How to Solve the ‘What Should I Test?’ Problem
One of the most common and mundane questions people ask about conversion optimization is “what should I test?” Usually, this is the wrong question to begin with. Good experimentation tends to be hypothesis-driven, led by quality research and a solid reason for experimentation. Starting off with the “what should I test” question tends to lead
One of the most common and mundane questions people ask about conversion optimization is “what should I test?”
Usually, this is the wrong question to begin with. Good experimentation tends to be hypothesis-driven, led by quality research and a solid reason for experimentation. Starting off with the “what should I test” question tends to lead marketers down the path of testing trivial things like CTA colors just because it’s what others are doing.
Instead, you can find opportunity areas with a relatively small amount of qualitative user research coupled with some digital analytics analysis.
This post will outline a few ways to dig into Google Analytics in particular to find areas you might want to look into improving on your site.
Since analytics is best approached by asking your own unique business questions and then seeking answers, it won’t be an extensive list of reports that help with optimization – but the ones listed below have helped me over and over in the past when looking at what to test and where to start.
Note: this article assumes a certain level of knowledge with Google Analytics. If you’ve just set up your account and haven’t created goals, events, or checked that your implementation isn’t broken, do that stuff first. The reports to follow aren’t complicated, but there are a few prerequisites. If you haven’t met those, read this post on setting up GA and getting started or take CXL’s beginner’s Google Analytics course.
How Can I Find Which Pages on My Site Need Optimization?
This is really the first question you want to answer: where do I start optimizing?
Not all landing pages are performing equally, so it’s important to prioritize based on bigger opportunities.
There are many ways to approach this, but one of the best is by using engagement metrics such as bounce rate as a proxy for performance, and comparing landing pages to the site average.
This is fairly easy to do. Just go to Behavior > Site Content > Landing Pages and use the comparison feature on the right hand side of the screen. Then select Bounce Rate from the drop down menu like I showed above:
This shows you all pages, but it’s likely that you’ll want to get more granular. For instance, if you have product category pages, you can narrow down and see the Bounce Rates of only pages in that category. Just search your category indicator, e.g. “drinkware,” and you’ll see only those pages and can nail down which ones need work:
Note: you can also compare other metrics, like Goal Conversion Rate. The point is the comparison feature. Instead of guessing which pages to start optimizing, you can see which ones are over and under-performing against the site average for that particular metric.
When you’re prioritizing A/B tests, no matter which framework you use, it’s likely that one input you use for prioritization is “impact.” This usually comes down to how much actual impact, or potential, the test could have.
Well, those with lower than average conversion rates and also high traffic represent the biggest potential impact. There’s a lot of room for improvement on these pages, and because they’re higher traffic, the impact is larger than the marginal increases in value you’d see with lower trafficked pages.
Is the Homepage Really the Most Important Page?
Most people think their homepage is the most important page, so they start optimizing there by default. Often, they’re right, but it doesn’t hurt to check.
You can also find this in the Behavior > Landing pages report. Just flip this information into a bar chart and order by sessions (which is the default). This shows you the pages with the most sessions:
In the case above, which is from the Google Analytics Demo Account, the home page is clearly the most important page, with ~60% of the total sessions.
This report is a quick and easy look at the potential impact of tests you’ll run. Sure, if a page is underperforming, it’s likely somewhere you should focus your effort. But also important is the number of users your tests effect.
If nobody is on the page, it doesn’t matter as much if the UX is broken, but if 60% of sessions start on a particular page, you can bet your efforts there are largely worth it.
How Many People Convert After Seeing a Specific Piece of Content?
As someone who started in content marketing, this is a particularly interesting report to me. We push out content all the time, and we track metrics like sessions, email opt-ins, and traffic by source/medium, but how often do we ask, “how many people purchase after viewing a specific piece of content?”
Normally, you might think to go to a report like Behavior > All Pages and find the conversion rate or transaction account per page, but it doesn’t work.
Why is this? Pages don’t convert, people do. So, you have to step outside of the basics, and do a bit of advanced segmentation:
- Copy the URL of the page you’d like to analyze.
- Click +Add Segment
- Look on the left hand side and find the options “Conditions” and “Sequences.”
You might be used to using “conditions” to create advanced segments, but for this case we’re going to use “sequences.” Why? Because if you just add the landing page parameter and the transaction parameter but don’t specify the sequence, they might have converted first and then later looked at that page. So set it up like this:
In this case, I like to set the segment as “Users,” because it doesn’t matter as much to me if they convert in that specific session. Though depending on your situation, you may care more about sessions. I usually like to look at it both ways, as it gives a deeper look at those who convert pretty quickly (session-based) and those who convert at all after viewing (user-based).
Doing this, you can analyze your most trafficked pages and see if people are converting after seeing them.
In addition, if you save that segment, you can see anything about them: what countries they came from, what age they are, what pages they visited, was what they purchased even relevant to that page? Advanced segments are a treasure to your deeper analysis, and the possibilities for insights are awesome.
For content marketers especially, this can be incredibly insightful. Are the blog posts that get the most traffic valuable for the business? I’ll tell you in our case, some of our highest organic traffic pages underperform when it comes to delivering business value. This type of information can help you plan out your content and also see which content is most useful to your customers.
Where Are People Dropping Out of the Funnel?
Every site that has a funnel (most do) should start with funnel visualization. This Google Analytics report is easy enough to set up and will tell you how much traffic is dropping off at each funnel step. For instance, here’s a 3-step funnel:
Where is the leak in this funnel? Every step has some dropping off, but here we can see that the first step (home) has the biggest leak.
However, depending on the site and situation, this may not be a bad thing. Home pages generally get more than simple, targeted traffic. People come to the page, shop around, leave, come back on mobile, etc. The home page generally doesn’t assume a linear path to the goal page.
But once they’ve entered the more targeted, linear stage of the funnel, where’s the biggest drop off? The Subscribe page. Only 25.78% of people proceed to the next step. This is clearly the best place to start optimizing, because if one has put the effort into getting that far in the funnel, it’s assumed that there is interest and intent. It’s probably easiest to plug the leaks in the funnel in this step, and the most impactful.
Funnels aren’t just for e-commerce, either. They can be used for a variety of purposes – any time there is a preferred linear navigation path, really. Some other use cases could be multi-step sign ups, contact form completions, page navigation, and more.
For more information on setting up goals and funnel visualizations, check out Krista Seiden’s excellent article on the topic here.
Which Promotion/Image Works Best?
If you have enhanced e-commerce setup, you have a whole suite of GA features to help you retrieve more granular data. One of which answers a common question: how are my promotional images actually converting?
Especially if you’re in the e-commerce industry, you’ll most likely have promotional images on your home page. These tend to change quite often, as you’ll have different specials during different seasons and dates. You might even have an auto-rotating slider, yet have no idea how well each individual slide converts.
To find this information, all you need to do is go to the Conversions > Marketing > Internal Promotions. This shows you the data on your internal promotions you have set up to track, such as views, clicks, CTR, and goal/conversion data as well.
Of course, this can work for those outside of e-commerce as well. You can just set your conversions drop down menu to another goal, such as a lead generation form.
Most people are just guessing when it comes to internal promotions and banners (or taking orders from HiPPOs who simply need their offer on the home page). With only a few minutes of effort, you can see the actual effectiveness of these promotions.
Post-Test Analysis in Google Analytics
Best practice: in addition to analyzing your test data within your testing platform, also send the data somewhere other than the testing tool. This way, you can slice and dice the data in a way that can uncover hidden insights. It also helps to double check the integrity of the data of any given system.
One of the most common post-test analyses is to look for greater effects in certain segments, though be careful: you can fall prey to the multiple comparisons problem.
This article probably isn’t the right place to jump into the intricacies of why this happens, but essentially, the more segments you look at, the greater your chance for false positives (especially with segments of very low sample sizes).
Can you fix this problem? Yes, there are ways around it (e.g. Sidak Correction), though that, too, is beyond the scope of this article. If you’re interested, Chris Stucchio wrote a great article on the topic.
The real value of passing your experiment data to Google Analytics (or another data collection system) is that you can then analyze the behavioral differences of different treatments beyond the goals you’ve set up. This can give way to far greater understanding of your customers as well as greater hypotheses for future tests.
This is an introduction to using Google Analytics to find conversion optimization opportunities. It’s not extensive, and it’s not specific to individual businesses.
However, if you take these basic reports and the principles from them – things like segmentation, granularity, funnel analysis, and curiosity – then you can perform similar analyses on your own site, but deeper and more specific to your business.
If you want more reports that are easy to produce, here’s a blog post with 10 specific to conversion optimization.