What is the problem of search?
by dbasch — No Comments »A post on Techcrunch questions Marissa Mayer’s assertion that “search is 90 percent solved”. There is an interesting discussion in the comments about where search is going, but there is something more fundamental that is not addressed. Let’s leave the 90% progress mark aside for the moment. What does it mean for search to be solved?
In order to solve a problem, we must first know what the problem is. How is “search” a definition of a problem? For many technologies it’s relatively easy to find potential improvements that could be considered problems to solve. As an example, let’s look at cellphones.
It seems obvious that there are many limitations of cellphones that stand in the way of useful applications. Battery life is an obvious one: an ideal cellphone may have a battery that lasts as long as the device itself, pretty much like a wrist watch. Perhaps it could be powered wirelessly. The point is that the user would not have to worry about limiting usage until the next recharge. Also the ideal cellphone would be extremely light, it would have a faster user interface and a better input mechanism, etc. It’s even easier to analyze the limitations of cars: lower prices, self-driving capabilities, better safety features are some issues that come to mind.
Now let’s think about Google’s web search. If we look at the problem of indexing all the text on the web so that a reasonable person can find a given document, how much better could it be? It’s already free for the end user, so “cheaper” is out of the question. It’s also hard to imagine a faster search, as Google returns results within a fraction of a second. In terms of coverage, there is a possibility that Google won’t find a document that is out there but its crawlers have not discovered yet. They have gone a long way in this respect and today it’s hard to make the case for the usefulness of a document that Google has not discovered within 24 hours of its publication. As for relevance, sometimes it can be hard to find a document that Google has indexed even after a reasonable number of tries with different combinations of keywords. I know this has happened to me in the past few months but I can’t remember a specific example (I think it was some product with an extremely generic name). Still, in my experience it’s very rarely an issue. In conclusion, for a certain definition of the problem it’s reasonable to state that it’s 90% solved.
Of course, it’s possible to describe a whole class of problems (many still unsolved) that can be labeled “search”. We mentioned some of them a while ago. Search for extraterrestrial life is an extreme example.
The main point of this post is one that happens all the time in software: someone creates a solution to a perceived problem. This solution uncovers other needs or problems. There comes a point when people lose track of the problems and it becomes very hard to discuss solutions. We can’t argue whether search is solved or not. Let’s understand and describe each specific problem first.
