Wednesday, August 1, 2012

DARWIN'S THEORY OF ECOMMERCE OPTIMIZATION


The two essential elements in ecommerce are an online store and a human being(with some money).  First, let’s consider the human.  The survival of our species depended on our ability to be efficient hunters and gatherers.  Natural selection caused us to evolve an eye-brain connection tuned to rapidly absorb information and act quickly.  The content we see and the order in which we absorb the information affects our behavior.  This is why one web page performs better than another.  Any change to a web page, however small, will impact customer behavior.. one way or another.  If we want to encourage a certain behavior we need to find out how changes impact the desired behavior.  We need to make changes to our website and measure the results.

So let’s consider the web page and focus in on a single element, say a security trust mark, for example.  Most of us would think that a simple AB test would determine if our customers are security conscious and therefore it will show a positive impact..  In reality, there are many questions to answer.  How much impact will it have and which style and in which position will have the most impact.  If you test the same style mark in two different positions you will get two different results.  Likewise, if you test different style or brand of the trust mark you will get different results even in the same position.  I learned this running 100s of trust mark tests for McAfee Secure.  In fact, in a typical test of 3 positions and 3 styles on the same page the resulting micro conversion rate(adding item to cart for example) varied from -10% to + 15% vs baseline without the mark.  So the lesson here is to make sure you introduce sufficient variety when testing even a single element.  Variety creates more opportunities for survival (clicks).  There is one optimal position and style of an element that will have the maximum impact on purchase behavior.  You could set up lots of A/B tests to find the best style and position but the most efficient method of testing with variety is known as multivariate testing. 

Now let’s introduce some complexity.  One page element like one gene may react differently in the presence of another one.  If we take 5 elements and introduce 3 varieties of each it will create 125 different page combinations(versions of the page).  Each combination will impact purchase behavior differently because the content will be absorbed in a different order.  One version of an element may be drawing attention away from another or the presence of one element without the other has less impact.  Given the amount of variety it is highly probable that at least one of those 125 combinations will perform well above baseline.  It is also highly probably that a large percentage of them will perform below baseline.  The task of finding the top performer with this much variety is beyond the capabilities of A/B testing.  You will need to use a multivariate testing tool or service. 

There are two important considerations when selecting a multivariate provider: 

1. The most import consideration is the potential to lose money while testing.  Remember, many of these combinations are going to perform worse than your current site.  If the test splits traffic evenly between combinations you are sure to lose money while testing.  There are several techniques to get around this problem including manual culling elements or combination and the Tagutchi sampling method.  The best solution is a real-time adaptive system that can predict the winners and automatically adjust what is shown while the test is running.
 2. Ease of use.  The best solution offer virtual implementation, meaning no changes on your production server.  There are full and self service options and pricing varies greatly so shop around.  There are two multivariate providers I am familiar with and can recommend:  HiConversion and Maxymiser.

One more note regarding variation and survival.  All people behave differently, the same person behaves differently at different times, and new people are constantly coming to your web store.  You cannot expect a top performing combination that you find while testing to continue to perform the same over time.  There are some techniques and new technologies that take advantage this time varying nature of people and websites. (topic for a future post).

Monday, July 30, 2012

80/20 RULE FOR ECOMMERCE


I have heard digital marketers talk about turning browsers into buyers but i am not so sure this is the right objective.  80% of visitors to an ecommerce site are there for information.  20% will tell you they came with the intent to buy and the industry average conversion rate is 1.8%.  We should do our best to help each type of visitor accomplish what they came to do.  
  
Browsers tend to be patient while shoppers are often in a hurry.  Product pages should be designed with this in mind.  This is an important page for both browsers and buyers.  Focus on the impatient buyer’s needs above the fold (visible area without scrolling).  Keep this area clean and simple with only essential information and a single call to action.  We need to help the 20% be more efficient shoppers and improve their success rate by minimizing distractions and friction.

Keep product descriptions to a minimum or consider not having one above the fold.  Product recommendations are normally best placed lower down the page but this is something that should be tested.  There is a tradeoff between conversion rate and average order value (topic for another post).

Below the fold you are free to provide robust product information, alternate selections,  as well as reviews, social links and more.  Trying to squeeze these elements in above the fold works against the needs of both buyer the browser.  Browsers have varied objectives and interests.  Do your best to meet their expectations with strong content and choices where you have the space to provide it.  They will find it and reward you by coming back when they are ready to buy.