Wednesday, October 21, 2009

6 Examples of Data-Informed Site Changes that Increased Conversions

You're not one of those designers still resisting the use of data to inform site changes, are you?

If you are, I'll let you in on a little secret: I was too just a few short years ago. I would make prolific statements, like: "Who needs data? I'm a designer, dammit! I craft websites based on my intuition, design aesthetic and vast experience." Sound familiar?

In hindsight, I was not taking the time to understand site visitor behaviors, friction points they encountered or frustrations they harbored. I was far from being an advocate for the user. As a matter of fact, I didn't have a solid understanding of what was working and what wasn't, and certainly not 'why'. This resulted in a lopsided balance between user needs and business goals. The business almost always won.

Embracing Data
Fast forward to today: I'm growing my agency's client relationships and winning new business by evangelizing data-informed design. Of course, we don't let user data drive all of our design decisions. That could translate into emotionless, marketing-driven garble. But there is a happy place where a quantitative and qualitative approaches live harmoniously together! I've been there and have seen how it can result in delighted clients, satisfied users and an invigorated design team.

Success Stories
Let me share some quick examples of how we got to that happy place. Below are 6 challenges I faced with a few select clients. For each, I've listed the evidence we found within the analytics + survey data and the design changes that helped achieve dramatic increases in conversion:

Client Question #1: Are our visitors sufficiently motivated to start the checkout process?
Evidence: 10% of visitors abandoned before proceeding to checkout, although self-qualifying themselves as "ready to buy."
Result: Set expectations around checkout process length, as well as improving the visibility of shipping and return policy.

Client Question #2: Why is there a high abandonment rate so late in the checkout process?
Evidence: Nearly 20% of visitors never clicked "Checkout," but did click "Update Cart," possibly mistaking this button as a means of continuing.
Result: Designed clearer visual distinctions between calls-to-action.

Client Question #3: Within checkout, are visitors confused by our process?
Evidence: Multiple page views of billing and shipping pages, with a high rate of visitors clicking "Help" within them, then abandoning shortly thereafter.
Result: Improved error messaging and contextual instructional copy.

Client Question #4: Are we appropriately allocating page real estate to our features, such as gift card redemption?
Evidence: 90% of visitors never attempted to redeem a gift card during their session.
Result: Removed gift card feature and re-allocated page real estate to revenue-generating features.

Client Question #5: Why are our visitors making only single-item purchases?
Evidence: Only 30% of visitors scrolled down far enough on product pages to see cross-sell functionality. For those that did scroll to see the cross-sell functionality, less than 50% of them interacted with it.
Result: Moved the cross-selling functionality along the right rail and added explanation of why the products are "recommended."

Client Question #6: What are the optimal number of pages for our application process?
 Is our current long 1-page design optimal?
Evidence: A/B test showed that over 70% of visitors to the 1-page control application scrolled and then abandoned, without any other type of page interaction. The 2-page design showed higher interactivity and 10% higher conversions.
Result: Put 2-page application process in market and conducted further multivariate tests to lift conversions another 8%.

Summary
Of course I have to write the disclaimer: "your results may vary." We apply best practices to every site we build. But more often than not, we identify a multitude of distinct user behaviors that are specific to each of our clients. Hence the embracing of an iterative and data-informed design process.

(Note: this post was originally intended to run on Carsonified's Think Vitamin blog in the late summer. But since they procrastinated, I decided to post it here on my own blog.)