How To Use Google Campaign Experiments To Optimize Adwords

Changing adwords campaigns, particularly established campaigns that are working, can involve a lot of trial and error. We don’t want to ruin a good thing, yet all of us want to optimize for best results.

In August 2010, Google rolled out a helpful tool called Adwords Campaign Experiments, otherwise known as ACE.

ACE is a free tool that enables you to run experimental campaigns, side by side, against your existing campaigns. You can run tests on a percentage of your traffic, as opposed to all of it.

For example, say if you are currently bidding 50 cents on a keyword, what happens if you increase the bid to two dollars? It might be difficult to tell if any increased conversions are related to the increased bid, or some other quirk in the market. ACE will tell you if the changes you see are significant, as you’ll be comparing two different bids, run at – effectively – the same time.

ACE is similar to a slit-run test, but enables you to test settings across entire groups and ad campaigns.

You can find out more about Adwords Campaign Experiments here. Google have since rolled out Campaign Experiments globally. Now, you can test more variables, including the ads themselves.

To access Adwords Experiments, sign into your Google Adwords Account. Under the Settings tab, you’ll find an option labelled “Experiment”

How To Test Using ACE

• Establish Control Group – Your existing campaign is your control group. You can set this on the status icon on your Ad group.

Create new groups and set these as experiments:

• Change Match Types – Leave you existing campaign as a control and start an experiment to test different match types. How do different match types affect your CTR? Impressions?
• Change Ad Groups – how does adding keywords to existing groups affect performance? How does splitting groups up into smaller, more focused themes affect performance? How does removing keyword affect performance?
• Change Bidding – Shift bids up and down radically. How does bid level affect performance? Be careful with the percentage of traffic you allocate to your test. It needs to be high enough to be statistically significant, however if you set it too high, you might compromise your existing campaigns performance.
• Keyword Level URLs – How does changing the URL affect performance?
• Display Network – Change the topics you target on the display network.

Understanding The Results

As the campaign runs, you’ll see separate control and experiment rows, so you can compare data.

Probably the most important aspect of this tool is how clearly you can see if there is a statistical difference in the optimization vs chance. Whilst the numbers may change in your experiment vs control, these numbers may not mean much to your business.

Google makes this data easy to analyze by providing arrows.

• One (up or down) arrow represents a 5% probability that the metric increased or decreased by chance rather than due to your experimental changes.
• Two arrows represents a 1% probability
• Three arrows represents a 0.1% probability

Further Reading:

• Adwords Campaign Experiments – How To Use Ace To Optimize Your Paid Search Listings
• Testing With Adwords Campaign Experiments
• Five Things To Try With Google Experiments
• Google Adwords Campaign Experiments 101