Part V: CONVERSION RATE OPTIMIZATION
The aim of conversion rate optimization (CRO) is to improve the performance of a landing page or website in terms of measurable actions users take that support business goals (that is, conversions). Depending on the particular business, desired conversions can be quite different from site to site, even within niche industries. Some of the most common conversions include:
• Purchases
• Form submissions (such as newsletter sign-ups or contest entries)
• Link clicks (such as affiliate links or email signature links)
• Social shares
• Downloads
Mastering CRO is crucial to succeeding in the online marketplace. Websites with high conversion rates are better at monetizing traffic than websites with low conversion rates. Their proven ROI allows them to invest more in marketing channels such as SEO and social media. Moreover, increased competition across online channels (both paid and organic) is driving down profit margins. As the cost of attracting website visitors continues to increase, conversion rates will become even more important in the future. A business' ability to convert the traffic it already has into paying customers can and will be the difference between growth and extinction.
There is little reason not to engage in CRO. Which leads to the next question: Is CRO hard to do? Is it really that difficult to tell the difference between high and low converting pages? It's probably harder than most of us think it is. According to Avinash Kaushik, “80% of the time you/we are wrong about what a customer wants/expects from our site experience.” (See www.kaushik.net/avinash/experimentation-and-testing-a-primer.) Although an 80 percent failure rate sounds remarkably high, numbers from successful companies support this claim. Internal experiments at Microsoft show that only around 33 percent of all experiments produce an improvement, while another 33 percent have no significant change in performance, and the last 33 percent perform worse than before. Similarly, Netflix considers 90 percent of new features to be wrong. Amazon is more optimistic, assuming a success rate on new features of just under 50 percent. Even if we are better at guessing than these companies are, we will never be right often enough to make good business decisions based on gut feelings alone. This is why it is so important to conduct experiments before implementing new ideas.
Statistics from the preceding paragraph are from “Online Experimentation at Microsoft,” by Ron Kohavi, Thomas Crook, and Roger Longbotham, 2009 (www.exp-platform.com/Documents/ExP_DMCaseStudies.pdf)
The Evolution of CRO
Although CRO is a fairly new discipline, it has roots in ecommerce lead generation, which has been around since the early 2000s. In the early days of CRO, the aim was quite simple: to improve the performance of a signup form or sales page by tweaking design elements (e.g., buttons, headlines, copy, and images). The CRO process was isolated to particular landing pages and decoupled from the overall business strategy. The aim was to generate more leads and sales, not change or improve the overall business. The term “landing page optimization” describes this process well.
It quickly became clear that bigger changes have greater effects than small changes. Making big, bold changes can lead to vast improvements—and equally large failures. Some of the case studies examined in Chapter 21, “Lessons Learned from 21 Case Studies in Conversion Rate Optimization,” are quite dramatic. On some pages, everything is changed. On others, the majority of the content is simply removed in order to reduce complexity.
Most highly successful CRO efforts result from big changes. But it is hard to make big changes when you are only working on a single landing page. Sure, you can change the presentation of the product, highlight different USPs, add great testimonials, and the like. But such changes only scratch the surface. Focusing on landing page optimization is too confining—too isolated—to create vast improvements. A broader perspective on CRO is needed, one that involves changing key benefits of products, creating engaging storytelling, improving customer service, nurturing community, and making binding commitments to customers (such as 30-day money back guarantees).
Testing and CRO
CRO is more complex than it was just a few years ago. Today, most businesses need to optimize for multiple goals. Ecommerce websites not only want to optimize their sales funnels, they want to optimize their entire brand for user generated reviews, social shares, and links. This makes the tests more complicated.
As marketers, we are getting better at testing. But we still need to be smarter about what we test and how we test it. In order to figure out what to test, we need to collect qualitative and quantitative data, and then use it to formulate hypotheses about what we think will improve the conversion rate and why. We need to spend more time reflecting on why a certain change affects behavior, so that we can eventually understand (a) what makes some potential customers convert, and (b) what makes other potential customers not buy.
Equally important to carefully considering what we test is how we test it. Experimental methodology is crucial in the CRO process. In order to improve the conversion rate on a page, different variations are tested through A/B/n tests or multivariate tests (these are further examined in Chapter 22, “An Illustrated Guide to Web Experiments”). If we fail to run statistically valid experiments, we will make decisions based on highly dubious grounds. This is why it is important to understand the basic statistics behind experimentation. For example, when conducting a simple A/B test, we often notice very large differences between the two variations when data starts coming in. But over time, the differences will become smaller. Because there are only few data points early in the experiment, there is greater fluctuation in the results. This helps us understand why it is crucial not to get excited too early on, and to wait for statistically significant results before making any conclusions.
At this point, the CRO field is still influenced by persistent myths: red buttons convert better than green ones; having an arrow next to the call-to-action (CTA) improves conversion rate; short pages convert better than long ones. These claims are true in some contexts, but not all. These myths do not evolve out of thin air, after all. Some case studies (either from blog posts or conference presentations) document that red buttons outperform green buttons by a significant margin. But the lesson from such case studies is not in the substantive results (red button versus green button), but in the process that leads to the results. If you are going to follow best practices, test them in the specific context they are meant to be implemented in first.
Which leads us to the CRO blogger's dilemma: Many people want fast and easy ways to improve their conversion rates. Blog posts and presentations that focus on the long, cumbersome process that generates test results tend to be received with less enthusiasm than quick wins presentations and posts. So bloggers do their best to provide substantive knowledge, even though it is highly dubious to what degree the results are valid outside of the case study. This means that we tend to copy successful business results when we should be imitating their processes.
Two of the chapters in this Part, “The 12-Step Landing Page Rehab Program” and “An Illustrated Guide to Web Experiments,” therefore focus on the CRO process rather than substantive findings. The first post provides a framework for inspiring and conducting tests, while the latter provides methodological insights into conducting scientifically valid experiments.
The Future of CRO
We will never be done improving and optimizing our landing pages and websites. While investments in CRO may have diminishing returns over time, it is very unlikely that you'll ever get to the optimal point. The web is too dynamic to reach this point. This makes it pretty safe to assume that there is a future for CRO. But how will it look? Optimizing conversion rates on a single page is good for a business. Optimizing the conversion rate on thousands of pages is much more valuable. This is why it is important to consider the degree to which CRO scales. There are at least two good ways to make CRO scale: hypothesis formulation and automation.
As mentioned above, hypotheses ought to be used to generate alternative variations that can be tested. Hypotheses also help us identify the boundaries of our experiments, which helps us to implement winning variations for similar situations without having to conduct near-identical tests. Understanding human psychology and social psychology helps us generate great hypotheses. Therefore, it is fitting that the first chapter in this Part is “An Illustrated Guide to the Science of Influence and Persuasion.”
Many CRO activities can be automated. For example, The Huffington Post tests different headlines and images for the same story in order to optimize clicks and shares. The winning variation is automatically implemented. This makes it possible to conduct a large number of tests without going through the tedious work of reviewing each test and implementing the winning variation manually. This is especially important in industries where each page/product has a short life cycle.
HuffPo does not test random headlines, either. They use—you guessed it—hypotheses to generate alternative headlines: whether angle X or Y appeals more to the reader, or whether a person's name or his role appeals more to the reader (i.e., John Barrasso or GOP Senator).
We will never be done optimizing, whether it's optimizing for search engines or conversions. It is a circular process of research and testing. Sound hypotheses and automation will help us do it more efficiently.
Let the optimization begin.