If landing page optimization and conversion funnel testing are not on the list of items that you're currently focusing on, then you're not getting the most out of the visitors that are coming to your site. This is true for all types of websites, whether you want your visitors to make a purchase or donate so you can increase revenue, download a white paper so they can learn more about a specific topic, watch a video, spend more time engaging with the content on your site, etc. etc.
During the Landing Page Optimization session at SES New York 2013, Justin Cutroni (@justincutroni), Analytics Advocate at Google, discussed:
- Why testing is important for your business.
- How to identify opportunities for testing through data.
- The different elements on a page you want to consider testing.
- What types of tests are available.
- Some tools you can use.
- The basic elements of setting up a test.
Learn Your Audience, Improve Your Results
Cutroni started off by talking about how to approach testing from a high level and putting it into perspective.
Start by thinking about your typical conversion funnel (i.e., the common steps that it takes for someone to complete an action on your website). The purpose of conversion funnel testing is to optimize that process ‐ you're trying to figure out how to get more people through the funnel more effectively by making it easier for them to take each next step through to the end.
To help you understand what's going on in the conversion funnel, you have access to more data than ever before, e.g., you have click path data, micro conversion data, macro conversions, etc. And you can optimize around all these different data points.
What's great about testing, Cutroni said, is that you can automatically increase the return on a lot of things you're doing today. If you can improve the efficiency of your website through landing page testing, conversion funnel testing, etc. then you should do it.
If you can increase the amount of purchases, donations, signups, subscriptions, downloads, etc. (all of the different processes) on your site then you're going to get a much higher return on investment at the upper part of the marketing funnel, whether you're running display or search ads, focusing on search engine optimization (SEO) or content marketing.
This is why every business should be testing, but not enough do. The concept of figuring out and identifying what works best for your company, rather than just going along with the common recommendations or best practices, has been embedded in marketing for a very long time and is nothing new. You can see this in direct mail, product catalogs, store coupons, etc. just like you can with email marketing, content marketing, pay-per-click (PPC) advertising, and so on.
How to Identify Testing Opportunities
Start by looking at your data.
You have tons of data, but there are some common places that people start to get an idea of how their pages are performing for their visitors. One of those is with bounce rate.
Cutroni recommended that you identify your top landing pages on your site and look for the ones that have a really high bounce rate.
So what’s a high bounce rate? According to Avinash Kaushik, Digital Marketing Evangelist at Google, "as a benchmark from my own personal experience over the years it is hard to get a bounce rate under 20%. Anything over 35% is a cause for concern and anything above 50% is worrying."
Bounce rate is subjective and depending on the pages you’re evaluating, what appears to be bad might not be. But in general, start with the pages that have a bounce rate over 50 percent, because if they’re bouncing they’re not taking the actions you want them to, e.g., they’re not adding products to the cart, they’re not viewing multiple pages in an article, and so on as Justin mentioned.
Once you have your bounce rates by top landing pages, Justin recommends taking a step back and asking yourself "why" you think the bounce rates are so high for a specific page or group of pages.
In order to help answer that question he created this handy Google Analytics Custom Report, called Landing Page Testing Ideas, which, in addition to bounce rate, includes the following engagement and acquisition metrics: average visit duration and % new visits. It also includes these outcome metrics: conversion rate, transaction, revenue and per-visit value (i.e., what’s the value of everyone coming through this landing page). Adding in these value metrics is really important and gives you a little bit more ammunition.
Why It's Important to Consider All This Data as You Build Your Testing Plan
Often times key stakeholders at your company may not know the value of website testing. So it’s your job to educate them. One of the easiest ways to do that is showing them how testing can affect the bottom line.
If you ever have any trouble trying to convince your boss, IT staff, or other teams that you want to start testing try and show them current conversion numbers, or better yet current revenue numbers, Cutroni said. This ends up being very helpful in the conversation because you frame the issue in a way that they already understand and likely already thinking and talking about.
Bounce rate is great, but when you can say, "look how much money we spent on this awesome new landing page but here’s how little revenue we received from it" then it starts to really resonate with people and get them on board with wanting to try and improve.
While you’re creating these reports with acquisition, engagement and outcomes metrics, don’t forget about the power of segmentation. For example, if you add another dimension that includes where people are coming from, with source / medium in Google Analytics, it gives you even more insight into bounce rate, and conversions and revenue, he said. A person coming in from email is likely to have a really different path and action than someone from organic, and so on.
Key Takeaways From Two Case Studies
Cutroni then went into two case studies he included in his presentation, one for Kapitall, a new type of investment company that added a level of gamification into their training platform, and the other for Amari, a luxury hotel brand primarily in Asian countries.
Think about the audience coming to your landing pages and how that matches with what you’re trying to offer them (i.e., what’s your value proposition). Create an experience that is compelling for them to take the action you want them to take.
In the case of Kapitall that was coming up with landing pages that moved away from kind of the boring and straightforward white-background pages to something that had a game feel to them – an "Angry Birds"-themed page was the winner there.
For Amari that was creating supporting information and a clearer call-to-action up-front by finding out that online users were ready to book a room on their first visit to the website, i.e., they were lower in the marketing funnel.
Cutroni pointed out that it’s not always just about identifying high bounces, etc. It’s about understanding the bigger macro behaviors of those people, your potential customers, to see how long does it take them to complete something, what’s their intent, and so on.
By having the data to support that, you can start to develop and come up with a strategy for creating engaging content on a page that you’re sending people to.
What Kind of Test Should You Run?
Landing page optimization and conversion funnel testing has become much easier over time, and there are many tools and options out there to get started (e.g., Optimizely, Google Content Experiments, Unbounce). That said, you still have to plan for testing, and there are some things you need to understand about the pages that are involved.
Once you’ve gone through and identified a few pages you want to start testing, the next thing you want to do is figure out what type of test you want to run. Almost all of the testing platforms out there today support two types of tests:
- A/B Testing: You have one landing page which is version A, or the control, and then you came up with multiple variations of that page to test against that control. This is probably the best way to get into testing, since the implementation is simple, and you have a little more free-reign over the creative.
- Multivariate Testing (MVT): Cutroni likes to compare MVT to using a scalpel – it's an opportunity to really fine-tune the pages and the content on your site once you start testing. MVT is different than A/B testing, since rather than just doing a bulk change and testing different variations of full pages with A/B testing, you can carve up the page in different sections and you can test in real-time different combinations of these different sections with MVT. This can include headlines, text or content on the page, forms, buttons, etc. So you come up with two or three different variations for each of these elements or sections and then your testing platform will put them in a magic blender, shake it up and show the different combinations in real-time to your website visitors.
How to Set up a Test
When you running a test, there's usually not a lot of information that you need to give each tool. Here are three things you need to consider according to Cutroni:
- What are you testing for – what's your conversion event? Conversion events can include selling products, generating a lead, scheduling an appointment, etc. It really depends on your business model. But you want to test against something pretty specific. For example, don't just say, "here's a page that has a high bounce rate and I want to drive more conversions." Rather, start small with reducing the bounce rate first, and once you make progress there, you might discover there are other issues with the site that you need to fix before you can actually start to increase conversions.
- If you need to convince stakeholders internally to run tests on your site, one tactic you can use to reduce anxiety is limiting the amount of people that are thrown into the tests. Something great that most testing tools allow you to do is create a threshold of how many people you want to include in the test. You might have employees that feel like testing is going to end their world ‐ $X in revenue comes through the site every single day, and "why try and fix what's not broken." If that's the case, one thing you can do in addition to giving them your estimates as to what difference you can make, is showing them you can start small – you don't have to throw 100 percent of your traffic into the same test. Show them you can start with 50% or maybe even 25 percent of your overall web traffic. Most testing tools allow you to have some control in that area.
So then once you start running the test, and your data builds, you'll end up with really pretty graphs that typically include the following:
- How many people are flowing through your tests.
- How your different variations are doing from day-to-day or week-to-week.
- Do the different variations actually have a chance at beating your original.
What's a Good Length of a Test?
Sometimes people aren't really sure how long they should let us a test run to get to statistical significance, and they're just eager to just let a test run for 8 months. Not only does that make you impatient, but if devices are changing and the environments are changing then the test isn't really valid if it's stretched out over such a long period of time.
You really need to go just a couple or a few weeks for a test – nothing longer than two months according to Cutroni.
How Can You Test Pages on Other People's Websites (e.g., Third Party Sites?
That said, if they're going to let you put code on the site, then you can set up your tool to test different variations and versions. But that's really the only way.
Can You Apply These Testing Principle to Ads?
Yes, surprisingly 90 percent of advertisers are hitting pause, starting new ads, stopping them – a lot of activity is going on, and if you were to just check your ads to see if there was 95 percent statistical significance in their tests, often times you'll find there's enough to determine if there's a winner or loser.
Do Long Forms on a Page Suck in Terms of Driving Performance?
Cutroni agrees, yes, 99 percent of the time long forms suck. That said, as marketers some of the data that we base our assumptions off of is good, but it's not always great. And the only way you can really understand what's going on with long forms on your site is to track every step along the way.
It's unbelievable what you start to learn once you dive into the data; you'd be surprised what people will do if they're really interested in your site or what you're offering. So you always need to test something for your own particular situation before making any conclusions.