then Apply the Winner

That is the whole method. Keep a live base campaign running, change a single variable in the trial, let the test run long enough to reach a significant result, and only then apply the winner. Everything else in this article is a way of protecting that sequence.

Your next step: pick one assumption in your top campaign that you have never actually tested. Maybe it is that phrase match beats broad, or that Target CPA beats Maximize conversions. Set it up as a custom experiment and leave the base alone.

While it runs, the routine work still needs doing, especially search term cleanup. Keywordme lets you remove junk search terms, build high-intent keyword lists, and apply match types instantly, right inside Google Ads. No spreadsheets, no switching tabs, just quick, seamless optimization. Start your free 7-day trial (then just $12/month) and take your Google Ads game to the next level.