There are not many retailers that have not heard of the long tail (if not go check out Chris Andersen's book). The final breakout session of the day featured Jack Jia, CEO, Baynote and Brian Elliott, President and CEO, Alibris discussing how the long tail applies to online sales.
Jack Jia from Baynote was first to speak. Jack defined the major problem as being poor search and navigation, 83% of leads to a site will abandon in the process. Sure you're driving people to your site, but are they conversions? In measuring loss, they found that 95% of visitors will give a site 3 clicks before abandoning. Compare this to most commerce sites which require 6 clicks to complete a sale (from inbound link to sale).
Enter the long-tail dream. Is it better to sell many of 1 or 1 or many? In Jack's words it's tail or torso. In order to benefit from the long tail you need to expand your inventory. Here's the problem, when you increase you inventory, you also increase you chances of users getting lost.
One way to help users navigate sites is the wisdom of crowds. The other concept is emergent behavior. Rather than relying upon what somebody 'tells' you what they do, watch what they actually do. Track usage patterns and learn from them. Recommendation technology is based upon these tools.
Jack showed an example using US Appliance, how they make recommendations based upon browsing habits of other users. Community involvement also helps bring relevance to keyword searches.
A side benefit of the long tail is that long tail products have a higher margin.
Jack then sat down to talk with Brian Elliott of Alibris. Jack mentioned that Brian and Alibris stumbled into the long tail way before it was fashionable. Alibris currently sells 75 million SKUs, 15x what Amazon sells.
Brian talked a bit about Barry Schwartz's book The Paradox of Choice. People are sometimes are paralyzed by choice. How do you balance offering more (long tail) without scaring it off (paradox of choice).
At Alibris they do have a number of options for single items, but they allow other customers to rank/rate and this helps other customers. They're also constantly refining their site based upon customer testing.
Alibris built their own recommendation engine. The looked to their vendors for data. Many of their vendors have a wealth of information about their products. The challenge is to hook into that data. The second factor the look for is demand. A high demand product doesn't always make a good recommendation.
Jack asked Brian about some of their SEO/SEM tactics. Brian stressed that the keyword strategies need to appeal to the machine (i.e. the spiders) as well as the consumer.
So much of the recommendation model looks at user behavior around purchases, but what about the non-purchaser? How can you take the data from those sessions and do something with it? That's the next step.
