Tell HN: Why Bing is the future of search.

[amichail]

Google-style search is essentially a solved problem. I wouldn't be surprised if Google's rankings are already close to "optimal" for the top 10 results.

This would mean that it would be impossible to ever beat Google because small ranking improvements would be insufficient to convince users to switch in large numbers.

The only way to improve search is using verticals as Bing is doing.

But you need to take this much further, opening up a vertical search API so that developers can extend the search engine in thousands of verticals competing with each other whereever there is overlap.

[pg]

I suspect there are less incremental ways to beat Google. In this field, as in many, the big advances will come from redefining the problem: from saying not "we could make search better if we did x" but "if we did x, people wouldn't need to do search."

Directed Edge is already looking at the world that way. I'm not claiming they're the startup that will make Google irrelevant. Historical precedents suggest it's still early for that. But it seems extremely unlikely that nothing ever will.

[shorbaji → pg]

Well said.

Microsoft dethroned IBM with "if we did PCs, [most] people wouldn't need to do mainframes".

Ditto Google to Microsoft with "if we did search & cloud, [most] people wouldn't need to do OS".

For someone to "beat" Google the the search bar needs to become irrelevant.

[bsd_junkie → pg]

Actually, any service out there that tries to do away with the requirement that people actively perform some effort would fit the bill for a potential Google killer in a particular vertical. In other words, any recommendation service that preferably can "learn" the habits of its users and then simply recommend new content. DirectedEdge is just the latest player in this game. I personally track the news recommendation niche. There used to be a service called Findory that in my opinion came very close to being "optimal". But they went under for some reason. The closest to Findory (and with some very cool features such as computational lie detector for news) is http://www.euraeka.com They have all three basic features - search (aka Google-style), discovery (aka crowd-wisdom style) and recommendation engine.

[amichail → bsd_junkie]

Recommendation engines tend to fail, with StumbleUpon being the exception.

[wheels → amichail]

I'm not sure that's a meaningful statement in the startup world, since I think it can be reduced to, "Startups tend to fail, except for the ones that don't."

If the incidence of such is higher within the recommendations space, it's by less than a standard deviation.

[anamax → pg]

Recommendation engines aren't the only way to change search.

Most people spend a large fraction of their online day in a couple of applications. Those applications could easily auto-generate useful search links - this kills the search box. It also solves some very hard search problems. ("dolphin" is ambiguous in a search box but it's unambiguous in most documents.)

The next step is to make the search results page more useful. In the context of a document, this "more useful results page" is likely to not have ads. Instead, the results will go into the document in a useful form.

FWIW, Microsoft does a better job monetizing the best candidates for this kind of search integration.

[eserorg → pg]

Starting a search experience by asking the user what they are looking for misses a lot of important use cases.

For instance, checkout some of the demos of Marti Hearst's research into faceted search at Berkeley: http://flamenco.berkeley.edu/demos.html

[tallanvor]

The danger of verticals is that people don't want to have to think about what site to visit depending on what they're looking for. Sure, it would improve relevance, but I think you'd find that the increased relevance isn't worth the inconvenience to the majority.

That said, what people forget is that there are still many areas where Google is not the king of search, especially at the enterprise level, so even with Microsoft, Endeca, Autonomy, and other search companies, there's still room for competition.

[amichail → tallanvor]

The danger of verticals is that people don't want to have to think about what site to visit depending on what they're looking for.

These verticals should be integrated into one site and vertical selection should be automatic (but taking into account implicit and/or explicit feedback from users).