Background Color Split Testing Results

A few weeks ago I wrote a post on using Google’s Website Optimizer (found inside your Adwords account). At that time I mentioned I was running a split test on a couple different background colors for a landing page I wanted to test. After only a day or two I was getting some pretty interesting results:

“As a side note, I’m currently split testing a different version of the page which at least in preliminary results, the splits are outperforming the original by 26% and 55%. The implications of a successful outcome from a split test can really make a big difference on the bottom line!”

I wanted to follow up on this to see if we can learn anything from this exercise. The first few days seemed spectacular; I thought for sure I had stumbled upon the ultimate background color. Then as more results steadily streamed in, things gradually started to equalize. In fact, the two tested colors suddenly reversed their leads! Anyways, I continued to run the test, from May 7th to June 16th. You can see the results here.

I should mention that the Original in this instance was color code #333366. Google doesn’t give you an option to edit that title. So you can see that color #0099CC came out on top, which is strange, because to my eye it seems the ugliest of the three. However, I guess that is neither here nor there. FYI, the original is kind of a dark navy blue, combo 1 is a light sky blue, and combo 2 is a light gray.

I don’t know if you can draw anything too substantial from these results, because I am sure different background colors work better with different page content as well. However, on the page that I’m using, apparently combination 1 turned out better. It is important to note that Google still didn’t come up with a conclusive answer, only that there was an observed improvement over the original of 5.86%, and that there was a margin of error of plus or minus 2%. Having run this for five weeks now, the numbers have not really been changing a whole lot, so I’ve decided this is as far as it goes. If I wanted to confirm these numbers I might try re-testing with only the original and one alternative.

You can see that there were a total of 2443 conversions during this period (I only used traffic from Yahoo PPC for this test). That gives me an overall conversion rate of about 50.3% – from there I still need to get them to apply for the offer after clicking through.

You can get some insight into how this tool works by looking at the conversions/visitors column. I think that the way they’ve built this, Google takes the assumed favorite combination and tests the heck out of it, while giving an equal share to the original. You can see that combination 2 got less visitors than the first two, but that the first two got exactly the same number of visitors. I think this is Google’s way of testing a hypothesis. First give everything equal traffic, then as soon as things start to gel a bit, pick the original and the favorite and push them farther.

Anyways, for the time being, I’m happy having found something that offers me a 5.86% improvement over what I had, for no cost. As I laid out in the previous post, anything I can do to increase this conversion ratio will net me more profit at the end of the day. I’ve already spent the money to get the visitors to come; from here on the more that click through the better.

For example, if I spend $100 on PPC to get 100 visitors, and previously got 49 to go through to my offer, and if I know that each time someone views my offer it is worth about $3.00 to me, then before I was netting ($100 – (49 x $3.00)) = $47 profit.

Now, with the new numbers, it would work out to ($100 – (52 x $3.00)) = $56 profit. That’s a $9 improvement, per day. That’s nearly $300 a month. Not bad. Of course, assuming I was actually paying a buck a click, and actually getting $3 per viewed offer – these are just made up numbers for the sake of illustration. However, it is close to reality, and the principle still stands.

I don’t recommend you all go out converting your landing page backgrounds to 0099CC because of this – but I do recommend you think of something that might improve your offer and test the heck out of it!

Sorry, comments are closed for this post.