This summer I took a couple weeks to really ponder what I saw developing on the real-time web. One of the conclusions I drew was that the success of real-time search would be based on a) context and b) identity.
Google’s social search is the first compelling foray I have seen into nailing the “identity” component of this equation. It’s local, as related to networks — not geography.
In this post however, I want to share my thinking around “context” in the hope that one of the real-time search engines will begin to embrace this, or get some smart people thinking about it.
The most simple example to show the “brokeness” of real-time search is to run the query “jobs + starbucks” in Summize (Twitter Search). The results you get will have two issues. First, results will likely have nothing to do with jobs in the sense of “available positions at Starbucks.” Second, many of the results probably look sketchy to the point where you don’t know if clicking will lead to a job description or a porn site. Oh, bots!
Job search is a great example of intent based search — when people search for the keyword jobs, more often than not, they are looking for an open position being hiring for.
My theory is that there is a much better way to distinguish between searches for jobs (the keyword) and jobs (the open position). It all comes down to the hyperlink.
Hyperlinks are company names, and company names are synonymous with context. For example, if I search for “jobs + starbucks,” rather than searching the real-time firehose for any mention of the keyword job, the search engine first query against hyperlinks from trusted sites –maybe even short URLs, optimized for this type of query — think: monster.com, theladders.com and doostang.com, etc. Using this type of search (which I do often) immediately reduces the number of false positives.
If my system was embraced, I also believe an entirely new form of commerce could emerge: what I think of as the disaggregated marketplace. It’s the marketplace that works for the “web of flow”, vs. the “web of static, destination sites”. It’s a combination of specific transactional objects (jobs, classifieds and promotions) coupled with the ability to search against and leverage the social graph (as opposed to simply keyword and geography). In my view it’s a hugely important concept.
On a separate, but related thread, I still find it shocking that Twitter doesn’t let me search against only people I choose to follow. This is where the identity portion should come in, but that’s for another post.
Note/Hack: Summize can actually be searched for hyperlinks using the advanced search tool: “show only” and http://domainname
