This 4:41 video tutorial from Straight North PPC Manager, David Ohm, takes you through how to set up a Google AdWords bid experiment to get the most out of it. Must-view for PPC specialists unfamiliar with the ins and outs of this extremely useful tool.
Hi, today we're going to go over setting up a Google AdWords experiment, specifically a bid experiment. Let's say you want to apply a bid to a given ad group or a group of keywords, but you're not sure how that bid will affect your key metrics such as clicks, impressions, click-through rate, average position, conversions, whatever is important to you.
So what we're going to do here is we're going to take that new bid that you suspect, you have a hunch, that it might help your campaign. We're going to take that new bid and only apply it 50% of the time. What the experiment is going to allow us to do is apply that new bid 50% of the time and our original bid, kind of as a control, the other 50% of the time. Okay? So this kind of gives us a backbone or a safety net when we want to apply that new bid that we're kind of unsure about how the results might come out.
All right. So I'm at the Ad Group level here, and I'm going to apply this to my branded campaign. I'm going to go into Settings. Go down. The first thing I'm going to check is my bid strategy. You want to make sure it's on Focus on Clicks, manual CPC, nothing automated within Google going on other than the experiment itself. You're good there and Save.
Then Experiment, and I'm going to set up the experimental settings. Name it so you'll remember what it's doing and what the experiment actually is, so Plus 10% Bids in this case, and notice the Control/Experimental Split, I have it set at 50/50 right now. If it's more of an aggressive bid and you're not exactly sure how it's going to come out, you're worried that while it could be a potentially positive impact, it might have an equally catastrophic negative impact, you might want to go more on the 60%, 70%, 80%, 90% control, so that way your original bid will still show most of the time, but you'll still get some initial data on that new bid that you're anxious to see if it works or not.
Okay. So I'm going to leave it at 50/50 right now and then No Start Date. I'll just start it manually myself after I've set it up, and then the End Date will automatically be 30 days. You can set this to whatever you like, though, if you want to.
Okay, so I'm good. I'm going to save it, and notice down here it still says "Start Running Experiment." Until I click that, it will still stay not started there.
All right. Go back to the Ad Group level. All right, and now this little triangle indicator indicates that I have started up an experiment. It's not actually started yet, I apologize for the wording there, but I set up a new experiment.
Now I'm going to go to Segment and divvy up by experiment. So now we can see Outside Experiment within this date range, so our original data, and then you're going to have from the start of the experiment, the Control and the Experiment kind of divided out here. So I'm going to go to the Experimental one. I can't click at all on the Control one. But the Experiment, I'm going to click on it. I'm going to add 10% here. So the experimental bid is $1.10 versus the original which was $1.00. Pretty extreme, I know.
So now I set up my parameters, what I'm going to do is go back to Settings, go down here, and now I'm going to press Start Running Experiment. See, now it just gives me the indicator right here that the experiment is good and it's running. 50% of the time a 10% bid increase has been made to every keyword in that ad group.
Whenever the experiment is done, it will automatically stop running, keep in mind, after 30 days if you left that setting on. However, whenever you're done with it, say after 20 days you have statistically significant data, you might want to stop the experiment early. So let's say 20 days have passed, and I stopped the experiment.
Now it gives you the option to apply and launch the new experimental bid across all the keywords permanently, or remove, go back to the control. Because this was an experimental experiment, if you will, I am going to remove and go back to my original.
Keep in mind, once you remove the changes you will no longer be able to see the data split up in the way that I just showed you at the Ad Group level. You will no longer be able to segment it. So if you want a report on it, I advise taking a screenshot or downloading that page before you remove or apply the changes.
I'm going to remove, and the experiment was, in this case, I guess not a success because I did not . . . the experiment was a success, but the experimental bid was not because I'm taking it away. Obviously, if it had done something positive, I would have left it.
I hope this has been informative and helpful to you. I hope you can appreciate exactly why you might use a bid experiment in this fashion. It's very easy to report on, you have a lot of control, and it's just a very helpful tool to kind of test bids out.
Thank you. Have a great day.