I made a post in a forum the other day, posing a small conundrum about which Google AdWords should be kept and which should be discarded based on the CTR figures which each ad managed to achieve.
And it's interesting how different people measure the success, or otherwise, of their Google AdWords campaigns. I'm always advocating testing, but there's to be a few caveats when you run tests.
The example which I gave in the forum was as follows, and was based on an actual campaign that I had managed for a client in the past.
Imagine you have an AdWords group with two ads running with optimization turned off, ie Google AdWords will just distribute the search traffic uniformly between your two ad creatives. After 160 you can see that Ad A has a CTR of 6.8% and Ad B has a CTR of 1.2%
Now the question is, which ad do you run with? Which one do you ax?
When the client saw the stats he immediately said we should remove Ad B because it was not performing well enough.
Oh dear.
Obviously he had not listened to a word I had said about how to manage a successful AdWords campaign. Ho hum.
Yes, CTR is important, but not as important as ROI. Return On Investment.
What's the point in buying a ton of traffic if all it does is burn a hole in your wallet? You need sales to run a business, not clicks.
"Oh yes Well what's the ROI figures then?" he asks, as if he knew this was what he meant to say all along.
Well, Ad A had a conversion rate of 0.9%, whereas Ad B had a conversion rate of 9.4%
"Delete Ad A immediately," was his suggestion.
Is this what you would have done? I did not.
The fact that the Conversions did not actually have a significant monetary value was not the point, because anyone who runs their own list knows that it's not always about the initial sale that's important, but the value of a customer over a whole lifetime.
Looking at it on the surface, and now that I've revealed that ROI is more important than CTR, you would expect to delete Ad A and run with the Ad B because of its high conversion rate.
Well, look again, my student.
You see, to read anything into the adverts that we test day in and day out, you always have to ensure that you make decisions that are based on statistical significance.
In the example, extrapolating the figures, you'll realize that we have the following:
Ad A, CTR = 6.8%, so for 2000 impressions that's 136 clicks. Conversion = 0.9% which results in 1 conversion for 2000 impressions.
Ad B, CTR = 1.2%, so for 2000 impressions that's 24 clicks. Conversion = 9.4% which results in 2 conversions for 2000 impressions.
3 conversions for our test. Is that a significant number of results for us to base our findings on?
NO.
Nowhere near enough data.
For a * minimum * we need at least 5 measurements in each group, that's 5 conversions for each AdWords Ad creative, before we can even start to run signal testing.
Of course, running Taguchi methods for multivariate split-testing would deliver a significant improvement over the simple A / B split test.
The benefit? It allows you to take the elements of the test which provide the most significant influence on the result. And in the example, it was neither Ad A nor Ad B that came out as the winner, but a mixture blend of the two.
Is not science wonderful?
I just wish I had stayed awake longer during my stats lessons at school so that I did not have to pay a ton of money on learning it all over again.
So just remember, when you come to measure your ad performance, make sure you have enough data to come to a statistically significant conclusion. It'll save you a ton of cash, I promise.
© Copyright Howard Sandford, Fast Improvement 2006. All rights reserved.
Source link :
Google Adwords – Do not Make Assumptions You Can No Longer Afford
0 comments:
Post a Comment