PPC advertisers face two common problems when they try to tackle split testing: no ideas, or too many ideas.
If you have no ideas about what to split test, that’s a sure sign that you’ve been spending too much time in your AdWords account. Log out, browse a magazine rack, check out your competitor’s websites, watch a movie, or get some exercise. Shake off the mental cobwebs and look for serendipity and inspiration.
If you have too many ideas, however, you can be too overwhelmed to run any tests. Or you may discover yourself testing an ad that you already tried eight months ago, and forgot about. Or you may run 16 ads simultaneously and grow old waiting for insight.
What you need when you have too many choices is a system for filtering and prioritizing those choices. I respectfully submit my “Who, What, Why, How” testing plan for your enjoyment and edification.
Testing Is About Wondering
If you hated high school science class, you may not like the whole idea of testing. So let’s try a quick reframe: testing is just wondering about stuff and figuring out ways to get answers.
My friend and mentor Perry Marshall remarked once that almost all marketing problems are symptoms of not knowing something. If your AdWords ads are not attracting enough visitors to your website, that’s because you don’t understand something important about who your prospects are, what they want, why they want it, how they want it, or more generally, how many or how few of them are there in the first place.
If your ads attract lots of visitors but few leads and sales, that indicates you’re either targeting the wrong searchers with your keyword/ad combinations, or you’re somehow turning them off on your website.
The healthy response to either of those problems (and hundreds like them) is, “I wonder what I don’t know here.” Split testing is simply a way to subject your best guesses to a rigorous market reality check.
Start Broad, Get Narrow
Once you let yourself off the hook and simply see yourself as a bumbling, curious Columbo-like marketer, you’ll probably be able to generate lots of split test ideas.
Now all you need is a testing framework to answer the biggest questions first, and then drill down into the minutiae.
Without a framework, you’re likely to spend a lot of time testing things like “comma vs semicolon” or “half price vs 50 percent off.” Sure, those are important distinctions, but not relevant until you’ve established the big picture. That big picture starts with the question, “Who is my prospect?”
Who includes two dimensions, global and situational. Global who possibilities encompass demographics (age, sex, location, job, income bracket, education level, etc.) and psychographics (values, lifestyles, attitudes, and interests).
For example, if you sell motorcycle riding gear, you could be catering to many different groups: 43-year-old suburban male accountants who ride Harleys on weekends, 20-year-old rural high school graduate who can’t afford to gas up the pickup truck, or a 36-year-old urban female bike messenger, among others.
The situational who dimension relates to the search at hand. Someone searching for [motorcycle pants] might be getting their first motorcycle and have no idea what type of fit or material is right for them. They might be a seasoned street rider looking to take up moto-cross. They might be the bike owner or the passenger.
Your first testing step is to write ads to appeal to each different potential market segment you want to attract and service. “Look good on your HOG” and “Extreme Motocross Leg Protection” will attract one group and repel all the rest. When you run these ads simultaneously you can compare the size and buying passion of different potential customer groups and focus on your most profitable market.
Once you’ve identified a who, you start wondering what they want. Let’s take Harley owners. Are they looking for denim or leather? Harley-branded MotorClothes® merchandise or some other brand, or no brand in particular? Full-on pants, or just chaps over their jeans?
If you started with what by just assuming the who, you wouldn’t know if your what categories were relevant. Once you determine a who, the what choices may become obvious.
You can test the what by making different offers: chaps vs. pants, denim vs. leather vs. heavy-duty polyester vs. all of the above.
Let’s say your Harley riders don’t need to pay a premium for Harley-branded clothing, but still want to look like Harley riders. What’s their motivation? Do they want to prevent bruises, cuts, and injuries? Or to feel more comfortable in the seat, either through more padding, softer material, more breathability, or more warmth? Do they want to look tough, or sexy, or accomplished, or rebellious and a tad dangerous?
You can elicit the why through your choice of words and personality, or “voice,” of your ad. Harley owners react emotionally to the sound, smell, and look of their machines. You can write text to evoke the roar of the engine, the looks they get from “civilians,” or the swagger they feel when they fit in perfectly with their HOG pals.
Finally, we get to the how level, the arena of details. Do they want multiple colors, or just black? Overnight shipping or free shipping? Do they want to buy from a shop with 2,000 items in stock, or a boutique motorcycle apparel website that stocks just 15 top-of-the-line items? Do they want a big discount, or a full-price experience with great customer service and personalized help?
Testing your how possibilities is easy in AdWords – in fact, that’s where most split tests begin and end. But in order to have the most impact, save your how tests for the end of the testing cycle.
If you’re old enough to have owned an analog radio, you might remember the two dials of the FM tuner: the main tuner, with a heavy weighted knob attached to a flywheel; and the lighter bandspread tuner, used for fine tuning. Your AdWords split tests consist of four tuners, the who, what, why, and how knobs. Start with the weightiest question of who, and then successively fine tune your market knowledge with ever more precise degrees of tuning.