Hi, My responses are mostly like Ian's. More below.
----- Original Message ---- > From: Ian Holsman <[EMAIL PROTECTED]> > Sean Owen wrote: > > I mentioned this to Otis but thought I'd expand it -- > > > > As some of you may know I am actually an MBA student. I have worked on > > the collaborative filtering portion of Mahout (formerly called Taste) > > for several years. And you still plan on working more on it, right? > > I have always been curious to know what, exactly, > > enterprises need out of collaborative filtering. What product would > > they like to see? I'm not as familiar with CF as I am with things like search, but from what I know the main 2-3 requirements are - give me GOOD recommendations - give me recommendations QUICKLY - make it SCALE to lots of people and lots of data Not too much to ask for ;) > Hi Sean. > as a ex-MBA student (I got it last year) and a guy who managed to get > this on his companies radar I might be able to help ;-) > > > Now I have the opportunity to use this as the basis of a business > > school project. So I am wondering if anyone can comment on any of > > these questions, without naming names if it's confidential -- > > > > - What do enterprises imagine they can get out of collaborative filtering? > > > > Firstly we are a 'consumer web site', different businesses will have > different value drivers. We don't sell things , but rely on advertising > as our primary revenue driver. (for Advertising the 2 things that > influence that is how many people click through/see the ad, and how much > we can charge for it). I'm looking at Taste for a consumer-oriented site as well - basic recommendation engine functionality is needed. However, Ian mentions advertising, and you can think of the ad engine as a recommendation engine, too, so this may be an interesting direction for Taste to explore. Keep track of searches, ad impressions, relate query strings and ad clicks and try to "recommend" top N best ads to display next time. i.e. implement AdSense - a piece of pie :) > we also have other KPIs that we believe are lead indicators. (that is we > increases these and the money will follow). > > They are engagement (time spent on the page), page views and return visits. > we see CF as a way of increasing most of these. Same analysis happened on my end - Visits, Visit frequency + duration, PV, PV/V, etc. -- my assumptions/hopes are that all of these should go up if the RE is good and if its results are presented well. What I expect to go down is the bounce rate. > In advertising we can use it to put more relevant ads (behavioral > targeting) in front of the user increasing the chance they will click on > it, and conversly give us a wider selection of ads to choose to put on > the page (higher RPM). Same thinking here. > on the content side, if we can use CF to create a better experience for > the user, they will a) come back b) tell their friends. you can check Again, same thinking here. > out baynote.com if you want some marketing whitepapers on how they sell > it for more detailed examples. > > - At what stage is the interest -- barely know about it, ready to use it, > other? > > > > actively developing several applications that make use of basic CF, and > researching other more advanced uses. Early stage of building a POC around Taste. Building a data-collection mechanism, storage mechanism, and figuring out how to feed the data to Taste, do so quickly, frequently enough, etc. > > - What do people use to solve these needs today, custom code, other > > products? > > > mahout naturally In my case this need was never solved. I know there are other commercial recommendation engines, and I believe some enterprise search vendors have recommendation modules or at least mechanisms to feed user behaviour/search history back into the core search engine for user-specic relevancy ranking adjustments. > > - What's the biggest need enterprises have from collaborative > > filtering products or services? > > > a way to measure how effective CF products are, and a easy way to > install it on a given site. I think it's that list I presented above: quality, speed, scale. As I described above, I'm at the point where I just want to see if the recommendations can really be good. Then, if they are, I need to figure out how to put things together so they scale and so on. > > - What's the one product or feature an enterprise would love to see in > > the market, in terms of CF? > > > > get your own business plan ideas ;-) I'll keep my itches to pitch to my VCs > ;-) That's going to be tough in this market :) My answer: I don't know yet, ask me in 3 months. Otis
