eBusiness Secrets is a series of in-depth articles about Internet mediated Business and Commerce, what makes a great eBusiness and how do the best ensure they keep thriving online.
The credit crunch is pushing shoppers online on the premise of wider reach and deeper savings. As you read this, people who may be new to online shopping are hitting your site, wallet in hand. How do you make sure it is you they end up ordering from? And how do you prompt higher basket values and more frequent purchases?
You could do worse than trying my latest coinage: Automated persuasion. Traditional persuasion architecture techniques address site effectiveness issues by setting commercial goals and applying user centered design techniques to the creation of a commercially effective user experience. In essence, you tell users what the site does and ask them how they’d like it to do it. If applied thoroughly, these techniques are sure to increase your site’s ability to sell as well as the value of each sale. However, applying these techniques thoroughly requires big commitments in terms of time and resources and any change introduced as a result of applying persuasion techniques is liable to input bias (the personal opinions of the users involved) and erosion (the change in attitudes by users over time as a result of long term market development and abrupt external discontinuities such as the current credit crunch).
You can, therefore, invest a lot of money and time in studying a few users and risk timing it wrong, or you can find a way to protect yourself. Input bias can be removed if all of your visitors contribute to the study. Erosion can be eliminated if the study never ends. So, you need a never ending study with universal reach. Impossible? Prohibitively expensive? Actually, much the opposite, thanks to the latest multivariate testing platforms.
Step 1: Collect information. You could let the web evolve: seed a large number of variations, run them all and let the best one win. Or you could collect data about the parts of the site/page/funnel actually contributing to your objectives, and in what way, and target your experiment to those. Collect your figures by running a few static combinations in parallel and comparing results against a large portion of control visitors. You’re not really looking for detail, just an indication of which elements have a high chance of materially affecting your results.
Step 2: Explore. Create as many variations as sensible with the elements identified in the previous step. Don’t test variations that break the brand, the design or just plain look wrong – not in the first few iterations anyway. As the number of variations increases you’ll want to move away from static measurement to a dynamic measurement platform. The free Google Site Optimizer or a similar tool will greatly help accelerate your results in this phase. Leave every combination of variations running for a day or two, then remove the losers and introduce further variations of the winners, and new designs, into the pool. You’re not looking to come up with the definitive design, just to understand how elements on the page tend to relate to each other.
Step 3: Automate. Use business rules derived from the previous step to feed a conversion optimisation engine (i.e. Optimost). The system will start from a set of well performing designs and learn how different groups of customers react to each, using your business rule to develop your designs over time in order to maximise the KPIs you set it. This is not a time-limited test: this system will be running your site all the time, for all your visits.
Conversion optimisation is no magic wand, and the tools still have to mature to a point where their use doesn’t conflict with a site’s accessibility or SEO requirements. Additionally, it is crucial to start from an informed (Step 1) and effective (Step 2) position. No amount of testing is going to make a bad site work well. But it can make wonders to squeeze the last drops of value out of every single visitor hitting a well-designed site. In a marketplace where often the winners take all, this slight edge may be all that stands between you and world domination.
I’ve heard some good things about this blog. Remember to balance the pics with the text tho. cheers!
Martin Puttock says
Hi – where do you stand on using on line chat to boost sales conversion levels?. We have found that by using real time scoring based on predictive analytics we can stimulate significant uplift. The trick is to invite only the hottest prospects to chat.
Roberto Hortal says
I think there may be limited opportunity for chat but unlike you I believe the opportunity is with the clueless, rather than hottest prospects. Hot prospects already on our site’s transactional pages should not need online chat’s help to convert. Their adoption of chat in that context may be masking a design problem with your site that prevents them from naturally converting unaided.
Clueless people, on the other hand, may find the option of talking to someone without being forced to change context to i.e. the phone a good option that would help them gain information and ultimately confidence about the transaction. These can also be helped in other ways via user-centered design methodologies and persuasion architecture but online chat may provide a valid shortcut to the same effect for them.
Finally, a word of caution about using chat as a tool to convert undecided customers. Don’t! Figures may show a short-terms heightened conversion after adopting pushy chat requests to people on a timer, but fixing the underlying design issues with a site that chat patches up will give much better, long lasting results.