AT&T: Solr as middleware
I’ve been doing a lot of talking about how search is about more than finding data; it’s about understanding it. And it might almost be conventional wisdom to suggest that application of search to the big piles of data and content boils down to ‘know thy data’. But add the other vital piece of software insight — ‘know thy user’ — and it gets interesting. Understand that search is not just about queries and indexes — it’s about matching particular users to particular data.
For AT&T’s mobile e-commerce online store (selling ringtones and bluetooth headsets), Shantanu Deo explains in his talk on Day 2 of Lucene Revolution, the decision to use search to generate their web navigation was a necessity. They needed a way to provide three levels of navigation, including over 700 unique URLs, customized for each user, based on membership in more than 60 profile groups. For example, you might be a New Yorker who is on the Family Plan and has Digital TV.
View the slides for this session:
The only reasonable way to do this, Deo and his team decided, was to model it as a search problem. They encoded each level in the hierarchy with a six digit number (2 digits per level) and indexed documents specifying which URLs should be shown on which users groupings in which levels. From there, it was a straightforward matter of showing users URLs they should see in search results, and none of the ones they shouldn’t.
The system works for AT&T; once they’ve indexed the data, the system just plain runs unless they have changes, and if they do, there’s no coding involved. All they have to do is re-index the data. No need to rearchitect the application or change the data model
The principle has the beauty of simplicity. Personally, I think this is an interesting harbinger of more search-based navigation. In this case, they’ve got a fairly specific (if large) set of data to manage, and Solr handles it easily. In the future, they could easily add the ability to display links based on more complex search factors, such as past history or interests.
You can also hear Shantanu Deo talk to Michael Cote of Redmonk, in a recent podcast, here.
Cross-posted with Lucene Revolution Blog. Nicholas Chase is a guest blogger.This is one of a series of presentation summaries from the conference.
Contact us today to learn how Lucidworks can help your team create powerful search and discovery applications for your customers and employees.