Here’s a common AdWords scenario: your campaigns are profitable but you wonder whether it makes sense to pay a little more per click, move into higher ad position and generate more traffic and conversions. Most advertisers would just increase the bids, monitor performance for a week or two and then compare the before and after results.
The problem with this approach is that your cost per acquisition varies week by week independent of the bid change you made. There might be factors other than your bid change during those two weeks that increased or decreased the cost per acquisition.
The solution is to this problem are AdWords Campaign Experiments (ACE). ACE allow you to split test campaign changes in real-time. For example, you could let AdWords run one of your keywords with a bid of $1.00 for 50% of the searches and with a bid of $1.20 for 50% of the rest.
Read the Adwords Campaign Experiments documentation for more details.
Click on an Adwords campaign -> Settings -> Experiment (Beta) -> +Experiment Settings
– Type in a name for your Adwords experiment
– In most cases it’s best to keep the split 50% / 50%.
– If you are testing a risky strategy that could significantly increase cost you should allocate a higher percentage to your control, e.g. 80% control / 20% experiment. That way your experiment has a lower impact on your average campaign performance but you can still evaluate the difference over time.
– Choose “No start date (I’ll start it manually)”
– Choose “30 days from start” as your end date. You can always stop your experiment early.
– Click “Save”.
At this point you are not ready to launch your test yet because you haven’t setup any “experimental changes” yet.
The next step is to go back to your campaign and setup your experimental changes.
In this example we will increase the bid by 20% for a high traffic keyword.
– Select your experimental campaign
– Click on the keywords tab
– Click on “Segment” and choose “Experiment”
The keywords are now broken down by “Control” and “Experiment”.
In order to make an experimental bid change for a keyword, click on the “Max. CPC” cell of the keyword in the “Experiment – 50%” column.
Enter the percentage increase (or decrease) you want to test for the keyword and click “Save”.
Once you launch your campaign experiment, AdWords will now run your keyword with a higher CPC bid for 50% of your impressions and with your standard bid for the other 50% of your impressions.
To launch your experiment, go back to “Settings” and click on “Start running experiment”.
– In order to monitor your experiment go back to your campaign and click on the keywords tabs.
– Click on “Segment” -> Experiment to analyse the performance of your control and experiment
– The screenshot below shows an example of a keyword that was tested with a 10% lower bid.
– Blue arrows next to a metric show that statistical significance has been reached. In the example below this means that we can be sure clicks, impressions, CTR, CPC and cost will decrease if we choose to implement the experiment settings.
– There is not enough data yet to determine what the effect on conversions, cost per conversion and conversion rate will be
– Make sure that you run each experiment long enough until your key metrics reach statistical significance (i.e. a blue arrow next to the metric). It’s tempting to pause these tests early because you “think” you know the results. You don’t, it’s a mistake to pause them early.
