Test Everything: An Online Business Manifesto

test everything

We make lots of assumptions when designing a website.

We put a menu at the top. A sidebar to the right of our blog. A call-to-action button somewhere in the mid-to-upper right.

But why do we do all this?

Benjamin Hunt calls it the “best guess” theory of design. You look at what similar websites are doing, and copy them.


The Power of Expectations

To be fair, expectations play a key role in a user’s experience. (Try making a visitor scroll down to find your main menu and see what happens to your site traffic.)

These conventions aren’t necessarily the most effective. But they work reasonably well, so we stick with them.

This still leaves a lot of room for experimentation. Large, successful websites constantly test their layouts and sales copy to squeeze every last dollar out of their traffic. And you can do this too — even as a small business. Read on to learn how.


What is Testing?

First, let’s clear up the definition of “testing.” (If you already know what I mean by this, scroll down to the next section.)

When we talk about “testing,” we’re usually referring to A/B testing, also known as split testing.

Here’s how it works: the business will create 2 versions of a page (sometimes more). These versions are commonly called the control (the original), and the treatment (the variation). These variations can be just about anything:

  • The price
  • The headline
  • The layout
  • The call-to action

Each visitor will randomly see one version of the page and not the other. Typically they won’t even know that multiple versions of the page exist.

Then the business will measure which variation leads to more conversions — whether that means sales, newsletter sign-ups, Facebook shares, or any other goal. This is a statistical process; you continue the test until one variation is statistically proven to convert better.


How To Test Your Site

The benefits of testing should be self-evident. You’ll ultimately get more sales as a result of running tests on your website.

What follows are some free, or low-cost, tools you can use to run A/B tests.

Lead Generation & Mailing Lists

I use LeadPages to grow my mailing list and find new leads. Their Pro Annual package costs $40/month and allows you to create beautiful landing pages and A/B test them. To see an example of a LeadPage I created, click here.

WordPress Blog Testing

Title Experiments is a plug-in that allows you to test different headlines for each blog article.

It’s dead simple to use; you just enter multiple titles when you create a new blog post. It will automatically test these titles for you. It occurs anywhere blog titles are automatically generated — your homepage, sidebar, etc.

As a rule of thumb, you’ll only want to test 2-3 titles at a time for each article. You need substantial traffic in order to draw statistically relevant conclusions; probably at least a few hundred visitors a day.

Manual A/B Testing

Several websites offer this service, but I prefer Convert.com in conjunction with Yoast’s Convert Experiments plug-in.

The service allows you to manually create 2 versions of a page — where you might test difference prices or headlines — and see which converts better. It requires a little technical expertise, so get in touch with me if you’re uncomfortable with this.

Convert.com lets you run a tests on 5,000 visitors/month for free for your first 12 months, and the rates are very reasonable after that.


Be sure to use different SKUs or item numbers on each page variation. Even though you’re selling the same product, you need a way to distinguish which page people bought something from.


Poor Man’s Testing

A true A/B test runs pages in parallel; both versions of the page are live and served randomly. But even if you don’t have the technical skill to do this, you can still look at your analytics (site stats) and get a feel for what works and what doesn’t.

If, for example, you notice that your menu links to a page, but that page never gets traffic, it’s a good indication that you should try something else. Either change what the text says or remove that menu option entirely.

So you don’t necessarily have to run A/B tests to make data-driven decisions. Because that’s what this is all about: using data rather than hunches.


A Potential Pitfall To Avoid

Statistics can be misleading. Sometimes you might be quick to draw conclusions about the effectiveness of a treatment.

Don’t do this unless you’re certain that your conclusion is mathematically valid.

Many of the services I’ve recommended (including LeadPages and Convert.com) do the math for you; they’ll tell you when you’ve gathered enough data to declare a winner in your A/B test.

For most other tests, you can use this online calculator. Just enter the number of visitors to a page and the number who converted (bought something, joined your list, etc.). The calculator will tell you if there’s a winner or if you need more data.

To declare a winner, you usually need a LOT more data than you think. And for every additional treatment you test, that amount of data goes up.

This is why I recommend just 1 control and 1 treatment if you’re a small business. (If you’re Google or Amazon, you can test a lot more at once.)


Test Everything

Stop making assumptions about how people use your site. Use the software and strategies outlined here and you’ll grow your mailing list and your revenue.

These tweaks can increase your conversion rate by 5%, 10%, or sometimes even double or triple your conversions if you strike it rich. This is why you should be fanatical about testing — it’s one of the marketing activities that provides the most leverage for your business.

Also remember that I do this stuff for a living, so get in touch with me if you need help performing a test.


, , , ,

No comments yet.

Leave a Reply