I've had this study on the back burner for ages and finally decided it was time to check it out. Here's the deal... Tons of marketing "gurus" tell you that the best way to sell something is with a sales letter. I'm not disputing that (at least not in this study). These "gurus" then tell you that you should set up a one-page web-site that simulates a piece of white paper against a _____ background. That blank is what has been bothering me for awhile. Here's why...



I keep seeing different colors for major marketers. I even see some major marketers switch colors from one site to another. It makes me wonder if they have done any testing. Does the color just not matter? If so, why do so many use dark blue, light blue and black?

Then I consider the major money makers on the 'net and wonder why only one uses a colored background. I'm talking about Yahoo, Google, Ebay and Amazon. They all use a white background surrounding their "page" that is set in a limited width table in the center of the page. Only MSN.com uses a medium blue background.

It was high time for a study to find out the real answer. I did it in the usual way. I built a list of profitable and unprofitable sites by looking at the duration of ads being placed on Yahoo (Overture) and Google. If an ad had been shown for six months or longer, I placed the target page on a list of "profitable" sites. If an ad was placed and completely dropped in less that one month, the site was placed on an "unprofitable" list. The latter list would be less reliable, but the comparison between the two lists would still be statistically valid with a sufficient sample size. The latter list might only represent "average" sites, but even "average" sites would be less profitable than the known profitable list.

The results were very mixed. First of all, the white background sites were the vast majority on both lists. You simply can't go wrong with a completely white background. However, it wasn't actually the winner. Here is the breakdown:

Profitable sites with a white background: 85.2%

Unprofitable (or average) sites with a white background: 92.6%

A non-white background actually was more profitable a higher percentage of the time. I then studied the "darkness" of the background in those minority sites for both profitable and unprofitable sites. Looking at the RGB values, I split the colors into two buckets... those that were dark (191 or less average RGB value) and those that were light (192 or higher RGB value). The results were even more clear:

Looking only at non-white backgrounds of profitable sites with dark colors: 92.8%

Looking only at non-white backgrounds of unprofitable sites with light colors: 96.1%

Wow! That's quite a correlation. Although a vast majority of profitable (and unprofitable) sites have a white background, those that do have a non-white background have a dark background whereas the unprofitable sites were more likely to have a light background... both with an extremly high correlation. That's hard to ignore.

I next had to get the actual answer. What color background had the highest correlation with profitability. What's the bottom line? The answer was just as clear:

53.3% of the profitable sites with non-white desktop background were black.

Now; I should be very clear about this result. The area under study is NOT the actual web-site text area. We are NOT talking about having white text on a black background. In a vast majority of cases, these sites actually had black text on a white background. The area under study is the theoretical area AROUND the virtual piece of paper. It is the "desktop" color if you pretend the web-site area is a piece of paper sitting on a virtual desktop. It isn't the color of the piece of paper that is under study. It is the color of the virtual desktop that the piece of paper is virtually laying upon.

So what am I going to do about this result? I'll probably eventually implement it. I don't like it, but the result is very clear. Why don't I like it? Well, take a look at one of my sites:

http://www.Glyphius.com

I use a virtual piece of white paper on a virtual white desktop. I like the fuzzed shadow I use around the edges. That fuzzed shadow won't be possible with a black background. The shadow will simply disappear leaving (in my opinion) a less professional look of a crisp piece of white paper against a black backdrop.

The other reason I don't like it is that it puts me out of the majority of profitable sites in general. It will make future studies more difficult. Instead of simply grabbing profitable sites and comparing them to unprofitable sites, I will need to grab only profitable and unprofitable sites that have a black background. Otherwise other results might not be valid if they come from the majority white background dataset. A good example would be a study of headline color. The most profitable headline color might be different between those sites with a black background and those with a standard white background.

I simply haven't decided what to do yet. Will I stick with the safe standard white desktop behind my virtual piece of paper? Or will I go with the black desktop that is less standard, but has a higher correlation with profitability? What will you do?

One result from this study is very clear though. If you do choose a non-white background, it is foolhardy to choose yellow, blue (light or dark), pastel green, or any other color. The correct color for highest profitability is clearly black. The 2nd place color (and the color of the vast majority of sites) is white. All other colors failed miserably in this study.