Author Topic: Alternatives to 99% algo-based search - how would you rank search results?  (Read 5614 times)

inbound

  • Jr. Member
  • **
  • Posts: 53
    • View Profile
There seems to be a fair amount of discontent building about the quality of search results on Google; so, given the huge task that general search is, what alternatives do you see as worth exploring for any competitor to Google?

I'd take an approach that isn't as commercial as Google, by lowering the commercial expectations it frees you to make choices that favour the results rather than the bottom line. An extreme way to do this would be to drop advertising completely - the best set of results for any query should surely not have adverts in it.

So a question from me, how would you monetise a search engine that did not show any adverts?

On the ranking side, it must surely mean that sites are whitelisted and blacklisted based on some form of human judgement in addition to automatic factors - how would you do te manual scoring? (big potential problems there)

Of course, if you move away from covering all searches you can adopt different ideas more easily. An idea I'm toying with for verticals I run is to show everyone, and automatically collect data about each company from their website (if they have one) but should a company wish to control what shows on their listing, they would have to pay a small fee (by ticking what services they do and don't offer as well as places they cover). I'm thinking that the fee should be trivial and a yearly "subscription" which does not alter their position for searches.


Drastic

  • Need a bigger hammer...
  • Global Moderator
  • Hero Member
  • *****
  • Posts: 3087
  • Resident Redneck
    • View Profile
Re: Alternatives to 99% algo-based search - how would you rank search results?
« Reply #1 on: February 21, 2011, 04:48:30 PM »
>So a question from me, how would you monetise a search engine that did not show any adverts?

Well, the market share would have to be there first obviously, but spidering/indexing per domain would be a possibility.


BoL

  • Inner Core
  • Hero Member
  • *
  • Posts: 1214
    • View Profile
Re: Alternatives to 99% algo-based search - how would you rank search results?
« Reply #2 on: February 21, 2011, 06:07:22 PM »
Free 'organic' search & paid inclusion/performance for commercial searches.

Google already has a pretty good idea what's an informational search and a purchasing search.

4Eyes

  • Hero Member
  • *****
  • Posts: 817
    • View Profile
    • Email
Re: Alternatives to 99% algo-based search - how would you rank search results?
« Reply #3 on: February 21, 2011, 08:49:58 PM »
Quote
how would you monetise a search engine?
... that bit is easy, the key question is how would you persuade people to use it 'post monetization'

Personally, I would settle for complete fraud - build a really good 'anonymous' website for each big-money niche and rank myself top - but it does rather depend on having a good solution to the 'key question' first.

ergophobe

  • Inner Core
  • Hero Member
  • *
  • Posts: 9394
    • View Profile
Re: Alternatives to 99% algo-based search - how would you rank search results?
« Reply #4 on: February 21, 2011, 09:08:34 PM »
[Gary, I know you're probably not interested in anything with a paid model, but these two things you say make me want to at least throw out the possibility]

Quote
So a question from me, how would you monetise a search engine that did not show any adverts?
Quote
if you move away from covering all searches you can adopt different ideas more easily.

Well, there's always the Lexis-Nexus model: make people pay. Obviously, your results need to offer something that other results don't. The key is, if someone's primary way of making money is by accessing things like your data, they will pay to have really good search on that data set.

I use all kinds of search tools that are subscription-only, but typically what they have built is not a better search algo, but
 - a database that nobody else has
 - a search algo that is customized for their database.

Attempting to "organize the world's knowlege, but better" is going to require something that scales massively, dramatically improves on Google and can be monetized by scraping some cream off the top.

Meanwhile, attempting to organize one small piece of the world's knowledge, but really really well is something people will actually pay for if it's a tool from which *they* can make money. How many people here, for example, pay to use a keyword research tool or competitive analysis tool?

So I would focus on trying to partner with someone who has incredible data and crappy search. The trick is finding those last untapped sources.

I think you'd be surprised what sorts of data various professionals consider essential to doing their job to the point that they will pay for search access. Almost every scholar of French literature or history pays directly or usually indirectly (through institutional subscription)  for access to the ARTFL database for full-text searches of old French texts, 90% of which are in the public domain.  ARTFL commands this subscription price by offering really good search on a set of documents that would almost certainly fit on one DVD and, as I say, are largely in the public domain (and for me, the ones that are not in the public domain are useless anyway, because they're too recent).

I can't imagine what sort of subscription price you could command if this were for something that people spend and make more money on than writing scholarly papers on French literature and history, read only by your 200 closest colleagues (if you're lucky).

In this model, by the way, the search is 100% algorithmic, but the selection of data available for search is largely manually determined, which has certain advantages/disadvantages in terms of scalability once you're up and running.
« Last Edit: February 21, 2011, 09:12:06 PM by ergophobe »

dogboy

  • Guest
Re: Alternatives to 99% algo-based search - how would you rank search results?
« Reply #5 on: February 21, 2011, 10:16:22 PM »
I wonder what ever happened to the 'Direct Hits' of yesteryear and all their patents... I imagine there are a bunch of patents that, if combined, would be pretty cool.

inbound

  • Jr. Member
  • **
  • Posts: 53
    • View Profile
Re: Alternatives to 99% algo-based search - how would you rank search results?
« Reply #6 on: February 21, 2011, 10:18:37 PM »
I think ergophobe has summed up how to do well in a niche quite well - we're working on stuff like that for a few companies with great data but "sub-optimal" search.

Re the pay to use model for a broader search engine; if a service could save you time in comparison to using a leading engine, would you consider paying for it? Along the lines of using a standard search initially, the results are pulled from a search API from one of the majors, junk is dropped, extra results are added in based on whitelisted sites (which maybe didn't appear due to an extraneous word in the search) - then also giving you advanced search options that are specific to the search conducted.