How does Google website optimizer work

google website optimizer control script and how does visual website optimizer work
ErrolFord Profile Pic
Published Date:03-08-2017
Your Website URL(Optional)
Making More Sales with Website Optimizer In This Chapter ▶ Deciding what to test ▶ Setting up landing page tests ▶ Interpreting test results ▶ Testing continually for massive improvement n the movie Groundhog Day, Bill Murray plays a surly, cynical weather- Iman fated to relive the same day — February 2 in Punxsutawney, Pennsylvania — over and over again. Along the way, he learns to sculpt ice, play piano, and dance. Eventually, he falls in love with his producer, played by Andie McDowell. In his quest to win her heart, Murray’s character fails repeatedly to impress and attract her. But he learns from every failure and adapts his behavior, until he has transformed himself into a suave, considerate, and heroic companion. When you split-test pages on your Web site, you can accomplish the same sort of trial-and-error adaptation with your visitors and prospects, without having to spend quasi-eternity in Pennsylvania. Your tool of choice, compli- ments of Google, is the free, powerful, and elegant Website Optimizer. Website Optimizer allows you to test different variations of your Web pages to see which ones give you more of the results you want. Will your visitors respond better to a product photo, video, or testimonial in a given region of the page? Will more people buy if you offer a 50 percent discount or a two- for-one sale? Should your headline read, “Natural Soaps for All Occasions” or “You Probably Stink”? Testing, part of the DNA of direct marketing, has been practiced for well over 100 years. In the old days, advertisers would test direct mail campaigns at a cost of thousands or tens of thousands of dollars, and would get results months after initiating the test. On the Web, you can generate the same quality of data with a few minutes of setup and a few days or weeks of dataPart V: Testing Your Strategies and Tracking Your Results 296 gathering. Most site owners never test, so they never discover that they’re missing 50 percent to 100 percent increases in sales — and that range is conservative, because you won’t believe the real possibilities until you experience them for yourself. In this chapter, I borrow heavily from the expertise of Richard Mouser, testing genius and proprietor Richard launched an e-commerce site,, in 2004, and tested his way from losing money to becoming highly profitable in a very competitive market, home water filters. Richard and I guide you to create a testing strategy, set up tests, analyze the results, and take action to produce more effective and profitable Web sites. Richard should get credit for all the good information in this chapter. But all the bad jokes are mine. Deciding What to Test Website Optimizer makes it so easy to set up and run tests, you may be tempted to jump right in before doing the necessary planning and prepara- tion. So before I show you how to crank your testing engine, I talk about how to develop a strategy that gets you going in the right direction. The first step is to figure out what to test. Here’s the problem: Your landing page consists of dozens of different ele- ments — headlines, subheads, body copy, navigation bars, images, fonts, color schemes, and on and on — and each can be changed in practically infi- nite ways. If you were to test each element randomly, you’d feel like a monkey at a keyboard aiming to produce King Lear. It would take forever, you’d have no guarantee of ever getting there, and you’d develop a nasty case of simian carpal tunnel syndrome. And, you’ve already created the best site you could build. So where can you go for new ideas? Testing Principle 1: Start big, get smaller In Chapter 13, you see some examples of very tiny changes in ad copy that lead to very different results. The addition of a comma increased my CTR four times So you might think that similar small changes would work for site testing as well. Generally, they don’t. Instead, go for giant differences. That way, you’ll get clearer results faster. For example, you may want to find out if your visitors will respond better to a video or a still image at the top of the page. That’s a big difference. Do they prefer a short page or a long page? A professional-looking site or one that could have been put together by your nephew in about 15 minutes? Chapter 14: Making More Sales with Website Optimizer 297 After you answer the big questions, you can look at the nuances and details. But if your first tests are showing no difference in results, the difference between options A and B aren’t big enough to matter. Testing Principle 2: Tests are just questions in action Maybe you still have flashbacks to 11th grade chemistry class, where you had to write down hypotheses and experimental methods in your marble notebook, and then try to avoid setting your hair on fire with a Bunsen burner? If so, please relax when it comes to scientific Web site testing. Running a test is just asking a question and inviting your visitors to answer with their actions. One of my tests started with the question, “How much do the visitors to this page know about me already?” Do I need to spend time introducing myself and establishing credibility by telling them I’m my mother’s favorite AdWords author or cut right to the chase and entice them to purchase the Look Over My Shoulder AdWords Video Series? My test had to put that question into action, by creating an option that assumed familiarity and an option that didn’t. I created two pages, cleverly named Long and Short. The long page introduced me in depth before getting down to business, and the short page started talking about their situation and why watching short online videos was the answer to their prayers. Testing Principle 3: Test to overcome objections While you brainstorm questions to turn into tests, familiarize yourself to all the reasons your visitors aren’t doing what you want. Is it too hard? That is, do you have a usability issue of complicated menus, cumbersome forms, and hidden buttons? Or are visitors unpersuaded? If so, you need to work on the elements that create comfort, build credibility, and describe relevant and desirable benefits. Karl Blanks and Ben Jesson make this simple with their objection/counter-objection approach. First, you deter- mine the likely objection. Then, make changes that address that objection. For example, if the objection is, “I don’t trust you or your company enough to buy something or give you my contact information,” then you add trust elements such as endorsements; photos of your staff, office or warehouse; years in business; Hacker Safe seals, and so on.Part V: Testing Your Strategies and Tracking Your Results 298 If they can’t differentiate your product from your competitors’, then test elements that show the advantages of your product. If they’re afraid of making a mistake, introduce and test risk-reduction strategies, such as a free trial or money-back guarantee. In other words, don’t focus your testing on the elements of your site (“Red or blue headline here?”). Focus on what the users need in order to feel good about doing what you want them to do. Testing Principle 4: Look for things that don’t work Testing isn’t always about finding things that work better. Sometimes you learn more when your great idea falls flat. Instead of shaking your head and saying, “Well, that bombed,” get an inquisitive look in your eyes and ask, “Now why didn’t my visitors go for that?” That question gets you out of your head and into the heads of your visitors — the people who are interacting with your site in an attempt to achieve a personal goal. When you understand what your visitors don’t like, and what gets in their way, you can apply that knowledge to the rest of your marketing — online and off. When you test, you’re developing your instincts for what works and what doesn’t work. Over time, this is where testing really pays off. By experiencing success and failure, you gain insight into your market and become better at creating effective promotions for them. Creating a Testing Plan After you generate your testing ideas, your “I wonder . . .” questions, and the likely objections, you can build your testing plan. Richard recommends three simple steps to get you started (and hooked on) testing. 1. Make a list of things you want to test. 2. Prioritize that list. 3. Start testing and keep adding to the list based on test results. The next few sections discuss these steps in detail. Chapter 14: Making More Sales with Website Optimizer 299 Making your list of things to test You might not feel like you have any ideas that are worth testing at this point, but don’t worry. You just need to start getting ideas down on paper regard- less of whether the ideas are good. If you follow this process, you’ll end up with more ideas than you have time to test and you’ll never stop adding to your list. The most promising ideas will bubble to the top of your list, your Web site will get a little better with each test, and your marketing skills will get sharper in the process. One important thing, carry around a piece of paper or an index card to cap- ture ideas. Inspiration comes at the strangest times; you don’t want to forget a good idea. Richard’s story When I launched my first Web site, www. what I wanted. I was disappointed, but I let the, I was just seeing design stand because I was in a hurry to take whether I could successfully drive traffic to an the site live. But when I got around to testing my e-commerce site by using AdWords. I didn’t original concept, sales increased. So good-bye want to spend a lot of money on the site. One artist’s rendition. thing I skimped on was security, or credibility Sometimes testing went the other way. Iknew seals. The seals were expensive at the time, so that making the first part of each bullet point I didn’t want to spend the money without being bold was tasteful and effective. It drew the eye sure that I could generate sales without them. to the key points that would help visitors decide I started making sales as soon as I turned on to buy. Except that when I tested it, I was dead the AdWords traffic, but I was losing money wrong. Bold text suppressed sales. So despite with an initial conversion rate of 1.2 percent. my love of a well-placed bit o’ bold text, I use Because the site was making some sales, I almost no bold on the site today. decided to take the plunge and test credibility My sales conversion rate from AdWords traf- seals on the site, and my sales conversion rate fic is 5.5 percent. Actually, it’s higher than improved to 1.5 percent almost immediately. that, because I’m not yet tracking phone sales. That may not sound like much, but it made the But the big lesson is that it’s rare to be profit- difference from losing money to making money. able with AdWords right from the start. If the It was the first time I tasted blood and saw that only modifications you make are to ads and testing could actually work. keywords, you’re missing the lion’s share of My graphic artist misunderstood some instruc- profits. tions of mine and created a header that wasn’tPart V: Testing Your Strategies and Tracking Your Results 300 Think back to when you launched your Web site In the frenzy of getting a Web site built, there are always compromises. No one has the time to create the perfect Web site on the first try. And even if you had all the time and money in the world, how could you possibly know for sure what “perfect” would look like? Take a deep breath, follow my pocket watch swinging back and forth, and feel your mind free-floating back, back, back to when you were creating your site and thought: “OK, that’s good enough for now.” “Not exactly what I had in mind, but I guess it works.” “We don’t have time to do it over again.” Browse your site to jog your memory. Look for areas of compromise, of “Good enough for now,” or “Gee, I hope that works.” Make a list of every ele- ment that just maybe could improve. Getting inspired by your competition Next, take a good look at your competitors’ sites. Compare their landing pages to yours. Ask yourself: ✓ What are they doing better than I am? Do they have better copy, head- lines, pictures, design, or something else? Can I use their site as a tem- plate on which to improve? ✓ What’s on their site, but missing from mine? Can I test adding that? ✓ What’s on my site, but missing from theirs? Can I test removing that? ✓ Is their site more focused on the customer, the customer problem, or the solution? ✓ Does their site seem more credible or professional than mine? ✓ Do they have more elements or different elements above the fold (visible without scrolling) than my site? Could I test these changes? ✓ Does their site deliver on the promise made in their AdWords ad? ✓ Is their checkout or sign-up process more intuitive and easier to follow than mine? Your competition can give you lots of ideas on what to improve, what to add, and what to remove from your Web site. Sometimes you can find good ideas by studying and modeling your competitors. But don’t assume something is good just because someone else is doing it — most online markets consist of the clueless copying the clueless. Until you test, be skeptically curious. Chapter 14: Making More Sales with Website Optimizer 301 For example, Richard noticed that his competitors showed only the manufac- turer’s stock photos. In many cases, the pictures weren’t even sized correctly, so they appeared stretched and grainy in visitors’ Web browsers. Richard replaced his stock photos with high-quality photos that he took, using an inexpensive light box and a digital camera. Even though the water filters he sells are in many cases identical to those sold by competitors, he achieves a competitive advantage by making it easier for his customers to visualize and evaluate the products. When Richard surveys his customers, they consistently tell him that good, clear pictures are an important factor in their decision to buy from him. Asking your visitors and customers Your visitors and customers can provide a wealth of information on how to improve your Web site. Here are a few things you should put in place to gather information from your visitors: ✓ Google Analytics to track visitors and sales, compare bounce rates on various pages, learn how users progress through your sales process, and more. See Chapter 13 for more information about Google Analytics. ✓ A high profile, easy way for visitors to ask questions (see Figure 14-1). This lets you know the questions you need to answer with your Web site. This also provides you with words that customers use when talking about your product or service. ✓ Surveys to determine visitor needs or satisfaction with your Web site. Surveys can give you an overview of what people are trying to accom- plish on your Web site and how satisfied they are with your site. Figure 14-1: MrWater encourages visitors to ask questions by soliciting them at the top left of every page. Part V: Testing Your Strategies and Tracking Your Results 302 Visit 4Q at for a free survey tool that’s easy to install and configure. You’ll be amazed at the valuable information you can get from a simple four-question survey. Your visitors will tell you — if you let them — what they like and dislike about your site and how you can change it to make them happier. With the information from the above tools, you can start to ask interesting questions, create some theories, and then test those theories. For example: ✓ If you find a high percentage of survey respondents are attempting one task and they have a lower satisfaction than visitors attempting other tasks, you should focus some testing in that area. ✓ If you find that visitors consistently ask a particular set of questions, you might test answering those questions on the product page or in a Frequently Asked Questions section (hence the name) or article with prominent links to your product page. ✓ If you find that there are fewer sales from one of your high traffic pages than from other lower traffic pages, you should test some changes to the high traffic page. Analytics allow you to focus your effort where the biggest payoff is possible. All this combines to create a steady stream of information from and about your visitors. Evaluate that information to design tests that look for ways to serve your visitors better. Here’s an example. By running a 4Q survey, Richard found that 47 percent of his visitors to the shower water filter page wanted to compare products, and that they were considerably less satisfied with their experience on the site than were people attempting other tasks, such as checking prices or making a purchase. The survey revealed an opportunity to improve those visitors’ ability to compare products. So Richard is now testing ways to allow compar- ison shopping for that category. He uses visitor and customer data to drive continuous improvement of his site. Prioritizing your list After you set up tools to capture information about visitors and use that information to generate ideas to test, it’s time to grade each idea so you can prioritize the tests based on expected ROI. You move the ideas with the high- est grades to the top of the list and test them first. Chapter 14: Making More Sales with Website Optimizer 303 Apply the following 11 criteria to each item you’re thinking of testing. The items that receive the most checks rise to the top of your testing list. These represent your first tests. (In case of a tie, I recommend flipping a coin — I like the 2005 Oregon quarter with a picture of Crater Lake, but that’s totally up to you.) ✓ The item is on a high traffic page. ✓ The test result, if positive, can be applied to other pages on the site. ✓ The item is above the fold (that is, most visitors can see it without having to scroll down or sideways). ✓ The item is the most prominent feature on the page. ✓ The item relates to a high leverage conversion, such as an opt-in or first sale. (High leverage items in most businesses are gateways to multiple purchases and customer loyalty.) ✓ The item addresses a frequent customer question or problem. ✓ This item affects all customer sales (as part of the checkout process, for example). ✓ The item fills a competitive gap. ✓ The item enhances the credibility of your site. ✓ The item contributes to your site’s uniqueness. ✓ The new version of the item is very different from what is shown on your site. This exercise provides you with a prioritized list of what to test on your site. When you come up with new ideas, add them to the list, but be sure to re- prioritize before you select the next item to test. Start testing (and never stop) Are you ready to pull the top idea off the list and start testing? First, you need to create the resource to test. You already have your original, that’s your A version. Now put together your B version. You might have to write some copy, create an image or video, and get some HTML code to tie it all into a working page. Upload that page to your site, and commence testing When you’re testing elements of your marketing, you’re acting like a scientist. Don’t just ask questions; ask clear questions in an elegant way so the answers mean what you think they mean. Exclude other variables so that the only reason for the difference in outcome is the different inputs you controlled.Part V: Testing Your Strategies and Tracking Your Results 304 Also, run the test enough times to reduce the effects of random chance. Finally, don’t stick the end of a Bunsen burner hose onto a sink faucet instead of the methane gas supply. Although hilarious, it also got me suspended. Here are four rules to help you achieve useful and valid test results: 1. Test one thing at a time. If you test the headline font and header graphic simultaneously in an A/B split test, you won’t know which element mattered and which element didn’t. (You can test multiple synergistic effects using the more complicated (and therefore left out of this For Dummies book to keep your blood pressure down) Multivariate Testing. Visit multivariate when you’re ready for some advanced testing methods. 2. Repeat each test to be sure. If you have enough traffic, engage in the luxurious practice of validating your tests by running them a second time. Website Optimizer makes this one-click simple (keep reading this chapter to find out how). 3. Keep traffic streams separate. Each ad group should send visitors to a unique landing page, so when you’re testing elements on that page, you’re looking at how a specific market segment interacts with it. If you send all your traffic to a single page, you will almost certainly miss chances to optimize the page for different groups of people. If all of Richard’s AdWords traffic went to his home page, for example, he would not be able to optimize the page for people searching for shower water filters in order to compare products. 4. Identify a measurable outcome. Sure, you want your Web site to be better, but what exactly does better mean? To conduct tests, you have to operationalize better into some visitor behavior that you can see and count. If you collect leads on your site, one measure of a landing page is how well it turns visitors into leads. Another measure is sales. You might also count downloads of soft- ware or marketing documents, or views of a key page. To use Website Optimizer, you must identify a page on your site that visitors arrive at after they convert — that is, after they take the action you want them to take. Just to make things really clear, I’ll call this the thank-you page. Two things must be true for the test to be valid: ✓ Every visitor who converts must make it to your thank-you page. ✓ No visitor who doesn’t convert can make it to your thank-you page. Chapter 14: Making More Sales with Website Optimizer 305 Testing with Google Website Optimizer Here’s the step-by-step method for setting up your first experiment in Google Website Optimizer (GWO, which sounds like Elmer Fudd encouraging his plants). 1. Start GWO. You don’t need AdWords to use GWO, but Google provides a handy link right in the AdWords menu bar. (See Figure 14-2.) Figure 14-2: Get to GWO from within your AdWords account. If you haven’t set up your AdWords account but you want to begin test- ing anyway, and log in with your Google account. After you log in, Google requires that you specify your location and time zone and then agree to GWO’s terms of service. The first GWO page you see is the Experiment List page. Naturally, no experiments are listed, so let’s fix that right away. 2. Click the Create Experiment link at the bottom of the Experiment List box to get started. (See Figure 14-3.) 3. Choose the type of experiment you’ll be running. The choices are an A/B Experiment (a simple two-option split test) or a Multivariate Experiment (where multiple elements of the page are tested at one time). Select an A/B Experiment, in which you create one or more variations of your page and compare results of each to the original. (See Figure 14-4.) This is the simplest experiment to create and understand, so it’s a good place to start. Even when you become a testing ninja, you’ll want to start improving a page with A/B tests because they give you big answers quickly. After you feel comfortable with A/B experiments, you can visit to see how to create a multi- variate experiment.Part V: Testing Your Strategies and Tracking Your Results 306 Figure 14-3: The Create Experiment link gets you started. The Create Experiment link Figure 14-4: Choose the simple A/B Experiment for your first experi- ments. Chapter 14: Making More Sales with Website Optimizer 307 The next page is a checklist of what you need to complete before the test can go live. (See Figure 14-5.) Identify the original page and at least one variation you’ve created. At first, stick to one variation — you’ll get results much quicker and they’ll be much clearer than if you compare several pages. Figure 14-5: Google reminds you of the three elements necessary to begin a test. 4. When you have this information readily available, check the I’ve Completed the Steps Above box and click the Create button to continue. At the beginning of the chapter, I give you a process to help you decide what to test. Now you tell Google the URL of the B version of your origi- nal page, as well as the URL of your thank-you page. In Figure 14-6, you see how Richard completed the setup form for an experiment that he was just beginning when we started working on this chapter. If you’re curious, visit to find out the winner and get some commentary from Richard and me on what it means for his site.Part V: Testing Your Strategies and Tracking Your Results 308 Figure 14-6: Setting up a test of two new pages against the original. The Add Another Page Variation link Specifying Experiment Details The following list helps you get the details right. Use it to specify the details on your first experiment in Google Website Optimizer: 1. Name the experiment. Enter a meaningful name for your experiment; this is how you will find it in the experiment list. (After you start testing and improving your profit- ability, you will create many more tests.) 2. Enter the URL for the original page. Chapter 14: Making More Sales with Website Optimizer 309 3. Enter a name and URL for the variation page you are testing. 4. Click the Add Another Page Variation link (refer to Figure 14-6) and repeat Step 3 if you want to add more variations. 5. Enter the URL of the conversion page (your thank-you page). The conversion page is the page that visitors land on after they com- plete the action you wish to optimize. For example, after a sale, visitors generally land on a thank-you page, which is an ideal conversion page for GWO testing. Google checks the URLs as soon as you type them, to make sure the pages exist. If every page receives a happy green check mark, you’re ready to gener- ate the tracking script that’s installed on each page involved in the test. Website Optimizer uses JavaScript tags on the pages to track visits and show different versions of the page to different visitors. Your job is to copy and paste these JavaScript tags into your pages. If you aren’t terrified of HTML, you can easily do it yourself, in which case you choose the option, You Will Install and Validate the JavaScript Tags. (See Figure 14-7.) If you don’t want to do this yourself, you can have your Webmaster do it for you by choosing the option, Your Web Team Will Install and Validate JavaScript Tags. Just remember that the whole process should take about five minutes, so make sure you’re billed accordingly. Figure 14-7: You can install the tags your- self or have the code and instruc- tions sent to your favorite Webmaster. Part V: Testing Your Strategies and Tracking Your Results 310 For the sake of your comprehensive AdWords education, I’m going to assume you selected the Do It Yourself option. The next page gives you the JavaScript tags and allows you to validate proper installation on your pages. (See Figure 14-8.) Figure 14-8: Stick this code onto the pages involved in the test and let the fun begin Chapter 14: Making More Sales with Website Optimizer 311 This feature of GWO rocks big time (that’s online marketing slang for “is a very useful feature”). It ensures that your tests actually provide you with valid results or the tests don’t even start. Take it from someone who once sent 250 of AdWords traffic to the wrong page, and almost stopped selling the product because he thought it wasn’t selling — you can’t be too careful here. Adding and validating your own JavaScript tags GWO uses an elegant configuration of JavaScript tags to send traffic to multiple pages without slowing down your Web pages or relying on external servers. Follow these steps to place the JavaScript tags: 1. Original page: Add the control script and tracking script. On the original page, you will install two JavaScript tags, the control script and the tracking script. The control script goes at the top of the page, just after thehead tag. (See Figure 14-9.) The tracking script goes at the bottom of the page, just before the /body tag. (See Figure 14-10.) To copy the JavaScript tag, click anywhere in the shaded blue area to select the entire code snippet. Then copy it as you normally would. Open your original page in your HTML editor, select the option to view the HTML source, and then paste the JavaScript tag at the top of the page. Figure 14-9: The control script goes near the top of the page, just below the head tag. Part V: Testing Your Strategies and Tracking Your Results 312 Figure 14-10: The tracking script for the origi- nal page belongs just above the /body tag on that page. 2. Variation page: Add the tracking script. Place the variation tracking script at the bottom of each variation page. Figure 14-11 shows the variation tracking script in its proper location, just above the/body tag. Figure 14-11: The tracking script for the variation page goes just above the /body tag. 3. Conversion page: Add the conversion script. Add the conversion script to the bottom of the conversion page, just above the/body tag. (See Figure 14-12.) Figure 14-12: The conver- sion script goes just above the /body tag on the conversion page. Chapter 14: Making More Sales with Website Optimizer 313 At the top of Figure 14-12, thewebsite optimizer code for wide landing page test comment helps you identify which test is con- nected to the script, so you can remove it after the test is over. Placing / before the comment and/ after the comment lets the browser know to ignore it, so it doesn’t affect anything on the visible page. After you’re hooked on Website Optimizer, you can end up with quite a few of these conversion scripts on your conversion page. Manually adding a comment helps you figure out which ones can be deleted because you’ll recognize the completed experiments. 4. Upload the changed files to your Web server. When all the JavaScript tags are in place, upload the files to your Web server using an FTP program. When the updated pages are in place, you’re ready to validate the JavaScript tags. Click the Validate Pages button at the bottom of the page (refer to Figure 14-8) to make sure everything is in the right place. Website Optimizer checks the pages and the JavaScript tags to be sure your experiment is ready to run. When validation completes, you see the Congratulations message, as shown in Figure 14-13. Click the OK button, and then click Continue to go the experiment console. 5. Preview the experiment pages. Before starting your experiment, take a minute to preview your pages. In the experiment console, click the Preview link to verify your pages are showing what you want to show. (See Figure 14-14.) Figure 14-13: Google lets you know that your tags are properly placed on the right pages. Part V: Testing Your Strategies and Tracking Your Results 314 Figure 14-14: Click the Preview link before starting the experiment to view the control and test pages. The experiment shown here includes the original and two variations. Figure 14-15 shows the original page within the GWO frame at the top. Choose Variation 1 from the drop-down menu at the top left to view the first variation page. (See Figure 14-16.) Figure 14-15: GWO allows you to preview the original and variation pages easily within a single browser window.