January 16, 1997 -- Techniques for personalizing Web sites

Transcript of the live chat session that took place Thursday, January 16, 1997.

These sessions are scheduled for noon-1 PM US Eastern Time (GMT -4) every Thursday.

These sessions are hosted by Richard Seltzer. If you would like to receive email reminders of our chat sessions, simply send a blank email message to or go to and sign up there.

For transcripts other previous sessions and a list of future topics, click here.

For an article on how to make "business chat" work (based on this experience), click here.

Since the chat itself happens at a rapid pace, it's often difficult to note interesting facts in particular URLs as they appear on-line. Here's a place to take a more leisurely look. I've rearranged some of the pieces to try to capture the various threads of discussion (which sometimes get lost in the rush of live chat).

Please send email with your follow-on questions and comments, and suggions for topics we should focus on in future sessions. So long as the volume of email responses is manageable, I'll post the most pertinent ones here for all to see.

Threads (reconstructed after the fact):


Richard Seltzer ( - 11:56am -- The scheduled chat is on Business on the WWW. If you are here for that discussion, please identify yourself.

We're here to share experiences about doing business on the Internet -- particularly the World Wide Web. What works? What doesn't work? Why? What are the trends that matter? How can you/should you adapt to the Internet culture and environment?

Richard Seltzer ( - 11:57am -- I work for the Internet Business Group at Digital Equipment in Littleton, MA. In that capacity, I end up talking to people from large companies about how they can use the Web for business. I also have my own personal Web page -- which is content rich and no frills -- which I do for practically nothing and draws a fair amount of traffic and attention.

Richard Seltzer ( - 11:57am -- In a chat session like this things can get pretty frantic. It's sometimes difficult to follow the threads of conversation. And there's no time to write down interesting URLs and facts. So last week, I took a copy of the raw transcript and edited it to make the threads clearer and posted it at my own little Web site so anyone could take a look. You can see it at I plan to do the same today. Barring technical difficulties, I hope to have a transcript up later today.

By the way, even if you tune in for a chat session and are there the whole time, you probably should check the transcript periodically to see what new material has come in as followup mail messages.

Sheldon Foisy ( - 11:59am -- Good day name as can be seen is Sheldon Foisy. I work for an electronic forms design company on the west coast of British Columbia B.C.

Richard Seltzer ( - 12:00pm -- Welcome, Sheldon -- are you working on Web-based projects?

Sheldon Foisy ( - 12:02pm -- Our products are more geared towards Intranets, but of course, the technology involved is very well suited to applications over the internet as well as the intranet.

Kathy ( - 12:00pm -- Hello, my name is Kathy Gill. I'm in Seattle and am a web consultant. This is my first time on this chat board.

John DeTreville ( - 12:00pm -- Hi Richard, this is John DeTreville from Digital Equipment Corporation.

Sheldon Foisy ( - 12:01pm -- Hello Kathy and John.

Richard Seltzer ( - 12:01pm -- Welcome, Kathy and John. John, by the way, is the lead developer for Each-to-Each, a unique technology for capturing people's preferences and tastes, which, I believe, can then be used to "personalize" Web sites.

John DeTreville ( - 12:03pm -- That's right, Richard. I'm a researcher in Palo Alto, and I've built sites like EachMovie ( that help connect people with content.

Ted Resnick -- ( - 12:06pm -- Hello, this is Ted Resnick. I am a marketing manager at Excite. ( - 12:01pm -- Sean Brunnock from Server Corporation

Bob Fleischer, ( - 12:01pm I work for Digital's System Integration organization, my particular group serves the worldwide Internet and Intranet Practice.

Richard Seltzer ( - 12:02pm -- Welcome, Sean and Bob, Sean -- what is Server Corp? do you make products for the Internet? or do you use the Internet to promote your business? ( - 12:04pm -- I've been consulting for years and developing for the Web for the past year. I have some ideas about Web products that I'm developing. One is a Web-based PIM. It's at

Todd ( - 12:02pm -- Hello, I'm a DEC developer.

Lauri ( - 12:02pm -- Hi. I work with Bob for Digital's Systems Integration org. too.

Richard Seltzer ( - 12:05pm -- Hello, Jonathan, I believe you are with Firefly. I'm very interested in learning more about your site and your technology and also trying to get a sense of how that is similar to and differs from Each-to-Each (John Detreville is here to help us understand that).

Richard Seltzer ( - 12:03pm -- It's good to see people coming on so quickly (too quickly to say hello to each). Today, we plan to focus on techniques for personalizing Web sites and how to make the best use of them. In particular, I'm interested in Each-to-Each and Firefly John Detreville the lead developer of Each-to-Each and Jonathan Sheena the lead developer of Firefly have both said they'll try to connect.

Jonathan Sheena ( - 12:03pm -- Good afternoon

Richard Seltzer ( - 12:04pm -- We also want to continue our discussions of intranet development priorities (What characteristics should a development group look for to determine which projects are good candidates for applying Web technology and which should be avoided?) and Millicent micropayment technology (spokespeople for Millicent, including Mark Manasse, the lead developer, should be joining us again).

Stan Hayami ( - 12:02pm -- Richard - hello, I'm online if anybody has any questions on Millicent, Digital's Microcommerce Payment Protocol

Mark Manasse ( - 12:14pm -- Hi, all.

John DeTreville ( - 12:15pm -- Hi Mark! (Mark's down the hall from me.)

Mark Manasse ( - 12:17pm -- No, I'm not. I mean, yes, John, usually I'm down the hall, but right now I'm still at home, about thirty miles away.

Richard Seltzer ( - 12:04pm -- And, of course, we also welcome any other questions related to business on the World Wide Web. This is an opportunity to share your experiences and brainstorm about your plans, and try to get answers.

Kathy ( - 12:08pm -- Hello, I'm working on an Intranet for a Fortune 100 company. Interested in personaization, evaulation, ID'ing user needs/preferences.

Sudha Jamthe ( - 12:10pm -- Hi Richard, How are you doing?

jpcarey ( - 12:22pm -- hello gentlemen

Sudha Jamthe ( - 12:24pm -- Hello jpcarey. Welcome

jpcarey ( - 12:25pm -- Is everybody here from the United States?

Kaye Vivian ( - 12:25pm -- Hi everyone. Kaye Vivian, New York communications consultant here. Sorry to be late, client matters.

Shin Tohi ( - 12:25pm -- Hi Sudha,

jpcarey ( - 12:28pm -- Hello Shin, Where are you from?

What Do We Mean by "Personalization"?

Sudha Jamthe ( - 12:12pm -- HI Jonathan (and John), Can you share your ideas about how people perceive personalization on the web?

Jonathan Sheena ( - 12:17pm -- Sudha, sure. People call a lot of things "personalized" -- from Levi's to electronic news. Personalization has come to mean many things. I like to differentiate between "customization" and "personalization" it's a fine line, but we're trying to differentiate between the *explicit* customization that most people associate with personalized systems ("give me news about the red sox") and implicit feedback-based learning.

John DeTreville ( - 12:20pm -- I have different vocabulary from Jonathan but the same distinction. Different kinds of personalization technologies can work in different domains. A learning system is great for a domain where there's a high degree of subjectivity, where there's no right answer.

Sudha Jamthe ( - 12:21pm -- Hi Jonathan, Yes, I agree personalization is often mixed up with customization. I think the user has the power to change their selection or look and feel of their site in personalization, while customization goes one step further and gives the user exactly what they want.

Richard Seltzer ( - 12:28pm -- Sudha -- That's a good distinction -- personalization vs. customization. In the extreme case, a user sees material that is absolutely unique, based on a profile. In the case of personalization, tastes are taken into account and it is made easier for people of common interests to link together and help one another. When the customization reaches the level of the individual there is no mutual benefit, and no community effect.

Shin Tohi ( - 12:33pm -- Hi Richard, Why do you feel when mutual benefits are lost when customization is at individual level? I think thats when the customer gets most benefit and will be willing to pay for it.

Richard Seltzer ( - 12:41pm -- Shin -- There are cases when customizing to the individual level is exactly what people want, and yes, they'll be willing to pay for it. There are other cases where your business model is going to be built around creating communities of people with common interests. In that case, tools that statistically determine tastes and hence "personalize" are far more valuable and important. While we all are usually very much alone when we type on our keyboards, and hence it's only reasonable that there are some private and individual things that we want to do on the Internet, it still feels like the really compelling draw of the Internet is interaction and community.

John DeTreville ( - 12:17pm -- We ask users, do you like having recommendations available?, and they say yes. In the movie domain, you would ultimately like to quantify things in terms of increased sales, etc., but we don't have that yet.

Bob Fleischer, ( - 12:17pm -- It probably would help to have a definition of "personalization" -- there are a lot of web implementations that "personalize" pages and entire webs for users and groups, most of which are NOT like Firefly or EachMovie in using the preferences of others to influence the results for a user. How narrow is "personalization" in this discussion?

Richard Seltzer ( - 12:20pm -- Bob -- I'd like the discussion to be as broad as possible -- to open our imaginations to the possible ways of personalizing a site and the kinds of benefits (and pitfalls) involved. Tell us about the sites/technologies you feel are related, please.

Bob Fleischer, ( - 12:25pm -- For one, there's the Microsoft Personalization System, a platform for delivering "personalized experience" --

John DeTreville ( - 12:29pm -- Bob, I've assumed from the little I've read that Microsoft's system is a pretty bare-bones "database" system, to invent terminology from my earlier example. You ask people what they want to hear about and match them with content. If you don't see the right content, it's your fault for not listing the right interests. With a statistical-based system (a "learning" system), you don't need to decide what you want to read months ahead of time.

Richard Seltzer ( - 12:29pm -- Bob -- can you explain a bit what's included in that personalization module of Normandy? I haven't been able to decipher it. How do they define "personalization?"

Bob Fleischer, ( - 12:34pm -- John, I actually like it being "my fault" that I don't get the right results -- I can do something about that! My point in entering the reference to Microsoft (and there are similar packages from smaller software firms) is that the term "personalization" is already ambiguous inthe market -- they have a legitimate claim to the term, but what they deliver is, obviously, quite different.

John DeTreville ( - 12:40pm -- Bob, I agree that Microsoft's stuff is annoyingly vague! As for whose fault it is if you miss something, the point I wanted to make is that we're giving the user an adaptive tool, and if it doesn't work, maybe it's not a good tool. Time to look for another....

Reservations about Personalization

Jonathan Sheena ( - 12:18pm -- Many people are afraid that personalization means a loss of serendipity, when really how broad or narrow my interests are is just as important a part of my "personalized" experience as what sports teams I like.

John DeTreville ( - 12:24pm -- I absolutely agree with my colleague from Firefly. A couple of years ago, when video-on-demand was hot, I heard a professor from Berkeley talk about how consumers would interact with such a system. That would ask, say, for a John Wayne movie, containing a certain line of dialogue, and the system would look it up in a database. (Did I mention this was a database professor?) This is a horrible way for people to pick movies; you'd like it to be a lot more subjective and to include the opportunity for serendipity.

Sheldon Foisy ( - 12:20pm -- Jonathan - Don't you find that personalizing sites beyond a certain point tends to push people away from that kind of site rather than entice them in?

Jonathan Sheena ( - 12:25pm -- Sheldon: it tends not to be the fact that a site is personalized per se that drives people away, but the fact that a user has to divulge more for less return. John's point is well taken, in that there is no one solution that works in all domains, and that methods for personalization need to be applied to the domain specifically.

Sheldon Foisy ( - 12:26pm -- Well put Jonathan...

Richard Seltzer ( - 12:25pm -- Sheldon -- There are many flavors of personalizing. Actually, that's what I'd like us to focus in on -- the varieties of approaches and when and where they are appropriate or inappropriate. Keep in mind too that pages dynamically created to fit a personal profile won't be indexed and won't be found by the major search engines, unless there is a static page with the same content.

Sheldon Foisy ( - 12:31pm -- Richard - Jonathan I think put the finger on the crux of most attempts it seems at "personalizing" pages, which is that the user has to divulge more than they are aptly rewarded for. There still is some appeal to the "mystery" of web shopping I think, and in order to entice the potential client into your site, there has to be as few obstacles as possible. Whether that be web design or content problems with the site. I agree that personalization is an excellent idea, but perhaps after the potential client has "gotten used to the temperature of the water" so to speak.

Richard Seltzer ( - 12:37pm -- John -- I love the statistical way Each-to-Each works. I don't have to define or categorize my tastes, and no one is presuming that my level of education or age etc. has anything to do with my tastes. You base your profile on how I react to particular movies, etc. and how others have reacted to those same movies, etc. So I can go to EachMovie and find out how I would have rated a movie I haven't seen before (not how some critic has rated it.)

Ted Resnick ( - 12:39pm -- Richard, the only thing that I was disappointed in about EachMovie was how few of the movies that I had seen on your list. Will you be able to let users add their own favorite movies? Seems that there are thousands out in videotape that can be thrown into the database.

Shin Tohi ( - 12:39pm -- I love the each-to each idea of personalisation I think this is what should be done not just for movies but all contents like magazines, and websites

Sudha Jamthe ( - 12:41pm -- John, Richard, With Each-to-Each movie rating, I as a user, have to take responsibility while the ratings were done by someone else so far. So, I think we need an attitude change to benefit from it fully.

Mark Manasse ( - 12:29pm -- In the design of EachRadio, we attempted to get at serendipity by tossing a bunch of things into the mix: things we thought you would like, things we didn't know about, and things we thought you wouldn't much like, just so that we could test our model sometimes.

Bob Fleischer, ( - 12:30pm -- I suspect that there are some lessons to be learned from the text retrieval community -- the idea of using the entire community's queries (and feedback on the results) to influence each individual's search results is an old one (and, as with so much text retrieval research, probably never has been used outside of a research setting).

Jonathan Sheena ( - 12:34pm -- Bob: There's oodles to be learned from text retrieval companies. In effect they've been down all these roads before. They *know* feedback. Firefly tries to build that community layer you're talking about on top of what you could learn from the statistical musings of the verity's and intext's of the world.

Personalization -- Each to Each and Firefly

John DeTreville ( - 12:08pm -- Since this is all pretty asynchronous, let me just jump in with a spiel. There's too much content on the Internet, just like the real world. This of a Blockbuster store, and imagine how much harder it would be to navigate on-line. I work on technologies that (should!) help make it easier. In particular, I do statistiocal analysis of other people's interests to help guide you where you might want to go.

Jonathan Sheena ( - 12:10pm -- Well, (and John, you should jump in to this) Each-to-each and firefly do do some of the same things with regard to personalization, but whereas each-to-each is concentrating on server side solutions, (like each-toy for example) firefly adds the network dimension by letting different firefly enabled sites take advantage of the collective knowledge of the network as a whole. -- and I can go into detail wherever people are interested.

John DeTreville ( - 12:11pm -- Hi Jonathan. That's pretty much what I'm doing too...

John DeTreville ( - 12:13pm -- The great thing about this kind of recommendation system on the Internet is how you can create a positive-feedback loop. If I enter information about what I like and don't, then I can get immediate gratification in the form of better recommendations, plus my information can be shared, ultimately, with every other person on the Internet. It's a very powerful win-win situation.

Richard Seltzer ( - 12:14pm -- Jonathan -- What do you mean by a firefly enabled site? Can you give us some basic definitions?

Jonathan Sheena ( - 12:21pm -- Richard, sure. When I say "firefly enabled site" I really mean a firefly enabled catalog. The firefly tool set today gives a site the ability to navigate a catalog (be it books, movies, travel destinations, mechanics, etc...) based on the opinions of the community. We call that "opinion-based navigation" This means a site can recommend from, and let the user search through items from its catalog with their preferences in mind (read: personalized). ( - 12:12pm -- John- Where is your company located? I realize that Firefly is in Cambridge.

John DeTreville ( - 12:14pm -- My company is located in a corner of my office in downtown Palo Alto. We're just up Route 101. I'm

Todd ( - 12:13pm -- Jonathan - how does FireFly make it's money? Is it all from the banner ads or are there other sources. How much does it make from the Ads? ( - 12:25pm -- Can software do a better job at recognizing people's tastes and making recommendations than they can at recognizing speech (wreck a nice beach)?

Kaye Vivian ( - 12:30pm -- Perhaps this was covered already, but I once again visited the Firefly site this past weekend and did the "choose your favorites" music thing to see what recommendations came up, I never did actually get any recommendations! I did notice after testing for about 30 minutes, and completing perhaps 200 or more choices that it was making more choices available that I recognized, but nothing more than that. I was rather disappointed. I agree with Sudha's comment that personalization is often confused with customization. It's nice, for example, for Netscape to set a cookie that says "Never show frames to Kaye Vivian when she's here" but what I would prefer is to say "Turn off all blinkings text and ads, cue up these five pages that I visit regularly, and on the basis of that, pull up anything new that you think I might be interested in, based on my past activities on the site. To me, that is the ultimate. I haven't seen any of the each-to-each sites that can yet do that.

John DeTreville ( - 12:34pm -- Kaye, I agree with you. For me, the most interesting problem is the "pull up anything new I might like" part. Right now, EachMovie and the other little eaches are pretty content-poor, but you could imagine using the technology as agents bringing you interesting things from the news, etc.

Sudha Jamthe ( - 12:35pm -- Hi Kaye, You are suggesting the ultimate pinnacle of cookies usage. The web is at a stage people are debating if Cookies are good or bad. You are asking about personalization using your cookie. Maybe we'll get there one day. Neat thought!

Ted Resnick ( - 12:42pm -- John, I agree that these are tough hurdles, but I think that it is necessary for them to be worked out. Like Kaye, I spent a half hour or more at one of the collaborative filtering sites, and I just was not impressed with the results that I got. I think that it is more rewarding to chat in a newsgroup for say, gothic music, to learn about other gothic bands than I had heard of.

John DeTreville ( - 12:42pm -- Ted, the only reason we don't do this now is that our experience is that users can't spell, and so you'd get a poor signal-to-noise ratio. Maybe it's time to rethink that....

Kaye Vivian ( ( - 12:45pm -- Very good point on the spelling, John...guess you would need an intelligent spell-checker to help the reader get to what they are after?

John DeTreville ( - 12:48pm -- Kaye, as for the spelling, you just need some back-office work to throw out inadvertent duplicates, fix spelling, etc.

Kaye Vivian ( ( - 12:43pm -- Ted, I agree!! I visited the EachMovie site and did the entire list of movies. Perhaps I did something wrong, but it didn't give me any recommendations back .. Anyway, being able to *add* titles would be a cool approach (of course, I guess it would be difficult to manage, since, for example, there are thousands of Indian and Russian movies that are so obscure that most people wouldn't recognize the title translations, let alone be able to find them in their local video stores.

John DeTreville ( - 12:48pm -- Kaye, if you rank every movie in our system, of course we can't recommend any you haven't seen! (Or am I missing something?) No matter. The people have spoken. We'll add the ability to add movies to the system.

Mark Manasse ( - 12:49pm -- I think there are significant differences in the underlying algorithms behind collaborative filtering sites, so don't dismiss the whole idea as junk just because one site didn't prove helpful. Also, you have to have enough users to make it work; that can fail. Kaye, clearly EachMovie might need a little UI tweaking, because you should have gotten some recommendations after your first dozen-or-so entries, if you had asked.

Possible Applications of Each-to-Each/Firefly Personalization

Richard Seltzer ( - 12:17pm -- John and Jonathan -- I believe that personalization is important for creating on-line communities -- linking together people of like interest. (I like the beginnings of that that I see at the firefly site, with chat sessions etc. created for people whose profiles are similar.)

Richard Seltzer ( - 12:18pm -- John and Jonathan -- I also see tremendous value for Each-to-Each and Firefly in store environments like Blockbuster and Toys R US -- helping shoppers on-line and in the store as well to make selections based on their taste.

Richard Seltzer ( - 12:24pm -- John and Jonathan -- I'm also very interested in another potential application of your kind of technology. When I was at Quantum Books in Cambridge recently, talking about AltaVista search, it dawned on me that little book stores -- even very Internet-savvy ones like Quantum -- are at great risk right now. If I can order any book in the world through, why do I need to order through any other on-line bookstore? (or from any on-line publisher either)? And it occured to me that stores with a particular niche -- which Quantum certainly has -- could let their audience rate the specialized books that they read and set up ticklers (by email or at the site) to alert people of similar tastes that a book they would like is now available. ( - 12:27pm -- Richard- I read in an interview that the founder of ( doesn't believe that bookstores will fold up shop. Even he purchases many books from his local store.

Kathy, ( - 12:33pm -- Sean, Richard -- yes, Amazon.Com doesn't think it will ever replace bookstores. Too much tactile, social experience (especially now that coffee houses and bookstores are co-habitating).

Richard Seltzer ( - 12:34pm -- Sean -- I don't mean that physical bookstores will fold up shop. I like the immediate gratification of walking away with my purchase, and like to hold and look through the pages of a book. What's at risk is the Web storefront extension of the physical bookstore or even the book publisher. There we're talking Web-based vs. Web-based. And the question -- not just for books but in general -- is how does the individual small store survive on the Web if there is a super-store selling virtually all of the same kind of merchandise (winning by immensity of choice). And I believe that the answer is by identifying niches and building communities and using personalizing techniques, so when I connect to The Computer Technical Bookstore I am participating in a community of folks interested in those kinds of books and with discerning tastes, and I can learn there about books that I want and need and wouldn't have otherwise heard about it. ( - 12:39pm -- Richard- has an interesting reseller program. Small websites can offer books for sale using as a back-end. In the small web-bookstore vs. Amazon, I think this is a case where it's easier to join than to fight. Is this a bad thing? Even is a reseller for a book warehouse in Oregon.

Bob Fleischer, ( - 12:45pm -- Sean, in the reseller program, do small book stores actually get to deliver the books? What I'm thinking I would like in many cases is to be able to select a product I'm interested in on the web, find out all the data on it, including price, but then be directed to a local dealer who actually has the thing. Some things I just want to handle in person before I buy, and have in my hands immediately afterwards. Even so, the web could help me select what I want and then help me locate where I can definitely get it near me.

Kathy, ( - 12:46pm -- Bob -- no does the delivery. The associate web site (doesn't have to be another book store) gets a commission. I'm an associate. ( - 12:48pm -- Bob- the model that you describe is very similar to Perhaps, books don't have a hign enough margin to justify this approach.

John DeTreville ( - 12:51pm -- As for margins in book-selling, Amazon is currently offering its partners 8% of the selling price of books that are not already heavily discounted. There's plenty of room for a business there. (I say that even knowing that Amazon has signed with Net Perceptions to do recommendations on their own site: there's room for both!)

Todd ( - 12:55pm -- Bob - I agree with the idea of a pointer to local sources. I like obscure videos, and it would be great if I could find a local store that has them. It could even tell me if they're rented or not. Also, some companies seem to leach off others; like funiture companies that rely on you shopping in competitor's showrooms and then buying cheap from them.

John DeTreville ( - 12:57pm -- Todd, that's called disintermediation, and I have a long rant already saved up for next week.

Jonathan Sheena ( - 12:59pm -- disintermediation -- that's my favorite word this week.

John DeTreville ( - 1:00pm -- Reintermediation is good too, and that's what we're doing.

Richard Seltzer ( - 12:58pm -- Wish list -- stores that don't just help with personalized selection, but also have on-line inventory -- so I find out about the great video I want to see and can tell which Blockbuster store near me has it, and can reserve it on-line before I go out in a snowstorm to try to get it.

John DeTreville ( - 12:59pm -- Absolutely. You'd like to rent from Blockbuster on-line and have the Domino's guy deliver it. This technology is one way to make that feasible.

Jonathan Sheena ( - 1:02pm -- Richard - I think the hardest part for online services (like the Blockbuster example) will be setting up those relationships between the service and the blockbuster. Inventory tracking is something libraries, and FedEx/UPS's do very well -- Blockbuster could certainly learn something from them.

Personalization plus Micropayments?

Richard Seltzer ( - 12:07pm -- I'm hoping we can all multiplex a bit here. We began discussing Millicent last week, and I know there were further questions about that, which Stan Hayami could field. I'm also interested in how technology like Each-to-Each and Firefly for personalizing/ understanding tastes, might be tied together with micropayments technology like Millicent. It's great to have neat new technology, but it's far better if you can get paid for it. And I'm not much of a believer in banner advertising as a long-term source of revenue.

Stan Hayami ( - 12:11pm -- Richard - tying Millicent and Each-to-Each would be a killer combination. However, because both technologies are in their respective infancies, the combination AT THIS TIME could be mutually detrimental. However, this could be a great combination once both products are accepted on their respective merits.

Mark Manasse ( -- I believe that there are a lot of things that connect collaborative filtering and payment, and that we're just starting to think of all the ways they can work together.

Todd ( - 12:19pm -- Stan - How would you tie the two together? How would that work?

Mark Manasse ( - 12:24pm -- I'll jump in for Stan; some people will be willing to act as guinea pigs (and receive whatever benefits of serendipity there might be), while others will want pre-filtered information for the most part. Those people who aren't willing to do the heavy lifting of contributing rankings early might well need to pay the others, and Millicent could come in handy there.

Mark Manasse ( - 12:26pm -- Moreover, the authors of the provided content deserve compensation related to usage; a micropayment system might make it easy to adjust the price based on the degree of confidence the system has in the rankings.

Stan Hayami ( - 12:27pm -- Todd - it could work a number of different ways, but the best application which come to mind would be to charge ($0.01 for example) for each query, then payback ($0.005 for example) for each time you fill out a questionaire, then payback ($0.005 for example) for each time a user views an advertiser's banner. These are just musings at this point, nothing definite.

Kathy, ( - 12:30pm -- Stan -- good idea, but that means we must all have a 'bank acount' ... and how are we going to move to micro-payments/credits when the mentality is still $3.50 per copy?

Stan Hayami ( - 12:35pm -- Kathy - good question. Please go to our web site ( to get the answer to your questions on bank accounts. As for customer mindsets, that is a challenge that every new technology faces when it is introduced into the market. But change mindsets we must in order for humankind progress to be enabled.

Kathy, ( - 12:42pm -- Ah, Stan, it's not the customer mindset I'm thinking of -- it's the bean counters at businesses! Go to the Harvard site -- they are selling HBR articles for $3.50 -- MAILING them to us. What a waste of web space -- and overpriced for electronic delivery. Same with Lexis/Nexus and other traditional info providers -- who are accustomed to charging $3-$5-$6 for a specialized newsletter article .. in part because the market for the product was low (relative to the potential audience off the Net). Make Sense?

Stan Hayami ( - 12:46pm -- Kathy - you make excellent sense. Publishers on the web are basing their business models on their real world experience, and not being too successful at it, I might add. Once they see the power of Millicent and the ability to collect revenue from volume rather than high pricing, their mindsets will be changed as well.

John DeTreville ( - 12:31pm -- I don't think that you need to get as heavily into micropayments as Stan is suggesting. Let people get recommendations for free, get your revenue from advertisers and from pass-throughs, then pay a share to your best contributors. I think Mark would disagree.

Mark Manasse ( - 12:34pm -- John, I won't disagree; ads are part of the mix that are going to make up the whole of electronic commerce. Personally (as I've told John), I'd like to see collaborative filtering technology used to better target ads!

Jonathan Sheena ( - 12:29pm -- Mark, you get into an interesting situation there. First off, using opinion based navigation (Automated collaborative filtering) the engine always gets smarter the more people have used it. So the later-ons do reap the benefits since fewer rankings will yield better results than they would have at the engine's inception. Adding payments into the mix gets you into an interesting situation since the rankings are just as valuable to the user as they are to the system as a whole.

Mark Manasse ( - 12:31pm -- Jonathan, that's exactly right. Your opinions (the early ones) ought to be worth money, time, or something to others.

Ted Resnick ( - 12:32pm -- Jonathan, I think that you hit on one of my reservations about collaborative filtering. The early users get back very few results, and usually very obvious ones. How does a system entice people in so that they can achieve a critical mass that would benefit later users?

John DeTreville ( - 12:38pm -- Ted, as for early users, it depends on the technology. We've built a number of prototypes wherein we can give useful suggestions with as little as one datapoint from the user, which can even be demographic information. Remember when you last went into a large drug store looking for a particular item? You find it by realizing that similar items are shelved together. Technology gives us the opportunity to shelve things in much more interesting ways. Remember that this technology does not replace others, so the shopper can still shop at random, or alphabetically, or whatever, but our experience is that even first-time users can benefit from this approach.

Jonathan Sheena ( - 12:38pm -- Ted: There are a number of fallbacks a system can take before it's reached critical mass in its domain. 1. fall back on using demographics to find similar people (mass marketers do this all the time). 2. use strict popularity as a measure of relevance, 3. use knowledge from another domain (use my movie ratings to predict what music I might like). All these solutions are imperfect, yet give a site (and a user) a real head start. In the end, it's a tough hurdle, but one that given enough data from other domains, you can stard jump starting new domains faster and faster. (I hope that made sense).

Jonathan Sheena ( - 12:47pm -- Ted, Kaye: Yes. There's a lot of work that needs to be done as far as algorithms, interfaces, etc. One thing that would help would be the ability for me to group my ratings under a label, or a "mood". For example, given my adventure ratings, find me a good adventure movie. We do this on the web with My Yahoo, and it certainly seems to help the recommendations, but does add interface complexity.

Personalization Plus Search

Ted Resnick - 12:23pm -- Sorry, my Netscape froze :-(

Richard Seltzer ( - 12:09pm -- Welcome, Ted. Glad you could join us. It's easy to imagine how technology like Firefly or Each-to-Each could be combined with search engine technology so people could tailor their searches by their tastes as well as by information. Can Excite do anything like that now? Is that a direction you would like to head in?

Ted Resnick -- I think that Excite is interested in these tools, and I would like to learn more about what the capabilities are these days.

Ted Resnick ( - 12:26pm -- One low level way that Excite Users can collaborate on site recommendations is on ExciteSeeing Tours. This does not use software like Ffly or Each-to Each, but instead is set up in a bulletin board style. That is, we might have a discussion topic for favorite football sites, and people can talk about why one is better - what unique areas one site may have above the others.

Bob Fleischer, ( - 1:00pm -- I'm working right now with a customer that wants to set up a very large governmental web site, and they would be very interested in personalized views of the vastness of govenment and government publications! They care very much about the perceptions of their "user community". I suspect that in some settings collaborative filtering will be a third way to findt hings, complementing formal classification hierarchies and full-text search.

Personalization -- Coola

Richard Seltzer ( - 12:13pm -- Sudha -- Welcome. Things are happening fast here. I'm hoping that some conversations can start up and continue independently in this very asynchronous mode. Could you please tell us about Coola? I took a look this morning but didn't have time to use it the way I'd like. Looks like it uses some new mail-based capability from Netscape. Can you tell us something about that capability and your implementation? and how that relates to personalization?

Sudha Jamthe ( - 12:18pm -- Hi Richard, This chat is exciting. I am trying to catch up in all directions. About coola: Yes, its true personalization. Coola delivers magazines to the user. The user can select just what magazine they want. He/she can look at it from their inbox mail at their leisure. We went to production recently and have got an amazing response.

Sudha Jamthe ( -- We are personalizing magazines by coola mail at, but are debating about whether we should customize for the user or leave it to each person to decide.

Shin Tohi ( -- Sudha, What do you mean by Coola sending personalised magazines? Can you elaborate?

Sudha Jamthe ( - 12:30pm -- Hi Shin, Coola provides the user with a selection of magazines. You as a user have a free subscription to Coola where you can choose just what magazines you want delivered to your mailbox. You can change your selection any time. Say, you are studying a particular course in school. You can make a Coola selection for exactly those periodicals and books that interest you this semester . then you can change it when your interest changes. You don't even have to visit many times. Once you visit the first time, coola will send you periodic updates of new magazines available with a form leading to to the Coola member lounge. For example, we are adding Richard's Internet-on-the disk from tonight.

Shin Tohi ( - 12:44pm -- Hi Sudha, I see a lot of magazines collection in Coola. But I don't know what I'll like. Can Coola recomend me some magazines the way Each-to-Each does?

Sudha Jamthe ( - 12:49pm -- Hi Shin, Currently coola shows magazines by various categories. So, you can choose the category of your choice. We are adding a "pick of the week" on our front page that will show the most selected magazines of the week. I think it will be great if we can use agents to just show you what your like. But, realistically this will take time to build. So, to begin with you have to make selection from the whole list. One workaround we are planning is to give pre-selected profiles for you to choose from. You can always choose the profile that matches you and add more from the general selection.

Bob Fleischer, ( - 12:52pm Sudha, where is "coola"?

Sudha Jamthe ( - 12:55pm -- Hi Bob, coola is at We have not registered in search engines yet because we are getting heavy hits already, especially since some schools like MIT are interested in their own personalized version of Coola.

Kaye Vivian ( ( - 12:55pm -- Reading your comment, Sudha, prompts me to ask...what is the price range on installing a collaborative front end/AI to aid in these selection processes? And aren't these really most appropriate for consumer-type products? I mean, I can imagine a lot of business sites with a lot of content available which might want to offer additional information to people who browse their sites, yet, as a user, I'm not sure I want a large accounting firm or my bank or the Department of Commerce "suggesting" additional publications to me. For me, it's enough to visit their site, have categories to choose from (so I can narrow down my own choices) and then let me browse or get in and get out, as I want. I'm not sure I'd have confidence that these collaborative front ends would actually help me do a conclusive search I could rely on, based on my three experiences so far with each-to-each. BTW...I do love the Eyes E-mail service Amazon offers. That is truly a service!

Sudha Jamthe ( - 12:59pm -- Kaye, you've hit upon exactly where the general consumer fear lies about collaborative front ends. If you want to truly use the power of the web, as we get more and more good content, you cannot afford to surf all the time. You have to develop trust on a collaborative agent. Its depends on who provides it. For example, any publisher can try to send reminders of their updates instead of Coola. But, think about how many sites you need to register and whose interest the flooding mails will serve?


Kaye Vivian ( ( - 12:38pm -- If I may interject a diversionary question... Two web sites I manage right now want to add conferencing capability similar to this one. Can anyone recommend any other really outstanding chat/conferencing rooms they have seen so I can check them out? (Richard, I did talk to Kathy Grobe this week). I'm wide open to options right now and think this format is about the best I've seen. Comments? ( - 12:40pm -- Kaye- I wouldn't recommend the chat software.

Richard Seltzer ( - 12:43pm -- Kaye -- Keep in mind that the human element is probably more important than the technical format for chat/forum. It takes some time and effort to plan scheduled chats, get out the invitations/ reminders, edit and post the transcripts. Without that kind of effort, a chat like this can quickly devolve into plain old chatter that really goes nowhere.

Kathy, ( - 12:44pm -- Kaye, I agree with Sean. This is not user-friendly software -- for one thing, the messages scroll upside down (for my mindset).

Richard Seltzer ( - 12:45pm -- Kaye -- What I'm trying to say is that whatever technology you choose for on-line conferencing, be sure that your customer understands that this stuff doesn't work automatically, that the success or failure of a given implementation will depend on the work and creativity put into the content side.

Jonathan Sheena ( - 12:44pm -- Kaye: You've hit on some of the really neat stuff we're looking into with personalizing news and chat. You might imagine this: now that I know some things about the people in a chat room, I can find rooms where there are people who are most similar to me (or most dissimilar, if I'm looking for a good argument). Then the next step is to facilitate chat around a topic or an item (like a news article, or a news topic ... take this chat here for example) and once you know what item or types of items a chat is revolving around, you can use that information to navigate chat space just as you navigate music space, or book space or any item space.

Kathy, ( - 12:45pm -- Have any of you used Chat software on intranets -- for employee communications/community building?

John DeTreville ( - 12:46pm -- Chatting about gothic music is great (for you, not for me!) but chat rooms and bboards have their own problems, such as an inability to scale very well. I was one of the first people on Usenet, and I read every article every day. Now that the Usenet has grown by several orders of magnitude, it's not as useful. With this technology, you could have, perhaps, a single, with the filtering engine helping decide who talks to whom. (You can always say, I don't like this, and the system will rematch you.)

Richard Seltzer ( - 12:48pm -- Jonathan -- I like the way you can set up chats with people of common tastes talking to one another. I'd suggest that it would be even better if you had some scheduled chats as well (with the same underlying mechanism), just so it isn't totally random that the right set of people happen to be on line at any given time. Try to inject a little order (at least as an option.)

Jonathan Sheena ( - 12:53pm -- Richard. Your point about scheduled chats hits on a much larger issue, that of how do you effectively ballance editorial input (such as that of yourself, or a newspaper editor, or a yahoo site catagorizer) with the intelligence of the system? Obviously the technology is no where near putting the Globe out of business, but there is room for a balance. Experts certainly do have a place in a personalized world, but expert voices shouldn't be the only voices -- and my experts, or editors might be different from yours.

Kaye Vivian ( ( - 12:48pm -- John S...I came late. What is your company/product? Also, Richard, I understand what you are saying about "how" to conduct the meeting, and I'm in full agreement. I have moderated quite a few discussion groups online and it's absolutely true that the moderator has to do a good job and go the extra mile in preparation and followup afterwards. You are terrific at this! What I'm interested in at this point is simply functionality...I want to see places or review software that enables conferencing or chats so I can get a feel for how we want to set up our own capabilities. I will take care of making sure the meeting facilitators know what to do and how to do it after we pick the vehicle. ( - 12:52pm -- Richard- any thoughts on new chat software? The stuff at looks pretty good.

Richard Seltzer ( - 12:54pm -- sean -- I'd also like to learn more about new chat software. I've gotten too used to using what's available here.

Ted Resnick ( - 12:53pm -- (Arrgh - got disconnected again!) If the goal is to find and participate in a community about my "favorite" topics, then I have gotten more out of the newsgroups, lists, and chats than I have with collaboration. Of course these are not mutually exclusive. I am just thinking that there may be a way to combine the best of all worlds. I would love to see a collaborative site that allows fans to identify each other and participate in a public forum on the same site(real chat or archived BB), as opposed to getting an email address that I have to follow up on later.

Richard Seltzer ( - 12:56pm -- Ted -- Firefly seems to do a bit of what you are talking about now. They do set up chats based on statistically determined tastes -- immediate connection, not just an email address. Of course, for that to really work well, they need lots of people connected at the same time -- and that will require the fastest and most powerful of servers -- good hardware that scales well. I don't know what they are using now in their beginning stages.

Ted Resnick ( - 1:00pm -- Richard, I like that they do on the fly chats, but because of the reasons that you mention, I think that bulletin boards go even further at providingmore meaningful information about the topic. I have seen traffic grow tremendously on ExciteSeeing Tours and I think it is because of that model.

Databases for the Web -- Python

Todd ( - 12:08pm -- Followup to my last week question about DBs for the Web - Mary Ellen suggested Python which is a language that has database extensions. That sounds like too much time to get going for me. I found MiniSQL (mSQL) from Austrailia ( that sounded promising, except it's Unix-only as far as I can tell. Turns out the commercial products are pretty cheap so that's probably the route I'll go. Personal Oracle Lite lists for $200 and MS Access is supposedly less than $100.

Richard Seltzer ( - 12:10pm -- Todd -- By the way, we got a followup email message from Jeff Black pointing out the Python is really a "language" rather than a database. The URL, I believe is I appended his note to the transcript at

John DeTreville ( - 12:10pm -- (Discussing DBs on the Web, I've been using MS Access with EachMovie for about a year now, and I'm having no problems. It helps to have a lot of RAM, and the database can't get much bigger than 500MB, but it's amazingly robust and stable, plus easy to connect to.)

Each Toy?

Robert ( - 12:44pm -- Richard, I've tried the Each to Each movie site. Is it basically the same for the new Toys R Us site? Is that one real, or just a Christmas trial? Thanks!

John DeTreville ( - 12:56pm -- To answer the question about EachToy... That was a quick prototype we put up in collaboration with Toys "R" Us for the Christmas shopping season. It's down now (sitting by my right foot as I type this), but EachMovie has been up almost a year and should stay up indefinitely.


Richard Seltzer ( - 12:50pm -- All -- I love the way this conversation is moving ahead and tying interesting pieces together. and I hate the way time is flying. I was planning to focus on Wireless Internet next week. Is there interest out there in continuing this discussion next week and moving the rest of the schedule out a week? How many of you want to/can join us again next week to continue this?

John DeTreville ( - 12:51pm -- I can come next week....

Mark Manasse ( - 12:52pm -- Richard, this has been great fun, but I'm not available next week (nor do I think that you need me!). In addition, I'm going to head out now, so that I can catch my train--bye, all.

Richard Seltzer ( - 12:52pm -- All -- Keep in mind that I'll be making an edited transcript from this -- reconstructing the threads as best I can. With as lively a session as this, I suggest that you all check the transcript to see what you may have missed on the fly. I'll post it at For the complete list of transcripts and the list of future topics see

Sudha Jamthe ( - 12:53pm -- I can come next week. I know many people in web-net user group who were interested in today's chat but had time-conflicts today. I think they'll join next week.

Richard Seltzer ( - 12:53pm -- All -- please post here your email address and URLs so discussions and further learning can keep going. Also, please send me email with your followon comments and questions -- I'll add them to the transcript. And remember that messages get added there from folks who for one reason or other can't connect live and just react to the transcripts.

John DeTreville ( - 12:53pm -- Bye Mark. (And what do you mean, the EachMovie UI could be a little better? :-) ( - 12:55pm -- Sean Brunnock,,

John DeTreville ( - 12:55pm -- I'm or or or or or or or or or ...

John DeTreville ( - 12:58pm -- I also think I'm, not to mention Let me think....

John DeTreville ( - 12:58pm -- And everyone should come and use !

Kaye Vivian ( ( - 12:55pm -- I can come back next week...not much interested in wireless internet though :)

Jonathan Sheena ( - 12:56pm --

Stan Hayami ( - 12:57pm -- Richard - Russ and I have a customer visit next week, so neither of us will be available. If anybody is interested in Millicent, they should visit our web site ( or send mail to to get updates on the progress of our pilot project. Bye for now.

Ted Resnick ( - 12:57pm -- I am interested in taking the discussion further next week -- Ted Resnick -- -- -- (discussions) -- (personalizations)

Kathy, ( - 12:59pm -- Thanks, Richard -- I'm particularly interested in intranet discussions. Kathy E. Gill,,

Richard Seltzer ( - 12:59pm -- Thanks to all -- please let me know your preferences for next week -- continue this or move on to other topics. (Looks to me like this one is hot.) Also suggestions for future topics welcome.

Richard Seltzer ( - 12:59pm -- Thanks to all. This has been great.

John DeTreville ( - 1:00pm -- Bye, all!

Sudha Jamthe ( - 1:01pm -- Hi Kaye, thanks. I'll visit millicent. Neat Job Richard. I hate its going to be 1.00pm now. Well organized and moderated chat. I am thankful you are posting the compilation of this chat online I have to go now. Bye

Ted Resnick ( - 1:01pm -- Till next week then!

Jonathan Sheena ( - 1:03pm -- Thanks, all.

Previous transcripts and schedule of upcoming chats --

To connect to the chat room, go to

The full text of Richard Seltzer's books The Social Web, Take Charge of Your Web Site, Shop Online the Lazy Way, and The Way of the Web, plus more than a hundred related articles are available on CD ROM My Internet: a Personal View of Internet Business Opportunities.

Web Business Boot Camp: Hands-on Internet lessons for manager, entrepreneurs, and professionals by Richard Seltzer (Wiley, 2002). No-nonsense guide targets activities that anyone can perform to achieve online business
success. Reviews.

a library for the price of a book.

This site is Published by Samizdat Express, 213 Deerfield Lane, West Roxbury, MA 02132. (203) 553-9925.

Return to Samizdat Express

| |