How to test websites?
Benefits of website testing
Let's underline that you can improve every page, but you won't be able to effectively improve a website without construing a measurable test for it.
You might succeed in tweaking one webpage element based on a hunch, but you won't know how change X caused a Y effect. Without construing a valid website test, you won't know how all the variables affect each other and bring about changes in web traffic, time on page, conversion rates or page bounces.
Warning: website testing is arduous and time consuming. It can also be boring. But, website testing carries with it valuable benefits.
Let's say you designed a new website, or added a new landing page. You spent time writing interesting and persuasive copy filled with relevant keywords. You touched up all design and layout elements to meet your preferences. You remembered to add links in all the appropriate places, plugging your site online. You're excited by the thought of how many new users/readers/clients this new website will help you attract. And what? And, nothing.
Seemingly visitors enter the site, but spend little time on it. Web users go to your page, but they don't click on any links. You think "hey, this is temporary... we're just warming up, let's give it some time to pick up". And then a month passes and your bounce rate is 3 times lower than average. This is when General Cambronne at Waterloo says "merde".
But the page is so cool, so unique
Here lies the problem. A 'cool' page is a question of taste, while a 'unique' design is a matter of perspective. Taste is something subjective, while perspective you acquire with time. Both conditions are common enemies of new websites. Both conditions, or lack thereof, are traps into which some website designers fall.
Something that appears magnificent and perfect to us, may appear completely horrendous to our readers/viewers. To find out how those readers/viewers perceive your site, it is best to test it. Let's get to work.
Step 1 - gather data
You won't test anything, nor make any relevant observations, without first having analyzed website data. We assume that your website already has some basic analytics software to measure its statistics (eg. Google Analytics). So, check your analytics across a few segments, like:
- how many unique visitors enter the site
- how much time does the average visitor spend on a page
- what are the most popular links clicked on
- what are the most/least popular pages
- what pages are used to reach your contact forms (or other relevant pages)
- what is the exit or bounce rate from that page, in comparison to the rest of the site
By gathering this data you will have basic information to build on in the following step.
Step 2 - measure and analyze web traffic
After carefully analyzing all the data, you will probably begin to notice some website elements that might not perform up to par. You will notice that visitors leave the page too early, before having read the content. Or, you might find that they're not clicking on your buttons, banners or links that point to relevant information or lead them to fill out contact forms.
After analyzing all statistics you will learn, which elements require overhaul and, which should be tweaked and tested to raise their conversion rates.
Step 3 - test
Having gathered all this information you are ready to begin testing your website. You can perform a website test by:
- changing a single element on a page and observing effects for a given period of time
- using a website testing software/application
Option 1)
If you have limited finances and cannot purchase software or an application for website testing, then you can perform a split A/B test. An A/B split test requires you to measure web data, statistics and general goals of a page for a given amount of time. Then, you change one element on that page and measure effects throughout the following week. Boring? Oh yes. Effective? Very.
After a week, you compare the original page (A) to the tested page (B). If the tested page shows improvement, you substitute it for the original. Then, you begin the process anew creating a copy of this page, changing one element and testing it for another week. If the new version (B) again proves better than the new original (A), then you implement the changes and begin another test until you reach satisfying results.
It is a very manual process. It is much more time consuming than if one were to use website testing applications. Nonetheless, an A/B split test is just as effective.
Note: it is possible to test two versions or more at the same time by redirecting 50% of site's traffic to version A and, 50% to version B. This will greately reduce the amount of time it takes to test all elements. However, this shortcut requires experience in writing scripts.
Option 2)
There are many website testing applications online. From basic tools to advanced professional software, or even website testing services offered by specialized companies.
To perform a website test individually it is best to choose a software application that will aid in this effort. Since software choices are abundant and dependent on your budget, and because the internet is a huge market place, all one needs to do is go shopping.
We believe that for beginners, small firms or companies with limited budgets, a good place to start is with 'website optimizer' offered by Google AdWords and, for free by Google Analytics. Google Adwords doesn't require significant capital (10 dollars per month suffices) and 'website optimizer' comes included. The only thing that's required is basic knowledge of HTML in order to imbed the code.
After having chosen a website testing application and gathered data, determine what areas of your website require testing and improvement.
How does a test look?
To illustrate a website testing process, we will use an example of a webpage that we once tested for a B2B company. That company created a landing page for one of its services in an effort to attract new clients. Regrettably, that landing page wasn't performing up to expectations.
To define what needed improvement, we examined their page's statistics and noticed that:
- it showed few clicks on main link-buttons (its click-through-rate was less than 2%)
- the content was in some way deterring visitors (exit rate was at 64%)
- most visitors spent on average only 37 seconds on the page (while the website had an average of almost 3 min.).
The purpose of this landing page was to convert visitors into leads via contact forms that were located behind links. Knowing that the main links had a low CTR and that most of the visitors quickly exited the page, we suspected that two elements required initial testing to improve conversion:
- link buttons
- content of the main two paragraphs
We proceeded by testing link buttons. We started by changing their position on the page. We moved them from the right column over to the left, directly under the navigation menu where the eye naturally looks when scanning a new page. We also tested their color scheme by changing them to red, which made them more distinctive from the remaining background. We also changed their call to action to be shorter and more direct.
In the end, we tested an image of a smiling woman over each link button, encouraging visitors to fill out a form. We used an image of a woman knowing that research done by Ogilvy & Mather proved that consumers best respond to images of smiling people and, trust women most. Our test also proved this rule to be true. In fact, SEO Magma has used this rule across its own website for a prolonged period of time.
We carried out multiple tests simultaneously, because they did not require previous results before initiation. An advantage of even basic website testing applications like "Website Optimizer" is the ability to test multiple variables at once, dramatically cutting down the time required to perform all tests.
In terms of content, it consisted of two paragraphs: top long one, bottom short one. We began our test by reorganizing them. The new top paragraph was shortened to include only the most important message and keywords. We also changed its format to include bullet points.
The bottom paragraph was expanded to explain each point mentioned in the paragraph above. At the text's beginning and end we added two links directing to the contact form. Over the header of the first paragraph we added a new 'call to action' and, at the end of the page we repeated our encouragement to contact. We turned the tests on and waited.
The time it takes to complete a test is conditional. It is based on the amount of visitors that page usually attracts, as well as other factors like seasonality, promotions etc. In this case, each test took us a week to perform over a group of about 1,000 visitors.
After a week's end, we gathered up all the results and began our analysis comparing initial results to the tested ones. Some changes we implemented right away. Others required further tweaking and testing throughout the following week, until we were convinced that they would positively affect results.
Those results
After a few more tests (in total after 3 weeks) we had proof, which changes to which variables would positively affect results. We also learned that aside from other modifications, some small cosmetic changes successfully helped elevate link-buttons' CTR up to almost 28%, while extending time on page to 1 min. and 56 seconds.
Simultaneously, these changes helped raise conversion rates up to almost 16% and increased the amount of potential clients (leads) by 20%. Mission accomplished. However, we believe that if this page had more variables tested, its conversion rate could raise even more.
It is important to note that without running a test, making changes wouldn't make much sense. Something would probably improve, but there wouldn't be a consensus why and in relation to what. Without testing, the risks of implementing bad changes would be significant and unjustified.
Want to learn more about improving your website's position in search engines? Read how to position websites.