Using a test-driven approach to ecommerce site design
Written 08.12.05 by: Tom Smith
There is amongst web designers and shoppers, to some degree, a level of consensus about what the experience should be like when buying online. There is a shopping basket (or cart) and "Add To Basket" buttons, there are lists of categorized products and a search engine. There is a checkout process where you either have to register or not and you will get an email sent to you in a few minutes about your purchase. And for the most part... that's it.
Of course, there is a huge range of different versions of these simple elements. Many sites find ways to cock up what should be fairly simple, or innovate in ways that don't really help anyone, or even worse, not realize that the standard model above doesn't work for their product line or customers.
If we had anywhere near all the ecommerce answers, small pure-play online stores wouldn't be stealing a large chunk of business from better known high street brands. If we had anywhere near all of the ecommerce answers, everyone's conversion rates would much higher.
Lots of people know lots of things about the psychology of shopping in general, about shopping online in particular and about interaction design and marketing. But despite all that expertise out there, when it comes to designing an ecommerce site, in the meeting rooms of new media agencies, you will still hear the mantra of "It depends". Sometimes "it depends" is used as an obvious kop out, but at the OTHER media we are proud to acknowledge that we still don't have all the answers. How can we?
Although it seems obvious to say it, selling cream cakes is very different from selling holidays, and not just because one of them is perishable. The back-end day to day business needs of a confectioner are completely different from a travel agent, the people they need to deal with, their customers' expectations, the companies and protocols all require a bespoke ecommerce solution of some sort. Both cream cake and travel ecommerce sites may both have "Add to Basket" buttons but they are in no way similar beyond that.
Of course there are lots of answers and solutions out there, backed up by research, anecdotal evidence and experience, but, as is often the way, lots of our design decisions are still based on personal preference and expectation rather than hard facts.
Which is why the OTHER media decided to build tools for experimentation into our content management system, OTHER objects.
Imagine you have launched an ecommerce site. Now, ask yourself these questions...
- Which promotions made the most money? 2 for 1, 10 per cent off or Free Delivery?
- Which colour shemes sell most products?
- How many links should there be to related products? Is 3 best or 10?
- Are bigger images better than smaller ones at selling products?
... the answer is probably "it depends". It depends on who your customers are, what you are selling and a whole heap of other things. There isn't one answer.
For a while, all most ecommerce sites did was follow the lead of Amazon, with the idea that "If it's good enough for them, it'll do for us", but buying books is very different from buying food, from buying experiences or tickets, for example. And people are beginning to expect an online buying experience more finely attuned to what they are buying and don't put up with poor shopping interface design. You see, the thing is, they really don't have to.
What we've been trying at the OTHER media is making it very easy to create design "experiments". We can split the site into two clear alternatives, say advert-rich and advert-light, automatically serve those two sites to two audiences and measure the popularity of each design against clear metrics, such as time spent on site, or number of sales or average sale order.
Site-splitting in itself isn't unique, I'm sure lots of companies offer it. The part I think is innovative is adding the idea of experimentation into the tools that you use to maintain your site every day. Additionally, all the numbers are related to the bottom line so that at the end of the trial period you can know whether green sells more than yellow or not. All this means that experimentation and continual evolution, design based on knowledge rather than yearly shot-in-the-dark re-designs, where often the most effective elements of your site's design are thrown out with the bath water.
If your content management system also shows you on which step of the checkout process people are bailing out, and which form items caused the most errors, you don't even need to go looking for the least effective parts of your site, they are already there in your face waiting for the copy to be improved or design altered.
Experimenting with design means that rather than basing certain design decisions on personal preference we can defer those design decisions and pass the decision onto the people who matter most... the customers. So whether product descriptions should be verbose and chatty or bulleted lists, or whether the checkout process should be 3 pages or 7, doesn't have to be set in stone, we can simply select the design or approach that works for a clients' particular customers. And because running a test is so easy, we can always check our site later to see if our customers have changed their preference over time.
Of course, you wouldn't want to test everything, unless you base most of your design decisions on hard won experience and reputable external research you will never get anything done. But in cases where, if you are honest, you really don't know, or in cases where you have a sneaky suspicion that something could be improved or made more effective then rather than implement it wholesale, you can "test your hypothesis" there and then (if the tools let you) and let the bottom-line decide if you were right or not.
If you have an ecommerce site you'd like to evolve using experimentation into being more profitable, give us a call now. 020 7089 5959.