The Internet moves at blazing fast speed. And, according to Avinash Kaushik, this allows businesses to rapidly test new products, new tactics, and new strategies so they can “fail fast” meaning businesses get an answer quickly about whether an idea is good, rather than shooting down great ideas in a smoke-filled conference room (ok, so they’re not smoke-filled anymore).
Now, firms can use analytics and online metrics to test, tweak, and move fast to capitalize on market opportunities.
Observe — Hypothesis — Test — Analyze and Adapt
Using the scientific method adapted from the hard sciences, helps mitigates the risk of bold ideas. Today, I attended a presentation by Adam Ribaudo, Director of User Experience, Velir. Below, I summarize how he recommends using analytics to improve business performance.
But, you have to start by observing what’s going on inside and outside the company. I usually recommend doing an environmental scan ending in a SWOT analysis to understand metrics inside the company, among competitors, and in the environment where the business operates — the cultural, legal, economic, and technical elements.
Firms then establish clear goals for performance. When performance doesn’t match expectations, firms need to develop hypotheses to explain why. Of course, businesses need to ensure the cost of doing this research is worth the payback they’ll get from better information.
Firms also use analytics, observations, and the scientific method to find opportunities such as new products.
Samples of hypotheses
Changing the call to action will improve conversion
Changing headlines or product descriptions will improve conversion
Reducing the number of clicks will improve conversion
So what kinds of results would you expect from using analytics? That depends on your goals — obviously your analytics should match elements that comprise your goals.
Remember our Hierarchy of Effects presented yesterday:
increasing positive sentiment
Increasing # of brand mentions
decreasing bounce rate
increase return visitors
increasing depth of visit
a sense of community
So, what types of analytics should you look for?
Looking at raw analytics matching the results you want to drive helps. But, testing provides deeper understanding about how consumers interact with your website and offer solutions for improving your metrics.
What types of testing work?
A/b testing – basically developing 2 or more alternatives for key elements on your website — such as offers or headlines
Multivariate testing – allows you to test multiple aspects of your website at the same time.
Usability testing – mostly using groups to talk through their experiences with your website.
Controlled experiment – this technique uses matched control and experimental groups to provide web analytics close to those you would find when the website is deployed.
What tools are available to conduct analytical tests?
Google Website optimizer — which has the advantage of being free
Visual website optimizer — $49/month
Omniture – from Adobe (enterprise level)
Sitecore – has powerful web testing as part of their CMS
A Case Study
To show how effective, and valuable, these techniques are, let’s take a look at a real live case from Amazon.
Amazon questioned the high rate of shopping cart abandonment on their website. Using focus groups, researchers found the biggest drop-off occurred when shoppers were asked to login to complete their transaction. Customers didn’t want to provide personal information — they just wanted to buy the products.
Instead, Amazon offered customers the option to buy products without registering first. They still needed to enter information to complete the transaction, but not having to register first increased sales by $300,000,000 — $300 MILLION. WOW, for such a small change that didn’t impact the operation, what a big payday.
Add option to buy without registering – results up 300 m in 1 year.