Channeling the Power of a Blind Ad Test
Let's say the marketing department at Generico Housing Insurance, Inc. wants to build an ad campaign to promote its new insurance coverage for egg damage caused by bored teenagers. It's decided to use the Googl Display Network (GDN) method to test its campaign. It creates two display ads. The copy in both is identical and only the images are different. Ad 1 shows a house being egged by teenagers. Ad 2 shows a pretty home with a nicely dressed, attractive family standing out front. The call to action in both is to click through to the website and request an insurance quote.
At first, Generico has no way of knowing which ad will generate more quotes. It doesn't matter: Google will show the ads to a few thousand people, and within a week or two, there's a clear winner. Generico now knows which image to use as it invests more advertising dollars in its website, GDN, print ads, business cards and even TV commercials.
This works because we're measuring actual physical behavior in real time -- not just verbal opinions. People are clicking their mouse and filling out forms . . . or failing to.
When they see the Generico ad online, they have no idea they're participating in a split test. They're simply responding based on whether the ad triggers them to want insurance or not.
GDN is never a branding exercise, and it's never opinion polling. It's observing your target market making live decisions and learning what excites their interest or leaves them cold.
Don't waste thousands of dollars creating huge portfolios of advertising and marketing collateral. Spend just a few hundred dollars up front testing ads, and you'll know which images, color schemes, headlines and calls to action give you the results you want.
"Test the forest" first
Our colleague David Bullock has given us a great verbal rule of thumb for testing ads and marketing campaigns: First test forests, then test trees, then test branches, then test leaves.
Test the big difference-making stuff first. Go after finer points later. How far down you drill will depend on your patience for this kind of work and, of course, your budget.
In this scenario, the "forest" is two things:
- Your headline, hook, call to action, offer or guarantee
- The overall aesthetic of your ad: color schemes, styles, layout and tone; bold versus subtle, traditional versus modern, serious versus comic, masculine versus feminine, young versus old and everything in between.
Let's say you plan to test images of people in your ads. Try different ages as well as both genders. Right off the bat, that gives you four variations: older men, older women, younger women and younger men. Which one will your target market respond to? After a few thousand impressions, you'll know.
The one thing split testing teaches you above all else is that there is no predicting what will and will not make a drastic difference.
There are so many things you can test:
- Image style: hi-res color photo, black-and-white classic, hand-drawn picture or even no image at all.
- Text formatting: color, font, size, bold or italic.
- Dominant color: Different colors and different levels of light or dark will evoke wildly different moods. If you're carefully targeting sites on GDN to show your ads, think about what color schemes will make your ads stand out on the page rather than blending into the background.
- Call to action: Ideally, this should be prominent. Even better if it's in the form of a button. We're trained to click on buttons. Your entire ad may be clickable, but if you feature a boldly colored button, people will click there more than any place else on your image.
Which type of ad should you start with?
Google's Display Network gives you a wide variety of ad sizes, but you need to stick to just one size until you've identified your winning ad style. It's simple math. If you create ten different styles of image ads -- a good number to start with -- but you attempt to test each of these in eight possible image sizes, you'll end up creating 80 different ads. That's unwieldy and a waste of time. There's little point in creating image ads for every size unless you're attempting to completely max out your impressions. Otherwise, it's a law of diminishing returns.
Related: 10 Laws of Social Media Marketing
The first size to test is 300 x 250 pixels. (This is officially called a "medium rectangle," "med rec," or "inline rectangle"). It's one of the most common sizes supported by sites on GDN, so you'll be sure to get plenty of placement. Most importantly, in our experience, it's the size that draws the highest click-through-rate, which is the key factor at this stage.
The leaderboard size (728 x 90) also gets good impressions. Clicks and conversions won't be at quite the same level as the medium rectangle, but this is often the next best size to go after. Once you've tested those, try the large rectangle (336 x 280), wide skyscraper (160 x 600), half page (300 x 600) and billboard (970 x 250).
Tip: Plan on your best ad outperforming your worst ad by a factor of three.