S: CRO–Conversion Rate Optimization, this is the science of getting more of your website visitor to buy your stuff or at least to do the things that you want them to do on your website such as download a white paper, or fill out an inquiry form, or sign up for your email newsletter. CRO is more definitely a science. It’s based on facts, on data—hard data, experiments are reproducible. In today’s episode number 140, our guest is Brian Massey. He’s a conversion scientist, literally. He founded Conversion Sciences in 2007. Brian is also author of Amazon bestseller, Your Customer Creation Equation. His rare combination of interest, experience, and neurosis were developed over almost 20 years as a computer programmer, entrepreneur, corporate marketer, international speaker, and writer. Brian, it’s great to have you on the show.
B: I’m so glad to be here. Thanks for inviting me.
S: Yeah, you bet. Well, let’s talk about conversion since that’s near and dear to your heart.
B: I do love the conversion.
S: How the heck did you get started in this conversion world? Did you start specializing in that in the beginning or did you start in more general web marketing space?
B: I almost had no choice. I’m a computer programmer by training. I’ve had Bachelor of Science in Computer Science that’s why I can call myself a scientist. But when I was co-oping in college, I saw my first IBM sales person with the red power tie, the blue coat, and the tan. I was like, “Maybe that’s what I should do with my technology degree.” After school, I got into sales—not IBM, unfortunately but semiconductor sales. I guess I rose to the highest level of incompetence in that role. It wasn’t a good fit after all. I came back and got back into coding, worked for a small company, started my own business in the 90s with an engineer there. We did internet graphics. We had Google maps working in the late 90s with satellite photos and everything across dial-up. We were still on dial-up in the late 90s.
S: I remember those days.
B: Well, dotcom crashed, took out a couple of our customers which took us out. I spent the beginning of the Ots, wrote my own analytics package because I was helping technology companies, mostly build out their web lead generation platforms—their web marketing platforms. It didn’t have Google Analytics until 2005. I was working on my own. In 2006, Bryan and Jeffrey Eisenberg did a presentation at this amazing business school here called the Wizard Academy. They started to put language to this thing called conversion optimization. No one puts up a website to get traffic, they put it up to get business, and that requires conversion. But there was no vocabulary, no discipline. I suddenly like that, “These are my people.” There’s really nothing else you can do with a computer programmer, corporate sales, corporate marketing, entrepreneur, than be a conversion scientist. It’s the only thing. I put on a lab coat and hang a shingle in 2007 and started mostly teaching and building the market. Because like I said, we didn’t have this vocabulary, we didn’t have these disciplines, and this process in place for optimizing websites. I’ve been doing it ever since.
S: Why the lab coat? I get that this is kind of a distinguishing branding feature for you. No one else does this, no other conversion expert wears a lab coat while presenting or anything.
B: No. In fact, it isn’t particularly original. I had enjoyed science as a kid. I was influenced by Bill Nye the Science Guy. I was honestly looking for a kitschy approach. If I came out all data science without a little bit of humor, it would terrify everyone. I was trying to introduce a bit of camp into it honestly. I was always afraid. Speaking and writing was how I went to market. Every time I would go out on stage in a lab coat, I always had this not in my gut that everybody’s looking at me going, “Oh god, he’s wearing a lab coat.” But I’ll tell you what, no one will let me go on stage without it now. I think only once had I presented without it and they very quickly apologized and said, “Next time please, wear the lab coat,” because everybody was like, “Where’s the coat?” I think it was part of that—the logos I chose were kind of campy science stuff and I work sciency kind of kitschy science stuff into my blog post and stuff. I made it fun. It was really kind of fun.
S: It makes you memorable.
B: It does. Everybody forgets what I say but they’ll remember the lab coat.
S: Yeah, that’s funny. I’m sure they remember what you say but they remember the lab coat even more because it really does make you stand out. It makes you remarkable. That’s what you need to do in the real world and online is stand out from all the noise out there. All the other folks who are doing the same sort of stuff as you and maybe doing it as well or better. What matters is people need to remember us. What is the worst mistake that you see happening these days with conversion optimization? I know two years ago or so, the slider carousel thing was being done to death. Finally, it’s been kind of slowing down, we haven’t seen as many of those sliders these days. I think people are starting to catch the drift that it kills conversion. What is it in 2018 that is the big stupid thing that people are doing to destroy their conversion?
B: There is a fundamental shift that’s happening. Gordon Moore started this, it’s his fault. Moore’s law, 1965, he says, “Every year we’re going to be able to double the number of transistors that we fit on a microchip.” What that means is that every year the microchip is going to get twice as powerful, twice as many features, and it’s going to cost half as much. For marketers, what that means today is that we have amazing tools from the cloud where we can requisition a whole rack of servers in the cloud for pennies. We don’t have to build out our data centers. High-speed internet all over the place, everybody’s on it. Mobile devices that allow us to query people in real-time, high definition webcams in our stock laptops. They are very high definition so much that we can tell where people are looking on the screen and entrepreneurs are taking advantage of this. It’s really easy for me to take some creative or an idea, put it in front of a panel of people, and see how easy they interact with it or how effective they can be with that particular thing. Just specific ideas, as I said, we can test layouts through these eye tracking studies, and it’s all done right from our laptops without leaving our desk for hundred of dollars instead of tens of thousands of dollars. We’re still kind of designing like madmen. When data was really expensive, it was hard to collect. When you have to have people with clipboard standing in front of the store, surveying customers. You really could only do this like one big research project and then your design team—your development team, your creative team—had to make assumptions from that research but we don’t have to do that anymore. Every step along the way of the design process or if you’re revising for every assumption that you make about how you can improve a landing page, a website, or a product page, you can put it in front of some people, click some data on it, and find out if in fact the change is going to be worth it. It’s everything from simple usability tests all the way to inexpensive A/B testing. Google Analytics has given me a way a competent A/B testing tool in Google Analytics. Now, A/B testing is easy to get wrong. I think anyone who’s still using these old design process—these madmen design process—is in for trouble. We just had an agency come to us and they said, “Hey, this major client of ours, we’ve been working with them for two years, wanted us to redesign the landing pages.” We came to them with wireframes mockups and they said, “Wait a minute, we need some data. You can’t just give us designs. I think the days a design firms giving a business a choice of three design mockups and asking them to pick are gone. Any manager, or anyone really, working with an agency is going to say, “I’m not a designer, go click some data on these and tell me which one of these is going to be the most effective with my audience.” The agency can go off and do that in practically an afternoon. These tools are so good. That’s what I’m most excited about. There are a lot of advantages that marketers are going to be able to capitalize and should be already.
S: Yeah, that’s pretty cool. If somebody is making this mistake of getting a design without him testing it first, without coming up with some sort of data to backup the fact that they’re going to put a hero image over here, the call to action button over there, and three benefit statements, or two, whatever. What does that look like in terms of a well-executed strategy? If somebody uses the right tools and collects the right data, what is that look like?
B: That’s a good question. It looks like this, it looks very much at the very beginning like there’s regular madmen style scarce data model. You go and you do this research. This might be easiest if we take an example. Let’s take an example, we’re developing or optimizing a landing page. A landing page has just two jobs: number one is to keep the promise it has made in an ad, in a link, in a social media post, or a link on your website. Number two, it needs to bring the visitor to choice. It needs to be very obvious that you’re asking the visitor to do something and it needs to have all the information on it that makes them feel comfortable and confident in taking action. The beginning, your question is, “I got to keep the promise so I know the offer is but what’s the best way to word that offer? Maybe we should use a different kind of offer. Should we be a $100 off, 20% off, or should we leave with the highest value, the best in industry? However you’re going to do that, “This piece of content is free in exchange for your contact information.” We generally take our first guess. We’re going to spend a lot of money, we just have to guess. We might show it to ourselves, our boss, our team, maybe there’s an agency involved. But this is a sample size of maybe 10 people at the most. That’s a very small sample size for getting something right. How do we get more people involved? Well, go look at your email service provider and see what subject lines are most popular with readers who’s got a blog. That will give you not only an idea of what they’re interested in but the words that’s attracting them. Have your PPC folks down in the basement behind that big oak door, running a spreadsheet for you, showing you what ads they’re running. Look at the ones with high impressions, high sample sizes, the ones that are converting the faster, giving the most clicks, what are the offers and the words being used there? That data’s just sitting around but it’s high sample size, it’s over a long period of time, it’s your customers and prospects that are clicking through–this is really good data that’s being ignored. You start there. You’ve got your offer and you got some choice, either they’re going to fill out a credit card or they’re going to fill out a lead form, or they’re going to pick up the phone and call. By the way, that’s the completely underused. If you’ve got any mobile traffic at all, go talk and go see about having mobile visitors, click the call. It’s underused. We’ve got to get some copy built out here. It’s the copywriter that is got to build a value proposition and handle a few objections that really needs all these madmen style, they really need to understand the visitors. They need to read your personas, they need to read your marketing studies, they need to do some surveys if you don’t have that stuff, and a good copywriter will do that. Once you got that, you want to pick an image and you’re going to have more than one idea for what the headline and an image should be. Run over to UsabilityHub and the images and headlines that you like best, put them in front of a small panel and find out, ask questions like, “What does this company do? What is the offer on this page? What would you do if you wanted to take advantage of this offer? Does this company seem credible?” Ask questions like that and find out what combination of those things quickly communicates to the visitor what the page is about because that’s so important. Then you hand it off to the designer now. He does the layout and he’ll come back and add a carousel, or he’ll add a fuzzy movie background with a hero, or some stylist take images—and on the landing page, that stuff usually hurts worse than it helps. We need to put an asterisk on that because I have a story to tell you about the rotating carousel. How often do you get something from a creative agency and you’re just like, “I don’t know what’s wrong with it but it doesn’t seem right.” “Well, put it up on sticky.ai. We’ll bring a panel to look at your page and you can see how their eyes track. Does it track naturally from the promise kept in the headline to the very clear understanding that we’re asking them to do something, then all of the reasons they should do it, and a clear call to action that they can execute on.” You’ll see that in the eye tracking studies. Again, we’re talking hundreds of dollars, couple of days to collect them, get the data, and the reports are amazing. These are just a few of the studies that you can do along the way of designing out a landing page. If you already got a landing page, you have ideas like, “We could change the image. We could change the headline. We need more copy. We need less copy.” All of those things can be put in front of folks. You can ask specific questions about how easy it is for them to complete tasks and find things on the page. That’s really what we’re trying to do. This is just exciting to me. I hope that kind of illustrates few of the things you would do along the way. Ultimately, you will launch and that’s when you get the most delicious data with heatmap reports. You’re watching session recordings which is so boring. You would see things in session recordings, you wouldn’t see anywhere else. Of course, you’ve got analytics running underneath it. You can also be doing A/B testing if you got enough traffic. The thing is you won’t be A/B testing crap. The only reason A/B testing programs fail is because you test the wrong things like button colors. Start with some ideas like, “We think we can improve this by x.” Those are the things that you filter through with the interim testing. You’re left with the things that are like, “Well, we don’t know. We can’t nail this down. AB test is a smart idea here.” That was kind of long for me to answer your question but did that help?
S: It did. We’re going to break this down even more. First off, you’ve mentioned some different tools like, UsabilityHub, sticky.ai. Let’s further explain what these tools do so that our listeners can follow along. UsabilityHub, is that kind of like usertesting.com? What does that do?
B: Yeah, it is. There is a spectrum. I’m happy to give you a graphic. I have a graphic of kind of where we use tools in different parts of the design process. The different kinds of questions that we have, hypotheses that we’re trying to answer about a page. We use these interim tools because we got to deliver increased revenue and we’ve got six months to do that. That’s our entry level package. We do free user testing—unlimited user testing—because it makes us better at picking what to AB test which is the Supreme Court. At the beginning, as I’ve said, you’re looking at your AdWords, your email service provider data, other sources of data that aren’t directly applicable to the landing page. Any existing analytics data that you have on your existing page or existing site can really help you answer those questions. There are tools that are really great for mockups. You don’t even have to have a live page. UsabilityHub, Helio, those are two tools that provide a suite of ways of testing. You can do a 5-second test which is great for that headline image question. You can do a click test which measures how long it takes someone to find out where they’re supposed to click to solve a problem. Or just a question test where they get to spend as much time with your page, your mockup, and you ask them questions about what they would do? What they think the price is? What do you think the difference in levels are, if it’s a pricing page. It’s very easy. These services provide the panels. It’s people coming, pretending to be your customers and your prospects. This is the rub, this is not as good a data as A/B testing, but it will easily help you narrow down the potential choices. Remember that day where on a Friday, you have this amazing idea for how you could position the product, completely change the way people see the business, and really do an end run on your competitors. You’re so excited and you spend all weekend imagining the landing pages, the ads, the messages, and the supporting blog posts. Then on Monday, you come in and you’re like, “You know what, there’s no way I can sell this idea internally.” You park and then it dies. But with data like this, you can try some crazy stuff. You can put that crazy stuff back in there. These are the reasons that we got into marketing, we wanted to motivate, influence, and persuade. Unfortunately, because we’re spending a lot of money on our campaigns, we got to be safe. Well, with data like this, you can try some of those crazy things. If you’ve got a prototype or a working page, then usertesting.com is an amazing tool for sizing up how well that page is working and where the problems are. For UsabilityHub or Helio, it’s a couple dollars per person to come look at your static images and your crazy ideas. I hope you guys will be trying more crazy ideas. It’s $100 a person with usertesting.com but the data is so rich that it’s really worth doing a test like that. Once you’ve got your best idea, we’ve mentioned sticky.ai, if you’ve got layout, great for understanding where buttons should be posted, if you’re button sticks out enough, or your form sticks out enough, or if there’s images on the page that are causing people not to get below a certain level on a page. These tools can be very easily used to develop a good first start so you don’t launch duds anymore. When you pass disparity of launch, we like Crazy Egg for heatmap reports. We like Inspectlet for session recordings. Hotjar, right now, gives you session recordings, form abandonment, heatmap reports, scroll reports, for very little price. That would be if you’re just starting out. That might be a great tool for you. Then you got Adobe Analytics or Google Analytics running underneath the site. Then you can pick up Visual Website Optimizer, Optimizely, Omniconvert, convert.com, or you just Google Optimize which is built right into Google Analytics for A/B testing. I guess a graphic would be helpful. I’d be happy to share that with the audience.
S: That would be fantastic. We will include a link to that graphic in the show notes for this episode. That will be at marketingspeak.com. Perfect. So, sticky.ai, that is an artificial intelligence-based algorithm?
B: No, it is not. Let me clarify that. If you want to find out how people are consuming your webpages, sticky.ai just uses the camera to figure out where people are looking on the screen as they’re interacting with your webpages. It’s that simple. They do the same thing for video. This is very interesting because video will either really help you or it will really hurt you. It’s such a powerful medium. The cameras are such high resolution, they can see the micro movements in the face and tell whether someone is enjoying or not enjoying, whether they’re reacting positively or negatively to your video as they watch it. They bring up 50 or 100 people to watch it. You get a pretty good idea of how that video’s doing and where in the video people are dropping off and go into the edit bay and fix those problems. This is stuff that used to cost hundreds of thousands of dollars and only Super Bowl commercial quality content was going to be put through this sort of a test. Now, for a few hundred dollars, I can put anything important in front of these audiences and see how my videos are doing or see how my layouts are doing. It’s very exciting.
S: Wow, that is so cool. Let’s say that you have a video for your homepage, some sort of an explainer video. You don’t know if it’s really working that well or not. You can run it through sticky.ai and these panel of people will get tested for the eye track and everything. Then, their reactions to your video will come back out through the output somehow?
B: Yeah. We’ve had good luck with explainer videos and if you watch them you’ll explain your product but then there’ll be this spike when you mention something like toll-free minutes for voice-over IP company. You mentioned it as kind of a side note towards the end. Suddenly, there’s spike like, “Wait a minute. That’s something important. Should we move that to the top? Should we lead with that? Should we emphasize that more?” It’s very, very helpful.
S: That’s cool. I mentioned that you could use these sorts of testing regimens to test book covers and things before you actually publish the book. I’ve heard, for example, of authors testing book covers by running Facebook ads and putting competing covers out there and seeing which one got the most engagement on Facebook ads as if the book was already out but it wasn’t, this was just a test. Is that something you would recommend? Or would you take it to the next level and run it through UsabilityHub or something like that?
B: Facebook is a great source. Facebook is really helpful because you have amazing control over targeting. You can take for instance, some names from your mailing list, people who are interested in your product. Ask Facebook to build a look-alike and go and look at it. It can be hit or miss but if you get it, Facebook will bring people to an ad that are very much like your prospects and customers. This is one of the things we want with the best data. Yes, you can put images in front of them and see which of those images and words get clicked through the most through an ad. Great way to test video. It is a little bit different so that it’s an interrupt-driven response but this data is to guide you to the things you want to A/B test. You’re not making absolute decisions with this sort of data. You’re really using it to eliminate options that perform poorly or buy some insurance for yourself if you’ve got it changed and you’re not sure what the reactions are going to be. You put it in front of a panel and it performs fine, it performs equal to the control, at least you know you didn’t shoot yourself in the foot with it. Facebook is a great option for that, absolutely.
S: Yeah, cool. You mentioned some tools like Crazy Egg, Inspectlet, and Hotjar for testing user engagement, things like heatmaps and scroll, activity and that sort of thing, even watching sessions as they are getting created. Somebody is interacting, filling out a form, scrolling, clicking—all that we can follow on a per user basis. Let’s kind of break that down a little bit further. Why would you use Crazy Egg over Hotjar or vice versa? Heat mapping, can’t you also accomplish a lot of that within Google Analytics? Isn’t there some heat mapping capability within that? Give us more details on this.
B: Alright, great. Let’s assume that we have a website that we’re optimizing. We typically go into Google Analytics and we’ll begin to understand what the primary segments are. Mobiles are always different from desktop. You always optimize mobile as if it was a different company, quite frankly, because the context is just so different. You might look at new and return visitors. You might look at different traffic sources, look at the landing pages, maybe you find a landing page that has a high bounce rate. That means one of two things, either you’re getting crappy traffic, some bad ad or there’s something on that page that isn’t communicating the value properly to qualify the traffic well. One of the first things you could ask is, “This is a pretty long page, are people seeing the call to action which is at the bottom of the page?” You bop over to Crazy Egg or Hotjar and you run a scroll map on that. It will tell you where most of the people are dropping off from a scrolling standpoint. Usually, it’s very blue when there’s only 10% or 5% of people who are seeing the page. We call it being underwater. You can very easily go see if there is a scroll problem on that page. Your pricing page, let’s say you’re an online service, and you have multiple levels of product—you have the free, the pro, and the enterprise—and people aren’t buying. “This is the money page. This has to work.” You can actually go into Crazy Egg and you can look at the heatmaps. You can see not only where they’re clicking—which is what Google will tell you—but where their mouse is moving. You can begin to see which features they’re most interested in. “Do you have the most interesting feature pressed down further on the page? Are you presenting too many features that are available on all the packages?” This page’s job is to help people choose. Heat maps can show you things where people are clicking on things that are clickable, whether they’re not clicking on things that are clickable, and let you make adjustments to the layout like that. You come up with some ideas for how to improve this page but you’re going to get with a lot of ideas. On our website, we will have easily 100-150 ideas—good ideas for improving the conversion rate. Because we’re experts at this, we’ll only be wrong on half of them. We’ve got to figure out a way to narrow down those ideas that are most likely to move the needle and those are the ones we want to subject to the A/B test which is a precious test that uses real traffic on the site. We’ll take you to these ideas, change the headline, reverse the order on our pricing page. That’s one of our go-to playbooks. The most expensive thing is on the left. Then we can put this in front of some folks at UsabilityHub–just a mockup, we don’t have to change the page display. Go to UsabilityHub and ask them questions about. “If your business is this size and this is your role, which of these would you pick? What would keep you from selecting one of these? What questions aren’t answered?” You begin to understand which of your ideas is most likely to move the needle. You work that in. We actually use a ranking algorithm that takes the amount of proof that we see in these tools into account, to highlight which hypotheses are most likely to deliver winners. Then you say, “Okay, this is the format that works best. Mr. Designer, will you please help us with this?” As I said, the designer will come back with something. You’ll like it, you won’t like it, it doesn’t really matter what you think. Let’s put it in front of some people and see if they’re getting to the parts of the page we want them to see. Are they scrolling? Are they seeing the important features? Are they confused? Go to the existing page and watch some session recordings and you’ll come up with another 10 ideas for what could be wrong with that page. Then you say, “Okay, this either does or doesn’t work.” You iterate on that–this is an iterative process. You’re expanding your development cycle or you’re testing cycle but for a very good reason because you’re launching stuff that’s going to work. You’re saving on the cost of bringing traffic to a crappy page. That’s what you’re saving on. Does that illustrate it? I feel a little bit I might be repeating myself but did that help?
S: No, this is great. Just to clarify when you said this concept of being underwater, that’s when the scroll map where it gets into certain colors scheme where people aren’t scrolling anymore?
B: Yes. That’s blue or dark blue. It starts off with a lot of people seeing the top of the page because that’s what open. 100% of people see the top of the page. Then you see the scrolling on these longer pages and then there’s a point where it just goes from green to blue. Then deeper and deeper blue. No one is seeing the bottom of your page. If that’s where, for instance, you talk about free shipping, if that’s for instance, where you talk about your return policy, or if that’s where the form is, 5% or 10% of the people are realizing that they’re being asked to do something. Some of these stuff could be brought up or you can put cues on the page that get people to scroll. I know you’ve seen on pages where there’s like a pointy arrow every few screens just to remind you to keep scrolling it’s because someone went in and said, “Oops! There’s a scroll problem here.”
S: Yup, got it. You mentioned Adobe Analytics. Most of our listeners, I’m guessing, are using Google Analytics. What would be the benefits of Adobe Analytics over Google Analytic or some other tool?
B: I don’t think there’s a benefit of one over the other. I don’t want to get in trouble here but traditionally, Adobe Analytics has been a little more fragile. It allows you to dial in everything on every page but developers release pages and they forget to put the right code on there or it changes. Tag managers, I think, are helping that. Google Analytics is great up until a certain point. The free version is a fantastic tool. It’s a little more stable because it’s really collecting URLs. We’ve just found the data to be more reliable over time but if you get over a few million visits a month, then you really got to bump it up to the paid version. At that point, the two tools are equally as expensive. Adobe Analytics is a lot more expensive than free Google Analytics. I’ll tell you honestly, we can’t remember the last time we were involved with a client that wasn’t using one of those two. Those are two I know about.
S: Got it. You mentioned as far as A/B testing or even multivariate testing. You could use Visual Website Optimizer, you could use Optimizely, you could use Google Optimize—there’s a bunch of different options. What’s the go-to tool that you use in that regard?
B: Visual Website Optimizer and Optimizely are the most common in our book of business. We’ve been bullish on a multivariate testing tool that uses artificial or machine learning to reduce the amount of traffic it takes to do a multivariate test. I would take a look at Sentient Ascend. There’s some caveats if you skip that set up right it can be really an amazing tool. There’s some that I know we can get less expensively. Omniconvert is a competent one we’ve worked with. Convert.com, I’ve known Dennis for a long time. I know that we’re all kind of looking at GDPR. I know that convert.com, Optimizely, Visual Website Optimizer are compliant. That’s kind of the spectrum of tools that we’re used to. We’re just released playing with Google Optimize right now. It doesn’t have the targeting capabilities that we like to have in some of these other tools but you can pretty much tease out what you need to know in post test analysis by going and looking at the analytic since it’s integrated with Google Analytics.
S: Yeah. Are you normally doing A/B testing or you’re doing multivariate testing?
B: We probably do 50 A/B tests for every multivariate test we do–something like that. Multivariate test is where you change multiple things on a page. If you have a landing page, and you have three good ideas for a headline, and two good ideas for images, and then two forms of cut links of copy that you want to try on the page, a multivariate test will present each combination of all those things to enough users to get statistical significance. With 12 combinations there—it might be 13—ultimately with the control combinations just from those few changes, and to test 13 variations takes a lot of traffic. At the end of the day, you don’t know which of the things was really the important thing to the visitors. You don’t know, was it the better image that really nailed it, was it the headline, so we like the A/B test because we can ask specific questions about how the page should improve, do a test that isolates that question. When we learn that, we can improve on it, we can double down on it, and we can do things like take that same wave. If it’s a headline, why don’t we take that headline over the email and use that as an email subject line? If it worked well in this situation, it might work in other formats. Would that be a good offer in Ad Group on AdWords? We like that capability of A/B testing to help us know what and to ask the question why.
S: Makes sense. Let’s say that you’re going to do an A/B test on the button verbiage, the call to action. Maybe it currently says “submit,” that makes me cringe whenever I see that, so many websites are still doing submit buttons. What’s a test look like? What are you going to come up with as far as the different options as a replacement for “submit.” Is it going to be in first person? Second person? Just walk us through this.
S: That’s awesome. What is testing statistical significance mean? Is the tool going to tell you that you need two weeks? Or two days? Or do you have to know some math in order to do this successfully? How does that work?
B: That’s kind of an art of it. The tools will tell you when they believe you’re at statistical significance. They’re kind of trusting you to set the test steps so that their results tell you the truth. We can make mistakes in how we configure the test in such that we can fool the tool into thinking we reached this statistical significance. We usually use a few different analysis. What the tools tell us, sometimes it nails it. We will use a statistical significance calculator and use that to estimate how long the test needs to run, or how many conversions we need to run. I think this is important with A/B testing. It’s not about how much traffic you have, it’s about how many transactions you have–that’s what determines statistical significance. We have 100,000 visits but if you only have three transactions, it’s not enough for you to test on. At least until you increase the conversion rate, right?
B: There’s a number of tools, number of rule of thumbs that we use and it’s all predicated on how big a difference there is between a control and each of your variations. Like I said, if it’s 50%, you’re going to come to statistical significance pretty quickly because it’s such a big win. Unfortunately, not every test is that big and we’re usually jumping up and down for a 10% or 20% lift on any particular test so that answers the question.
S: Yeah, that’s great. You mentioned there is a calculator for statistical significance. Is that an online calculator?
B: Yeah. I think we tend to use the Visual Website Optimizer traffic estimator. You type in how much traffic you’re getting, for how much lift you really want to dial in, and it’ll tell you how long the test needs to run.
S: Got it.
B: As you’re going along, you can also evaluate the confidence interval using abtestguide.com. Our friends at online dialogue and then everyone’s put this awesome site together. It will overload you with statistics so you can explore the power and all the things. But they have a Bayesian calculator which is more probability and then a frequentist calculator which is a thumbs up, thumbs down, sort of thing. Frequentists is an all in or all out approach whether or not your test is of statistical significance. I would check that out for sure.
S: Okay, great. What would be some tests of your own stuff that were counterintuitive or surprising? Like, you’re testing things like your online course title and stuff like that before you’re officially launching it, that sort of stuff.
B: Yes. On landing pages, you really need to nail the top of the page. For those spontaneous, those quick decision makers, spontaneous to competitives, you really need to let them know everything—all the important small points and then expand on them below for the more deliberate visitors. You talked earlier about the last couple of years where everybody’s adding carousels to their pages especially on landing pages. When we tested carousels, they work against conversion more often than not. Having said that, we just finished a test and the client wanted the agency we were working with to put a carousel on there. We’re like, “Well, let’s test it. At least we’re testing. We’re not just slapping it up there.” It was over a 30% lift on these pages. This is why we don’t do best practices anymore because you can’t do too much A/B testing without being completely humbled. Your beautiful, expensive pages just too often don’t win against ugly. It’s what your visitors want. You’ve either got to bite the bullet and say, “Well that’s the cost of branding,” or give your visitors what they want. But I’ve got a blog post that shows a great example. Udemy went through this process, very much this process iteratively to get a landing page, had great results, and it’s a shining example of how you should nail the top of a landing page with this process.
S: Can you give me that to share on the show notes page?
S: Cool. You have an online course that you’re currently testing the title of? This online course looks like the winner is Everyday Behavioral Science for Digital Marketers.
S: I’m not actually doing this test. This is a great example—Everyday Behavioral Science. I’ve been speaking about behavioral science and really talking about all of these things because you don’t have to be a data scientist to use these tools. If you are not using these tools and you’re in digital marketing, you are not going to be in digital marketing for a very long because you just can’t keep up with people who are using these tools. I don’t care how brilliant you are. You can’t guess your way into these things. Audiences are too fickle. I’ve been speaking about Everyday Behavioral Science and have had crowds in my presentations. I’ve always been a little, “Isn’t that geeky sounding? Am I going to scare people away?” I probably could get more people if it was like, “How to make $1 million from a traffic you’re already getting.” But I think I’d be getting the wrong people. I really do want those people who are like, “Alright, I might be a Russian Literature major but I’m in digital marketing now and I’m going to figure out how to do these things. It’s going to make me better. It’s going to let me be more creative. It’s going to help me manage my helicopter executives. It’s going to help me keep my agencies honest and it’s going to let me do more crazy stuff. That’s what I’m in this for.” This was on list because we had some data from my other presentations that it was attractive to the audience we wanted. We tested conversion optimization master class. Basically hijacked the masterclass.com things where you can have Aaron Sorkin teach you screenwriting and Somerset Maugham teach you how to write a short story or something like that. Those are two very different ones. I think our next test will probably be, and this is a low traffic site right now, we’re not spending hundreds of thousands of dollars a month to drive people to it. We’re doing it serially. Our next one will probably something like become a creative marketer. Playing on that idea you can really add more creativity to your marketing using science and see if we can use the copy to get people across that bridge in a scope of a landing page. Does that help?
S: Yeah, that’s great. Are you testing also things like the number of modules to offer in the course, and the price point, and all that sort of stuff.
B: No. We are testing the headline and we are testing the trailer video. We’ve got two versions of the trailer video that we’re testing. We just don’t have enough traffic to really start testing a lot of different variations. The scroll map tells us that people are scrolling and consuming the page.
S: Alright. You want to share what the price point is for this course?
B: The price point for this course is $345. Right now, we’re offering $100 off right now, so it’s $245.
S: Nice. If this episode airs in a little while, will that offer still be valid? Is there like a special offer code that Marketing Speak listeners can use to get that?
B: You know what, I will make a special offer code for your listeners that doesn’t expire. How does that sound?
S: Very nice. Thank you. Very kind.
S: What’s that offer code going to be?
B: What would you like it to be?
S: How about Marketing Speak?
B: Let’s call it MarketingSpeak.
S: One word.
S: Perfect. Very good. What would you say would be the best takeaway that somebody would get out of this course?
B: The course is designed to teach them how to do all of these. We’ll go into more detail on setting up what we call a digital lab—which is just a process of choosing these tools, as well as free plugins and things that we use to snoop on analytics and spy on either our competitors or any site that we want to steal from. It’s kind of an eye-opener whenever we show our clients that their competitors already have A/B testing software installed or they already have Hotjar installed. A tool called Ghostery which is designed to block these things will actually report on what is on in somebody’s site, so you’re going to get an idea how sophisticated they’re being. That one’s fun. Setting up that lab with your A/B testing tool and your heat mapping tool and stuff like that. I talk about how the design process has changed in a little bit more detail. We go enough into analytics that you can answer the important questions which is really understanding what the tides are in your site. Where your seasonality is, where some things are broken so that you’re using a period of data that’s right for your business. Then we go through what we do, we use our testing process, we do an entire landing page process—best practices for landing pages, how to choose copy, headlines, and images—all the way through right into A/B testing. It’s talk full of downloads, the spreadsheets that we use, the tools that we use, when we’re doing our research leading up to A/B testing and going through and do the A/B tests.
S: Wow, that sounds really good.
B: Yeah. That’s a lot of takeaways. The one takeaway that you’re going to come away from is–you can do this, you can be a scientist, you will earn your lab coat, and it will fundamentally change your success as a digital marketer or someone who’s managing a digital team.
S: Very nice. This has been very enlightening and if our listeners are already excited and have some great tools just from this one hour of interview on this episode, just imagine what you could get by going through the entire online course. That’s just very awesome. I appreciate the $100 offer code. Again, that’s MarketingSpeak is the offer code listeners. If somebody wanted to hire your agency to do a whole overhaul of your conversions, where would you like them to go to?
B: You’ll learn more about the agency at conversionsciences.com. Our primary offering is a team of people that will just come in and light all of this stuff up and do it for you while you keep working on your business.
S: Very good. Thank you, Brian. Thank you listeners. Action-based checklist of things that you can do to implement is all at marketingspeak.com along with the transcripts, show notes, with all the links to the various resources that were discussed. We’ll catch on the next episode of Marketing Speak. This is your host, Stephan Spencer. Signing off.