On a scale of digital marketing practices, conversion rate optimization (CRO) is the endpoint of web efficiency. CRO is the process of improving user experience in order to increase conversions. A practitioner of CRO analyses user behaviour to see where drop-offs occur, and creates hypotheses on how to remove roadblocks and change a website for the better, which are then tested. These could be changes from on-site design, to backend functionality, with the objective placed on getting users from A to B as easily and quickly as possible. In this guide, we’ll look at the process from beginning to end.
What is CRO?
A conversion rate is determined by the number of people who reach a defined goal, out of another number of people who don’t. In order to determine a conversion rate, you divide the total number of conversions by the number of visitors to your site. For example, a site with 10000 visitors and 100 conversions would have a conversion rate of 1%. As such, conversions are usually only a fraction of the amount of people who visit any given site, depending on the goal. Goals themselves could include purchases, sign-ups, or page views.
So why does this all matter? Well, you could have the best SEO in the world, a PPC campaign that runs into the millions, or a social strategy that brings users to your site in droves. Yet without thinking about how your site works for a user, traffic could have a significant drop-off, and for all the good work you’ve put into marketing, you may not receive a worthy return. Let’s look at two factors that could cause this:
- On-site friction from slow site speed, excess pages, misdirection, multiple 404’s, and others that cause users to leave.
- Cognitive overloading or a complicated funnel that creates doubt and indecision in the user’s mind.
It’s the job of conversion rate optimisers to gather information on user behaviour, analyse the issues, test features, and suggest improvements that can help take these pain points away. In essence, CRO is:
- A systematic approach to improving the performance of websites, campaigns, or marketing channels.
- Informed by insights – specifically, both analytics and user feedback.
- Taking the traffic you already have and making the most of it.
Why bother with CRO?
CRO is important because it improves sales, whether by direct or indirect means. It also potentially leads to increased customer satisfaction. Let’s make an example; a buyer places an item in their shopping cart while on a bus to work, but before she can finish the transaction she turns off her phone. Since she is signed up to the website, it’s a likelihood that she would like to purchase the item at a later time. The recommendation would be to save any item placed in a shopping cart by a user, as opposed to emptying it on sign-out. This may have the effect of making the shopping experience easier, as our user doesn’t have to find the same item again. However, this is a logical assumption, and only by testing one case against another would we be able (as CRO’s) to enact the change. Think of CRO like detective work. You have to investigate issues, create hypotheses, and find the culprits that prevent conversions, or discover those missed opportunities.
To break it down further, here are three more reasons why CRO can be beneficial to you:
- CRO is cost effective when compared to paid advertising. With a smooth conversion funnel you could gain more from the same amount of traffic, without having to boost numbers to reach conversions. In fact, doubling your conversion rate means halving your cost-per-acquisition (or how much it costs to attract new customers.)
- CRO avoids the problem of shortened attention. By giving the user what they want, quicker, you’ll reduce the amount of lost traffic from those who have come to expect a fast experience, and who have a very entrenched understanding of what user experience should be.
- There are always ways to improve. No matter how well-designed your site is, it’s likely that something is missing. No matter how granular, there are always means to improve conversions, even by 1%.
One final note, CRO isn’t necessarily just about websites. You can also work on the effectiveness of email campaigns, social traffic, or advertising. Any channel can be optimised, but for our current purposes, we’ll focus on sites, and talk about campaigns and wider activities in a later article.
A Note on Snake Oil
Before we look at the how CRO accomplishes its aim, I need to cover off some myths that everyone falls into when they first learn about this kind of marketing. Within CRO there can be a lot of talk about seemingly arbitrary changes that lead to huge improvements, talk which becomes a sales tactic, as opposed to definitive marketing methods. Keep in mind that CRO is not about:
- Guessing what can be improved, or following the crowd.
- Basing decision making on the opinions of the highest paid stakeholder.
- Traffic (although small SEO gains can be an incidental benefit of good CRO).
- Simple “tricks!”, such as colour changes, single word alterations, that are guaranteed to boost conversion rates.
The basic tenet of CRO is that just because it worked for one site, doesn’t mean it will work for yours. You need to test any hypothesis you come up with, and hypotheses need to be grounded in logic. Think like a user first, do you really care about buttons? Perhaps later in a campaign it could become a consideration, but what really bothers you now, as a newcomer to the site?
The Conversion Funnel
Before you can begin diving into data, you’ll need access to your site’s Google Analytics. Once you have it, check to see if goals and conversion funnels have been created. Sometimes websites may not have these set up, and it’s a necessary step to either create new ones, or fully understand the current path to a conversion on the site, before you can continue.
What is a Conversion Funnel?
A conversion funnel is the journey a purchaser takes from beginning to end (from discovery to sale). However, in Analytics, a conversion funnel has to be defined, and it is a vital means to understanding the customer journey, as well as where users often drop off. Without it, you can’t hope to understand how a site’s conversions work, or where the problems or potential areas of improvement might be. Setting them up is simple, but would require endless image manipulation (though they’re in the works!) For now, watch KissMetric’s video, which will illuminate the process.
The Planning Phase
In order for any test to be a success, you’ll need to do a lot of the work up front. CRO begins by learning about company identity, pain points, and consumer opinions. Creating surveys, and utilising analytics data are essential for finding problems perceived by both you, your colleagues, and the consumer.
With funnels set up, you will soon be able to dive into data. While this is being gathered you can lay some groundwork by studying business identity, and working within the company. Speak to people within the organisation who are outside of the marketing team. Sales staff can tell you about how they sell products, what they view are unique propositions, and provide you with customer feedback. You could potentially then follow this trail, talking to customer support staff to find out what day-to-day problems they and their customers face. Two essential pieces of information to dig up are typical demographics that the site receives (a consumer profile will be helpful if there is one available), and the amount of traffic the site receives, either weekly or monthly, as well.
Big or Small?
Site size determines a lot about your strategy. Rather than riff on the subject, keep this general rule in mind: go big if your site is small. Don’t test changes like text size, instead design complete new landing pages. Why? Tests take much longer when your sample size is much smaller. If bigger websites get 10000+ clicks a week, but yours only receives 100, it will take a much longer time to reach the sample size required in order to recommend changes. Furthermore, if only an average of 40 of those 100 move through the first stage of the funnel, how easy would it be to determine that a 10% increase or decrease of visits (in this case, four people) per test would be due to a new banner, or offer, and not just an regular change in traffic? It’s best to work big with little, and tinker with big.
While I talk a lot about problems that CRO could solve, there is also a lot of scope for CRO’s to define positive strategies that go beyond day-to-day operations. Say your marketing team is trying to improve sign-ups to their newsletter, with a secondary KPI of improved amplification through social sharing. You could suggest running variations on:
- Email sign-up modals (whether on time to appearance, or size, or text)
- Content upgrades
- Social sharing buttons within emails
Which could all see benefits previously unattained. By taking the initiative here you’re not only reacting to user need, but you’re proactively shaping the campaign. CRO’s aren’t just problem solvers, but can play an active part within the marketing team.
As part of your consumer research, creating a survey at the end (or beginning) of the conversion funnel is one of the best methods to finding out what users think about your website. Although surveys themselves could be considered their own aspect of CRO (with conversion goals attached), you should use them for planning. Survey Monkey and Qualaroo (see the list at the end of the article) are simple tools that allow you to create surveys placed on site. Qualaroo seem a little more focused on CRO than others, but try as many as you’d like, as there are dozens of survey alternatives available. For more information about surveys and customer interaction, Kissmetrics have a solid write-up here.
Between company research and analytics data, you should begin to have a feel for the problems that your site faces. It could be any from the list below:
- Site navigation: is there enough information? Too much? Is this intuitive? Are there many error pages?
- Call to action (CTA): is this clear and visible? Would a new visitor understand what to do next? Could the language change?
- Images: are your images adding to the page or distracting from the call to action?
- Copy: is the written text on your site well done? Are there errors?
- Translation: For multi-national websites, how does copy fare in different markets?
- Value propositions: are these stated well?
- Testimonials and social proof: do you have them? Do they speak to your customers?
- Site security: is it obvious to a new user that you have the proper protocols set up to ensure their safety?
- Number of clicks to complete an action: are there unnecessary steps?
- Site speed: is your site loading in a time that won’t cause frustration?
- Mobile friendliness: with the majority of traffic coming from mobile devices, how well does your site work on a phone, or tablet?
- Browser friendliness: Safari, Chrome, or Firefox? How well does your site work on each?
Ultimately, what you’re looking for are pages with higher than average exit and bounce rates, and any of these reasons above (and more) could be the cause. While the above examples deal with specific site factors, emotional needs of the user also need to be factored in too:
- Relevance: how closely does the content on your page match what your visitors are expecting to see? How closely does your value proposition match their needs?
- Clarity: how clear is your value proposition, main message, and call-to-action?
- Anxiety: are there elements on your page (or missing from your page) that could create uncertainty in your customer’s mind?
- Distraction: what is the first thing you see on the page? Does it help or hurt your main purpose? What does the page offer that is conflicting or off-target?
- Urgency: why should your visitors take action now? What incentives, offers, tone, and presentation will move them to action immediately? Are you aiming for a purchase now, or purchase on return?
The simplest thing you can do is to analyse the page as though you’ve never seen them before. Is navigation clear? Are links visible? Refresh pages to see what stands out (or doesn’t). Read the content aloud to see how it sounds Take notes on anything you can think of for later testing. Even follow the purchase funnel yourself, from beginning to end. Write down anything you notice along the way, noting anything that doesn’t seem as efficient as it could be, as these will be the seeds of future testing.
Once the problems have been mapped out, consider sitting down with stakeholders and walk them through the sales process, and include each touch-point between the customer and the website or marketing materials such as email. From here, discuss where problems can occur, and where they feel you might best focus your efforts first. This could help you to identify more problems.
A Word on Bounce Rate
Bounce rate is a phrase that will come up again and again in your discussions on metrics. Bounce rate is “the percentage of visitors to a particular website who navigate away after viewing only one page.” The action of leaving includes closing the window or tab, typing in a different URL, clicking on an external link, using the back button or timing out. For many clients and marketers, it is somewhat the holy grail of metrics. After all, if people have navigated to the page, why should they leave?
As it happens there could be a multitude of reasons behind bouncing, and only a few might have anything to do with your site. For instance, misclicks (or, with touch technology, misprints) and wrongful search intent account for many users who bounce. Also, spam bots, and users with malicious intent, may jump from site to site, bringing overall rates down. In other cases, it could simply be a case of the user finding the information that they need in the ten or so seconds of being on your site. They leave because they are satisfied with the page. As such, while a bounce rate that is over 50% per page may be a concern, many sites (particularly larger ones) will have a high bounce rate regardless of CRO. It all depends on intent, and as such you should not let it define your approach to testing. See more on bounce rate here.
Now you should have analytics data, a list of problems you have identified on site, and the beginnings of ideas that you can use to tackle the problems. If you haven’t already, the end of the planning phase is the time to define and list the problems, because they will be at the core of the execution phase. Once completed, it’s time to define your goals, which will inform your hypotheses, your tests, and how you interpret your results.
Defining goals is no more than a sentence. It could be “I want to to increase the traffic to my new offer,” or “I want to increase sales in time with promotional activity,” or even “I want to decrease the amount of negative customer feedback.” It’s from these goals informed by your research that you then have the basis for your activity, it even can inform the kind of test you will undertake.
A/B or multivariate?
Within CRO there are two main kinds of tests you can conduct: ‘A/B testing’, and ‘multivariate testing’. At its core, A/B is the practice of testing a single aspect of a web page against a control, hence its other name of ‘split testing’. Multivariate testing is the practice of splitting up your traffic towards multiple versions of the same page. So you test many aspects all at once. There are advantages and disadvantages to each, so let’s quickly run through them.
What’s good about it?
A/B testing is supreme for small sites with big changes. For instance, by creating a new landing page, you can see how well it improves against the old site. It’s easy to interpret; there will either be an increase or decrease in conversions against the control, and you won’t have to to delve into three or four other combinations of results. A final benefit is that it helps win converts with stakeholders, because it demonstrates without question what works and what doesn’t, which is great for new clients.
What’s not so great?
They take time, because you can only take on one aspect at a time. While this is effective for big alterations, when you want to drill down into improvements that may only swing a percentage or so, multivariate is a better option (but only if you have the traffic to create statistically meaningful results.)
What’s good about it?
Multivariate drills down to testing individual elements on the page, such as banners and images. The goal of multivariate testing is to determine which combination of variations performs best out of all possibilities. It’s great for late stage optimization.
What’s not so great?
Your site will require a huge amount of traffic in order to test the larger number of combinations successfully, so many sites won’t be able to conduct statistically relevant multivariates at once. Because of the nature of these tests, the number of variations in a test can add up quickly, and results may not be so easy to interpret. The result of a many-variation test is that the allocated traffic to each variation is lower, causing statistical issues.
By now, you’ll have thought hard about what problems the site has, or what changes you want to make to reach KPI targets. So now it’s time to create a hypothesis informed from your data gathering. Then, you can move onto testing.
A test can be anything. Altering site copy, images, the way price is displayed, varying calls to action, etc. But while you have a freedom of choice in what you test, always base it on your goals. For instance, if you want more email signups, you may want to test a modal pop-up when the site is opened, or you may want to alter the text around the site to entice people in. Or you may think that improving the emails themselves is its own test, especially if you have a more fundamental objective of converting receivers of emails to return to the site. There are so many action you can take, but remember to always go back to your goals when you create a list of testing opportunities.
Defining your sample size is absolutely crucial to test success. To be confident about your results, they will need to be statistically significant, which means you need combination of enough users, and a margin of error as small as possible. As a rule of thumb, if you have a larger sample, your margin of error will be lower. Basically, big sample, small margin; small sample, relatively huge margin. Tools such as Evan’s Awesome A/B toolset, and in fact many of the paid A/B testing services, will tell you what your margin will be, and if you want a lot more detail on the subject, Craig Bradford of Moz does an excellent job of talking about it here.
Setting up a Test
Tools such as Optimizely, Unbounce, and Google Content Experiments are the platforms through which you can run your tests. This is the easy part, as the heavy lifting is done for you, but the machinations depend on which tool you want to use. Once set up, it’s a time to wait on the results.
When segmenting traffic, you’ll want to aim for specific users. Consider your demographics: a mixture of new users vs. returning visits may skew results, and you may get more reliable data by only testing one faction over another. Also, begin from the start of the funnel, and follow the process in order. Then, go back and start again.
Just a small note about SEO, as you may fear that conducting tests may have an adverse effect on your current rankings. Google is very encouraging when it comes to CRO, it even has its own public tool for A/B. However, there are some best practices you may want to look at before testing.
- Add the rel=”canonical” tag to the original version of the page if you have not already done so. This helps prevent crawlers from attempting to rank different versions of the same page
- If it comes up, use 302 (temporary) redirects instead of 301 (permanent) redirects. This lets crawlers know search engines that the redirect isn’t here to stay, and as such a new URL won’t replace the current one
- Don’t cloak. Cloaking is the practice of showing search engines different content than a typical visitor would see. It can be achieved by directing human traffic one way, and bots another. By doing so you run the risk of de-ranked, and in a time of decreasing significance for technical SEO, it’s an almighty risk to run for something that doesn’t benefit anyone
- Running tests for longer than you need won’t only waste your time, but could be perceived as gaming SEO. Only test until you’ve reached statistical relevance, then take it down
Results, and How to Understand Them
After a test is complete, you’ll see one of three variations. Either conversions will have improved, stayed the same, or decreased. If there is no doubt of an improvement, you can recommend that the change should be implemented on-site. That part is simple. However, what if your hypothesis was incorrect, and nothing changes, or what if your conversions are lower than when you began?
Fortunately, neither result is necessarily a bad thing, it’s a crucial part of testing and stakeholders should have been warned that these are potential outcomes. What you have achieved is ruling out a single avenue of improvement, learned a little bit more about your users want, and have more data to cater to your audience. The process to continue this goes something like:
- Review your analytics and user survey data with this in mind to see if you can gain new insights
- Form a new hypothesis
- Design and conduct a new test
Asking “why?” can really help, for instance, if altered copy focuses on price, but loses engagement, would this mean that your users are more focused on brand, experience, and so on? These tweaks make up the backbone of CRO, and should all inform future decision making. Whatever the result, there will always be more tests, and the only variance of course is on whether you repeat testing on the same or on a different feature.
There are the names you’ll see pop up again and again in discussions of CRO tools. Some are free, but most require payment after a short trial.
These three are the paid tools that most look at when wanting to do CRO.
Perform A/B tests (but not multivariate) against your Google Analytics conversion goals. Free.
A rundown on statistical relevance of your results. Also free.
Usertesting recruits users to test your site.
Watch recorded sessions of users visiting your site. Includes heat maps.
- Google Analytics User Explorer
Allows detailed analysis of anonymous user interaction. Free
You create the survey, Google recruits the participants and provides results and analysis.
Poll ‘em with Qualaroo. That’s not their tagline, but maybe it should be.
Probably the most well known polling tool out there.
Heat and scroll maps
View where visitors click with heatmaps.
Site speed test
A critical but easy check, site speed needs to be fast, especially on mobile devices. This tool allows you to see how Google views site speed, and provides simple methods in how to fix what’s preventing a quick time.fda
A slightly more comprehensive, non-Google site speed checker.