Skip to content
Paid Search (PPC)

Using Drafts and Experiments In Google AdWords

Optimising a Google AdWords campaign can involve large amounts of data analysis, time spent, and an element of trial and error. However, you can use the campaign drafts and experiments tools within Google AdWords to test changes to campaigns, easily measure results, and apply any changes that have been identified as working well.

Published on:

Jonathan 'Cat' Catton

Head of TechOps

Campaign Drafts in Google AdWords

Campaign Drafts allow you to create and review changes to a campaign before they are implemented. This can incorporate any change that you wish to make to a campaign before it’s made live within the account. If that sounds a bit like AdWords Editor then you’re right, it essentially encompasses the same need. Hence why drafts aren’t used that commonly. However, there are still some features that you might want to draft that are not available in Editor. For example, you can only copy-paste Ad Schedules in Editor, whereas you could plan out your whole Ad Schedule in Campaign Draft, rather than in a live campaign.
To make a new draft, you need to navigate to the “Drafts & Experiments” tab in the Campaign left nav:
How To Find Drafts & Experiments
Draft mode essentially looks identical to the normal interface, and you can make as many changes to a campaign as needed before pushing it live.  The main downside is you’re making changes in the normal interface: you can’t download the draft into Editor to make bulk changes or .csv imports, which limits its use. Once the changes have been completed, click apply for the ads to go live. You can then “run an experiment”, which is the most common use of drafts.

Experiments for Google AdWords

Experiments allow you to test how the changes made within the draft will impact the campaign within a semi-controlled environment. Say we’ve developed a new landing page for a campaign and we want to confirm that it performs better than the old one. We could have just copied all the old ads and change the landing page, to compare the results for a one month period. But that’s not really comparing like-with-like, and it’s highly likely to deliver inaccurate results. The slightly better technique would be to duplicate all your ads and have the duplicates run the alternate landing page. This is much better than running them back-to-back as you’re testing both simultaneously. However, it can be very tricky to manage, requiring usage of labels and/or pivoting results in Excel to pull out results.
Instead, you can just run an experiment to send 50% of your auction entries to the main campaign, and 50% to the experiment (or whatever % split you want to use). This is easy to use, keeps your account neat, and the results are easy to obtain.
Experiments allow you to test a number of element changes within your campaign. Some of the more obvious things that you could test include:

  • Adding additional keywords
  • Raising keyword bids
  • Implementing new ads
  • Experimenting with new ad placements
  • Adding bid modifiers
  • Testing other significant changes in a controlled and structured way

Measuring the Results

Perhaps even more so than with any other marketing activity, it is essential to understand how the experiments have impacted your results. With campaign experiments, it is possible to view results in two ways:

  • The Performance Scorecard: The performance scorecard makes it very easy to compare the experiment campaign in comparison to the original. A good feature of the performance scorecard is that it will also outline whether the results should be considered significant.

In this example, there was no significant change from experiment to original


  • Evaluate Ad Group Performance: While the performance scorecard is useful for top-level statistics, sometimes it’s necessary to review individual ad groups within the campaign, to determine how the performance of the ad group impacts the campaign overall. Monitoring ad group metrics will give you another level of insight into the impact experiments have had on the campaign.

Once you are happy with the experiment results, it is possible to quickly apply the changes from the experiment into your existing campaign. If necessary, you can also convert the experiment campaign into a brand new campaign.

Benefits of Using Campaign Experiments in AdWords

Campaign experiments allow you an easy way to test changes to campaigns methodically. Some of the main benefits include:

  • You can choose the proportion of the budget (or clicks) to split into the experiment campaign.
  • You can choose the length of the experiment. The length of time you test for will vary depending on how much data you regularly get. The important thing is to make sure you get enough data to make it significant. Definitely anything shorter than two weeks is too short. Three months is probably a decent upper bound for length on the other end. So, something of that scale would probably be appropriate.
  • It is possible to end an experiment early if your advertising or business objectives change significantly.

Campaign Experiments and Quality Score

An interesting element of the ability to carry out campaign experiments is that the performance will isolate any changes in quality score to the experiment. That is to say, the keywords in the experiment will have a changed quality score, but the original will stay the same. This makes it quite a good platform for testing changes that you’re making to improve the quality score. Should you chose to swap out the original for the experiment, it will then run off the quality score that was used in the experiment.

Tips and Tricks For Using Experiments and Drafts

Whilst you can’t edit drafts in Adwords Editor, you can edit experiments in there. Therefore, if you want to test out a large change, what can often be a very effective tactic is to create the draft and then immediately make it an experiment, but then pause it. Now you’ll be able to edit it in AdWords Editor to your heart’s content.
Don’t end the experiment early because you think you already know the victor. Even if the experiment says it has a statistically significant result, you should always wait for the time on the experiment to finish before making a judgement. The reason for this is long, complicated and very mathsy. But if you do so, you bias the results and increase the likelihood of a false positive in your experiment.
For help with your PPC campaigns, get in touch with us today.

Get content like this in your inbox, weekly.