This week I’m writing for the SME audience on Econsultancy, a sweet homage to the much debated yet scarcely implemented topic of website testing, something which should be an integral part of your strategy, the testing plan being the culmination of analysis of web data and voice-of-customer. 

If you are not yet testing and are spending money on marketing/advertising, then read this post and do something about it. Testing is not the preserve of major brands with big budgets, it’s the mental gap you need to cross not the financial one.

Smack in the face with numbers

Amazon tested a form with two fields, two buttons and one link. Couldn’t be simpler could it? The form was preventing customers from buying. After investigation the designers changed one button from ‘register’ to ‘continue’. 

The results: the number of customers purchasing went up by 45%, in 12 months this generated an additional $300m. Yes I said $300m. Don’t believe me, read Jared Spool’s article on User Interface Engineering.

Why should I test?

How does anybody know the best possible page layout and design to cater for the needs of thousands of different visitors? Unless you have a zero bounce rate and 100% click through, your page can be further optimised.

Plus personal opinion and politics always rear their ugly heads. With the best intentions, people often second guess what customers want because they “know the business”.

If your finance bods are tightening the purse strings (after all it’s still not rosy out there in the economy) then give them the basic equation. Take your worst performing page that has high visitor numbers then project the revenue uplift of conversion increases of 10%, 20%, 30% etc. Show them the numbers.

How should I test?

Knowing what to test requires analysis and the understanding of relevance. You might have a page with 100% bounce, should you automatically invest in testing to improve? Not necessarily. What if that page has only two visitors per month? Even with 100% conversion the revenue return is minimal. What if that page is a static store information page with contact phone number? Bounce could be a good thing as you are directing people to the store.

So you need to find relevant pages. Build an engagement matrix for all site pages – look at visits, time on page, bounce rate, click through, conversion, revenue, average order value etc. Then filter with a relevancy formula:

Traffic * Bounce * Conversion * Average order value

Pick the pages that are losing you the most customers/money. Put these at the top of your testing plan.

If you haven’t already got a customer survey running, add it quickly. Surveygizmo is an excellent low cost solution with a Pro license at $49 per month. You can create sophisticated surveys with inbuilt logic and add them to your website as well as emailing your database. There are other tools out there but this has the best recommendations (endorsed by @OptimiseOrDie at the Econsultancy Online Marketing Masterclass event in London). I’ve used it myself – it’s intuitive and easy to publish surveys quickly.

Where do I start and what do I use?

Start with simple A/B testing. Your data analysis has told you which pages are losing you the most business. Your customer survey will highlight some of the potential reasons. Now create some hypotheses for testing.

e.g. customers said they don’t like the registration form on the checkout, it’s too complicated and the bounce rate from this page is 60%

  • Create two new versions and test against the current version:
  • Version one: clearer signposting & help information.
  • Version two: data capture fields in two columns instead of a scrolling page.

You don’t have to spend thousands doing this. Follow these six Simple steps (even I’ve managed this!):

  1. Set-up a Google Website Optimizer account for free.
  2. Generate the testing code yourself.
  3. Give this to your developers to add to the page.
  4. Get your designer to create the test versions of the page.
  5. Tell Optimizer what the URLs are.
  6. Let the testing begin.

As you learn more, you can increase the sophistication of testing and embrace MVT (multivariate testing). A/B tests different versions of a page against each other, MVT tests multiple variations of elements within a page. MVT is ideal for pages with multiple elements such as text links, buttons, images, javascript functions etc.

Reports will show you which page is performing the best. You can then implement the winning page. The testing doesn’t stop there, you can then create a new test for the same page and re-run the fun.


  • Testing is a simple concept – you don’t have to rely on an analytics specialist or agency
  • With simple steps you can do this in-house to prove the business case
  • Testing is an ongoing process – unless you’ve got 100% conversion, there is room for improvement
  • Focus on pages that can deliver the greatest financial uplift – that will keep the finance bods content
  • Use online forums, Google help centre & official Google Website Optimizer blog to get the basic facts and learn from others.

Don’t bite off more than you can chew

Analytics and testing experts are worth their weight in gold. However, many businesses struggle to take the first steps because they can’t afford to pay a specialist, you can implement simple A/B and MVT testing yourself to prove the business case.

Once the business case is established and your Board is behind you, it then often pays to bring in the specialists to manage a more complex MVT program, integrating this with your analytics tools to drive customer insight and commercial value.

Good luck!