I’ve heard statements like this countless times: “We should use Add to Cart instead of Buy Now because it’s the best practice.” But the fact is that in web design, best practices simply don’t exist. What works for one site, might not work for another. It will depend entirely on the context in which the element in question is displayed, and the website’s audience. Add to Shopping Cart might convert better for Amazon, whereas Buy Now might convert better for Dell, and Buy It Now might convert better for eBay. They have the data to back it up. But just because the Amazon uses Add to Cart, it doesn’t mean that’s the right option for your site. You will never know which option is truly the better one until you’ve identified and tested other options. On the web this is relatively inexpensive, not to mention a lot of fun, so you have no excuse!

These “best practices” aren’t just limited to calls to action, although CTA’s are notorious for having so-called best practices. Other elements which are subject to the tyranny of best practices are: colours, fonts, size of icons/design elements, advertising, navigation, general layout, style of writing, type of visuals used, design style, and everything else under the sun.

So, how do you determine what’s best for your site? Here are some paths to enlightenment…

Usability Testing

Usability concerns should be the first issues to be addressed – think of it as the low-hanging fruit of conversion optimization. If any element of the site is inhibiting visitors from converting, it should become apparent through usability testing. For example, during the design phase of a relaunch, you can present two versions of mockups to the subject of a usability test: one that uses Add to Cart, and one that uses Place in Shopping Bag, and see if the subject can successful complete the task in either scenario with as little friction as possible. This is obviously an overly-simplified example, but if the subject is confused by any of the variations, you’ll know to completely eliminate the potential for using it. If both are easily understood, it’s up to multivariate testing to determine which one actually converts better.

Multivariate Testing & A/B Testing

Once you’ve addressed all usability concerns, you can start testing different hypotheses through multivariate testing. The key to multivariate testing is to start with a hypothesis – one that is well-thought out. Don’t be afraid to test radical changes – we’re going for big wins here! With multivariate and A/B testing, you can say with confidence that “our tests conclude that on our website, Add to Cart converted 20% more orders than Buy Now, so we should use Add to Cart.” But that doesn’t mean that Add to Cart is a universal best practice because what works for you, might not work for another site. Furthermore, you don’t know if Add to Cart is actually the best converting CTA for your own website – all you know is that it’s better than Buy Now. It’s impossible to ever know what’s “best” because you cannot possibly test every scenario.

Multivariate & A/B testing can be done very easily for free using Google Website Optimizer, however there are several situations where you’ll run into trouble with multivariate testing & A/B testing, so if multivariate & A/B testing isn’t an option for conducting your test, sequential testing is the next best thing, as long as you understand the limitations.

Sequential Testing

Sequential testing attempts to measure the impact of a web design change by comparing the your analytics data for the period before the change to the period after the change. For example, your web analytics tool tells you that conversion rate before June 1 was 2% when you use the Buy Now CTA. On June 1, you change that CTA to Add to Cart. After enough data has been collected, you see that after June 1, your conversion rate is 3%. While it does have its obvious limitations, it can be used instead of multivariate testing so long as you’re careful and follow these guidelines:

• You select time periods where the amount of traffic is relatively the same
• You select time periods where your traffic mix is relatively the same
• Rather than comparing all traffic in aggregate, segment your traffic as much as possible and compare within segments
• Do not introduce any new changes, tools, features, irregular content, or anything else that could potentially influence the outcome you’re measuring
• You gather enough data to have statistically significant results (use the Google Website Optimizer calculator)
• You only conclude the design change has had an effect if you notice a big impact

It’s important to keep in mind that Sequential testing is suboptimal for a few reasons:

• Your traffic sources are naturally going to change from day to day, week to week, and month to month, so it will likely be different audiences seeing each variation
• You must observe a large improvement to conclude there was a statistically significant effect
• You can only test a small number of variations (depending on your traffic level)
• For big websites, it’s almost impossible to leave the site untouched for an extended period of time, so there could be other content/design changes that are influencing the test

But if this is the only method available to you, and you are careful with implementation, it can provide some insight – it’s better than not testing at all.

Surveys

Analyzing clickstream data can show you what is happening on your site, but qualitative data such as surveys will help you answer the “why”. Surveys can answer questions that your data cannot, giving you further insight into exactly what works and what doesn’t on your site. Survey answers can give you great ideas for testing. For example, you could ask your paying customers what factors affected their purchase decision. If there are any surprises, it would certainly be an opportunity for testing.

Continual Testing

Audiences change over time. What works for your site today might not work for your site next year. It’s important to keep testing new and old hypotheses to make sure you’re always using the most effective tactics for your current audience. Amazon is famous for this. At any given time, they’re running several multivariate tests, which is one of the reasons they’ve been able to consistently generate unheard-of conversion rates of over 15%. Amazon has embraced the fact that best practices don’t exist in web design, and so should we all.