Split Testing: Where the Hell Do You Start?

by Ryan Miller
on August 18, 2015

TL;DR

If you’d like to dabble in split testing but don’t know WHAT exactly to test, start with the basics. What would you like to see improved? What about your boss/the client/your business partners?

As a marketer/writer/entrepreneur/business strategist, you’ve probably heard of split testing. And if you haven’t, here’s the skinny: you’re wasting valuable clicks and your landing pages probably aren’t converting.

Actually, your landing pages may not be converting because you don’t know how to split test effectively. If you’re like most people, you may have heard about this novel concept from a blog or webinar. You probably thought to yourself, “gee, why am I not optimizing every click to maximize my ROI?”

And then, when the time comes to actually conduct the test and craft the perfect landing page, you find yourself confounded by every little detail on the page. And you can’t help but wonder what the heck you’ve gotten yourself into and where you should even begin.

A Brief Primer on Split Testing

I’m not ashamed to say I’ve been in those shoes. I think we all have. But I’ve come up with a few starting points that can universally apply to just about any ad campaign.

First, however, let’s briefly go over what split testing is. In the abstract, split testing is a series of randomized experiments conducted on landing pages. The purpose of these experiments is to ascertain which kind of page produces the most favorable results (purchases, signups, leads, etc.).

There are a few different kinds of tests that can be run. There is the A/B test, which takes two completely different landing pages and compares them side-by-side. There is also the multivariate test, which involves changing one or more elements of the same core design, such as a slider or CTA button.

Regardless of the kind of test you want to run, you’ll likely find yourself lost in the Bermuda triangle of copywriting and web design. There are millions of tutorials that will tell you how to conduct your split test, which variables to measure, and how to gauge its statistical significance. But few — at least out of the ones I’ve read, and I do read a lot — go to the trouble of telling you where the hell to begin.

Trouble Running a Split Test
Don’t be this guy, questioning your whole self-worth just because you can’t make heads from tails with this damned split testing thing!

The Gift & the Curse of Perfectionism

To start off, let’s unabashedly admit that web design and copywriting are never really finished. Here at TM34 Marketing, we like to say that we can’t do shitty work. And it’s true. We’re all perfectionists in our respective disciplines, and we get extremely meticulous about what we do.

Oftentimes, this is a liability more than an asset. If we have a project that’s due at a certain date and time, we can’t afford to revise our work one thousand times before it’s finished. We will give it 110% effort, but there’s always going to be that one line of copy that we’re second guessing. There’s always that voice in the back of our heads telling us we should have used a different font.

The good news is that this kind of perfectionism provides the perfect starting point for a split test. If you honestly think that something could use a better design or some better copy, then by all means have at it. Run a test on the elements in question and let the numbers speak for themselves.

The “Two Quarterback” Dilemma

Arm wrestling over the best split test techniques
There can only be one!

I’m not sure if this is an actual maxim, but my mother used to say “you can’t have two quarterbacks.” Any organization needs uniform leadership, and if you need any proof of that fact, ask yourself why Roger Waters ever left Pink Floyd.

I’ve had plenty of disputes with clients, clients’ employees, and even my former employers over which design elements are most appropriate for the task at hand. I considered myself the marketing expert, and so did the other individual. With two quarterbacks and no room for compromise, no work got done.

The ego is a tough obstacle to overcome, but nothing is more convincing than cold, hard data. Running a split test can deliver a solid verdict on who’s right in the situation. So tell your boss that, if he really wants that spammy popup to be the first thing visitors see on his website, he should at least see what his conversion funnel would look like without it.

To Be…Or Not to Be Above the Fold

“Above the fold” is another great place to start. As I’ve stated previously, the eternal quest for conversion-driven web design largely revolves around prioritizing above the fold content. You want the most crucial parts of your landing page to be the first things your visitors see. This sounds simple enough, but in reality it’s extremely hard to do.

In their article “How to Make a Landing Page That C.O.N.V.E.R.T.S.,” Kissmetrics lists at least 8 qualities that an optimized landing page should exhibit:

  • Clear Call to Action
  • Offer
  • Narrow Focus
  • “Very Important Attributes” (i.e. benefits and pain points)
  • Effective Headline
  • Resolution-Savvy Layout
  • Tidy Visuals
  • Social Proof

You don’t need to be well versed in landing page best practices to assume that the CTA and the offer should be prioritized above all else. As for the remaining ones (headline, testimonials, pictures, benefits, pain points, features, contact form, etc.)…good luck fitting it all above the fold.

There are a lot of interesting, creative things you can do with a landing page. But packing it all into a responsive layout with generous whitespace is impossible. It’s that simple.

What isn’t that simple is figuring out which genius ideas to place above the fold. But, depending on how crazy you want to get, you can test dozens of landing pages with different visible elements and see how your visitors react.

When the results are in, you can stop being so neurotic about the design. I don’t care how hard you worked on that perfectly designed infographic. If it’s producing 0.05% the conversions of your other pages, ditch it for something that works.

Ask Yourself Whether Split Testing Is Even Necessary

Landing pages used for split testing
Did you ever stop to think that maybe these landing pages aren’t the problem? Just sayin’.

Okay, so this one might be a bit of a cop-out. But seriously, it never hurts to wonder whether you’re applying the right solution to the wrong problem. Remember that split testing is concerned with landing pages and landing pages only. This is a huge component of any conversion funnel, but it is just one component, nonetheless.

If, for example, you don’t actually have a landing page, then you have much higher priorities than split testing: namely, creating a dedicated landing page. And no, your homepage is not your landing page (in most cases, at least). I’ve even seen ads that land on Google+ profiles — don’t get me started on these kinds of amateur-hour shenanigans.

But before you get too ahead of yourself, remember that you first have to actually create and place your ad on the medium of your choosing. There are a variety of platforms and different ways to go about this, including:

  • Search engine advertising, i.e. “pay-per-click
  • Advertising on social media
  • Placing ads on 3rd party websites

Some ads are text-only, while others emphasize imagery. Some that allow for graphic creation also moving or “fluid” ads. There are different limits on the images you can use, the text you can place, and who your ad can target.

Thus, there are a variety of factors to consider before your audience even gets to your landing page. Here are a few of them:

  • Are you targeting the right audience?
  • Is your cost of acquisition best measured by clicks or impressions?
  • Are graphics necessary or distracting?
  • Could the ad copy be written better?
  • If you’re targeting by keywords, are you bidding on the right ones?

There are also tests you can can run on individual ads, and of course analytics software you can use to measure, evaluate and compare your results. But, alas, that’s a different kind of test that requires an altogether different discussion.

Like This Article? You'll Love Our Work.