Writing For Conversions: Part 3- Testing New Content

Our exploration of utilizing content to increase conversion rates began with analyzing statistics to uncover opportunities for improvement, followed by how to create and implement good content to compel visitors to take action. The natural next step is testing.

Unless you do some sort of controlled, measurable testing, you will never know what changes in copy, titles, calls to action, color schemes, layouts, etc. increase your conversion the best. Simply put, you should test variations of your content elements in different combinations to see what achieves the highest conversion overall.

It may be helpful to picture the structure of journalism, the formula for writing a news story – “who, what, when, where, how and why.” When you are testing new content elements, defining and controlling for the first four W’s (and the H) just might offer you the answer to the last one – “why” your changes get the results they do.

The elements of change

Before we look at the elements individually, you should get a feel for the number of variables involved in this undertaking. First of all, of course, there is the copy itself. You will need to consider not just the individual words in the copy, but its tone, purpose, goal, and of course audience.

If you are a good writer, you can do the writing yourself. If you are not, then get a professional to assist you. Even if you feel you can do it yourself it may profit you more to hire someone else to do it. Not many business owners or employees have an extra 1-2 hours in their day, nevermind 4-6 hours, to write the content needed to succeed. Plus there are large opportunity costs – time spent writing is time stolen from focusing on core business objectives.

There are lot more elements to test than just the copy and more kinds of “copy” (words) than just the main sales message. Besides the main copy you will have title tags, article titles, headlines, subheads, calls to action, etc. Besides the sales message there is informational content to help visitors educate themselves to their decision. A change to any one of these is a separate test.

What you’re testing

Making copy variations might include changing the wording to create more engaging phrasing or language, changing the format from short description to comprehensive information, or changing the tone from hard to soft sell or from technical to layman terms, etc. It depends on your situation and industry. Headline and subhead variations may involve wording, placement, font, even color. With the calls to action, you may be rewriting, or repositioning, tweaking aesthetics, or all the above. If you cannot define what you are testing, you simply cannot test. Determine which elements are relevant to your situation and goals.

After you have defined your points, there is one crucial point to always, always keep in mind: only change and test one thing at a time. If you change more than one thing it is impossible to know which change actually impacted the results.

How now?

You should decide up front whether you are doing open- or closed-end testing, that is, testing that is always ongoing versus testing done for a certain period of time. Although, generally speaking, you always want to be testing and improving, it doesn’t have to be one, long, unending one. You may get better results doing a sequence of short-term tests.

Before you change or test anything you need to establish a control and a baseline. Put together what you feel is a solid page to be your control, then gather statistics on it for a period of time to get a baseline number from which you can judge your progress.

The number of elements and their variations you are testing will help decide which what testing method you use. Let’s take a look at two testing approaches: A/B testing and multivariate testing.

In A/B testing you compare alternate versions of an element against each other to see which one is more effective. If A/B’ing one element you can see which version works best and move forward. If A/B’ing only a few elements with limited variations, combining and testing them against one another is pretty straightforward (e.g. two elements, each having three variations, gives you six possible combinations). If you are testing multiple elements that each have multiple variations, you are going to have a huge number of combinations, so you will have to be very methodical about making one change at a time and tracking the results.

A/B testing is good because you can typically do it in-house with resources you have on hand, and it doesn’t require complicated data analysis, so ease and low expense put it within the reach of most companies. However, testing multiple elements requires diligence and significant time expenditure.

Another way to test multiple elements is multivariate testing. Multivariate testing involves blending multiple variations of multiple elements in various combinations, then collecting and analyzing the data to see which combinations work best. Different elements impact conversions in different ways, so multivariate testing helps you see both which specific elements impact conversions the most and which combinations produce the highest conversion rates overall.

Multivariate testing requires more technical savvy, planning, and upfront work, but can save you time and cover more ground by testing more items in a given testing period. It also gives you a very scientific approach to your conversion improvement and establishes a system for open-ended testing. On the downside, it focuses on one page at a time and doesn’t take into account the pages before or after your test page that may significantly affect conversion rates.

The “who” matters, too

You probably have at least a few demographics, and each demographic contains various subdemographics. Certainly you can test with your visitor population as a whole, but it may benefit you even more to target specific people with specific recipes to see what works best with each particular audience. You can do this using identifying markers, such as IP addresses, referring URLs, unique IDs, etc., or behavioral characteristics, such as frequency of visits, pages viewed, time between return visits, previous purchases, and so forth.

Some visitors you really want to pay attention to are the search engine spiders. When you are doing A/B split testing or multivariate testing you don’t want the search engines to see a different page each time they come by. Be sure your system can identify the different spiders and show them the same consistent page. After you have finalized your changes, remember to channel them to the new page.

The “when” of it

You could do an A/B test of your body copy, with Copy A for a month and Copy B for a month, and accrue and analyze those figures. Or you could set up A/B split testing, whereby Copy A and B are called up on successive page landings. Split testing is efficient because you can run the test in half the time and you lessen the impact of seasonality in the comparison. Whether doing A/B or multivariate testing, you want to make sure you allow enough time to gather sufficient data to draw useful conclusions.

Timing is also important in establishing controls for your testing, particularly when making a succession of changes in A/B testing. Put the best, most creative effort first, which often (but not always) means the new body copy. As you monitor performance over a set amount of time and proceed from change to change, having a regular timing schedule will put you in a better position to tell what change had what effect.

Also, be mindful of any seasonality when it comes to your industry. For example, if you sell flowers, data spikes around Mother’s Day and Valentine’s Day will naturally skew your results, so you have adjust for them or avoid those times. On the other hand, if your business can allow for it, you may want to target those days for specialized testing precisely because the high volumes and focused environment give you excellent data.

The “where” factor

It is not just the makeup of your elements that effects conversion rates, but also their position the page. Where is the copy most effective on the page? Do you place your lead form or products above the copy, below it, in the middle, or somewhere along one of the sides? It all depends on your industry, your situation, your audience, and what you want visitors to do. You have to try different formats and layouts to see what works best.

The “where” factor may affect the call to action more than any other element. Whether it is a button to click, a product to buy, a lead form to fill out, or something else, the placement of the call to action and how many times it appears on the page can impact conversion rates tremendously.

Don’t be shy about putting it in several places on the page. The call to action has both incentive and convenience aspects. You want it to be compelling enough to entice people to take action just by viewing it. But you also want it in several places so that no matter where the visitor is on the page they can take action if your copy compels them to do so.

Another “where” is where the test page exists on your site. The pages that visitors saw before they got to the test page will impact your conversion. If the pages pointing to your testing page have misleading messages, no amount of changes on your test page will give you a clear picture, fix your problem, or give you the optimal conversion rates you want.

Summing up

The wildcard in all this is that your visitors are becoming more sophisticated every day, that is, they are constantly changing. You are working with a moving target. In addition to all the other business reasons, this in itself is reason enough to be continuously testing and improving.

There is an art and science to all testing: an art in composing creatives and changes, a science in analyzing the data, and both in making the adjustments. The fact remains that you are dealing with humans and human emotion – how people perceive your pages (and thus your site) and whether they are compelled to follow your call to action. What visitors do may be tracked statistically, but why they are doing it is based on very human factors.

Despite the qualified, quantified results you will get from testing, never overlook your own instincts and insights. Don’t go horoscope on it; just remember that your mind works in more ways than you know. While you are testing, test your own hunches about copy, placement of page elements, title tags, or calls to action. When you consider results, you must also allow yourself the freedom to make unexpected connections among data points, as well as the leeway to challenge preconceived notions about what the results “might have” or “should have” been.

Gather and analyze the hard numbers and facts, but also include the art and psychological side, because it is emotion that drives people to take action (and thus convert). Remember, we are dealing with human beings, not robots, and people are nothing if not unpredictable, fickle, and confounding. Control your tests as closely as you can, identify and define your elements, run clean and clear tests, review the results, analyze the data, but also use your insight and intuition to figure out how to trip those emotional switches in people. Don’t use statistics only to force human emotional response into tidy rows, neat columns and pre-defined boxes. One of the advantages of “thinking outside the box” is that you won’t paint yourself into any corners.