A Clue...to Internet Commerce

by Dana Blankenhorn

For the Week of September 15, 1997

Volume I, No. XXIX

This Week's Clue: Collaborative Filtering

One of the biggest mysteries in electronic commerce may be why "collaborative filtering" techniques offered by such vendors as Wisewire , Firefly , NetPerceptions , LikeMinds and Gustos haven't done better.

The promise of collaborative filtering is wonderful. Gather the personal tastes of many individuals, mix them in a database then deliver recommendations based on those tastes to participants. Do you like Stephen King? Maybe you'll like Anne Rice. Do you like Blues Traveler? Maybe you'd enjoy Little Feat. That kind of thing.

There are two reasons such systems have failed to take hold. The first is trust. We don't trust that our private tastes will stay private. The second, I feel, is the application, which is being used for impulse buys.

Getting over the trust hurdle doesn't take some bureaucratic seal of approval Trust is built slowly, over time. You don't run a computer through someone's buying psyche on your first date. It's the organizations with long-running trust relationships which are in the best position to take advantage of this technology. So I think Charles Schwab might use it to sell mutual funds or real estate trusts, building model portfolios based on income and risk appetites, copying what Merrill Lynch brokers do now for their clients at lower cost. Geico might use it to simulate the "family insurance check-ups" of rivals like State Farm, and enter new lines of business. Consumer Reports might use it to help car buyers select among new and used vans, based on cost, crash worthiness and other factors. (Charge for it - they charge for reports on crash-worthiness.) Internet upstarts like E*Trade and AutoBuyTel are at a disadvantage here before the game starts.

The second problem lies in what the technology is used for. Mainly it's offered on impulse buys like books and records . It's not being used for real estate, insurance, investments, or cars, considered purchases where recommendations have enormous value but don't control. It's true that it can be difficult to collect data on tastes from Web surfers, but the databases whose results are offered could come from existing warehouses.

What's needed, then, to make collaborative filtering take-off? Here are some clues. First, integration with a higher-level of database technology, represented by companies like Oracle , Sybase and Informix is needed to match the algorithms of today's players with the data warehouses of companies that can help people. Second, we need careful handling of users before, and after recommendations are made. Don't just give this to "lookie-lous" but to serious prospects. Offer ample help - even personal help - in developing input so you'll get good output. Then, use recommendations to close sales and draw value from big commissions. Don't just sell a book - sell a house.

SSP (Shameless Self-Promotion)

We got game. In May it was USA Today , calling us a "hot site." Next came Jayde , which gave our PPN site, run by Sean Cafferky of Houston, its coveted "Jayde Gold Diamond" award. Then the "San Jose Mercury News" Online Today columnist, Patricia Sullivan , wrote in her "Good Morning Silicon Valley" column we've got things to say which are worth hearing. Now it's Investors' Business Daily, for the second time, mentioning us in their computing column. Thanks so much.

Want to know what you can do to help? Remember that it's Journalism -- checking the news, calling people, listening carefully, writing on deadline -- which keeps the clues coming, although I also handle consulting and commercial writing (ask about those rates via e-mail). If you're looking for excellent work -- like that found in my column at Atlanta Computer Currents or my regular work for Net Marketing magazine don't wait for the e-mail -- give me a call at 404-373-7634.

And now back to our show...

The CompuServe Breakup

The collapse of Compuserve had been in the works for years. Their obituary is that of a bureaucracy that could not make decisions quickly enough to suit the fast-moving industry it created.

Who wins? Not who you think. Conventional wisdom has it that America Online now has a monopoly in online services and the consumer end of the Internet market. Its stock price rose sharply on news it acquired Compuserve's subscriber accounts. But wait. AOL access remains unreliable, and business customers who depended on that will cancel. Compuserve customers who disliked AOL before won't change their tunes now. There's also a lot of duplication between the lists. Most of those who'll make the switch are in foreign markets - still a big number, but they won't adapt quickly to its English-language requirements or its American business practices.

The biggest winner is Worldcom , and the biggest advantage for AOL might be sending its access problems to it. Worldcom is best-known right now as the long-distance outfit Michael Jordan endorses. But they also own UU.Net which not only gets the vast, underused CompuServe network, but AOL's ANS backbone access unit as well. That makes it a solid #3 in business Internet access, behind MCI and Sprint, with a better worldwide network than the latter and easier growth prospects than its growing international rival, IBM Global Network. Getting CompuServe Network Services also brings it many top-drawer business Intranet customers it couldn't touch before.

"The way I read it, AOL gets out of the infrastructure game ENTIRELY," notes Jack Rickard, editor of "Boardwatch" magazine. "All of their ports, and all of CompuServe's ports, got to UUNET. Then AOL leases infrastructure from UUNET as needed. In a way, it's a big business windfall for UUNET - they get paid for infrastructure as AOL grows. It's also a big win for AOL - they're out of the hardware business ENTIRELY. They don't have any problems with this - the problems become UUNET's to solve, and AOL pays with checkbook."

But how much is this going to cost me? Before, my $19.95/month was paying for both access and service. With access and content de-linked, as it is on the Web, where does UUNet get its money for linking me to AOL? (There's a five year agreement on access pricing between Worldcom and AOL in this deal, but look at the fine print. Can Worldcom get out, and what are the implications if it can't?) Isn't UUNet head John Sidgmore insistent that "all you can eat" pricing must end for the industry to move ahead? All valid questions, all (so far) unanswered.

Who will win in the end? If it can play its hand right, ATT. It's got a solid franchise in U.S. consumer online access through its WorldNet unit. It still has plenty of cash to build its worldwide network. Through previous agreement, it's also set a low, low access price for AOL members. And it has one less rival. With MCI still wrapped-up in its marriage to BT, with Sprint strapped for cash and unable to merge, this may be the Death Star's last chance.

Don't Spam

Until the rules are clear and users accept it, here's a Clue for anyone thinking of sending commercial e-mails - don't.

Experian, formerly TRW, has backed-out of a a consortium called EHI designed to build "opt-in" e-mail lists after Mark Rotenberg of the Electronic Privacy Information Center screamed bloody murder. Credit databases need business' cooperation to keep their cash cows, and they've been rightly sensitive about using any of their data for marketing purposes. You should be, too.

The fact is, even opt-in list managers like PostMaster Direct can get in trouble, fast. Using the wrong address, or delivering an offer a recipient finds distasteful (even if they previously indicated they'd take ads on the subject) can get the service flamed and turn to dust the goodwill of the company buying the list. Some months ago I wrote I was considering using PostMaster Direct to seek-out paid subscribers. But I no longer think I can risk my limited goodwill that way. It's not their fault, but too many people have pooped in the nest - the risk for now is too-great.

Even a reputable message can draw an angry reaction. AT&T Worldnet recently sent me an e-mail asking that I "upgrade" to Internet Explorer from Netscape Navigator. My boiling anger gave me a Clue. People are hyper-sensitive to e-mail solicitations. Even the delivery of technical means for protecting mail servers won't change that immediately. Even the current wave of lawsuits against spammers won't change things quickly. Your e-mail box is something like your bedroom, and you may not even want friends dropping by there unannounced.

Clued-in, Clueless

Clued-in are the new "angel investor" networks like PriCap , run by Chuck Richards of Hanover, NH. "Angels" are first-tier venture capitalists who get new businesses started - there may be as many as 250,000 of them. This is what outfits like Wit Capital , still trying to build a public market for what's essentially venture capital, failed to perceive. Risk capital is different from public capital - it's money that can be lost and must be patient. By building secure, private links between these investors and potential deals, people like Richards build real value, which they extract through membership fees.

Clueless is Experian , formerly TRW, for trying to cross the Spam Highway without looking both ways first. Now it's backed-off, but the stench of the egg drying on its face will remain for some time. Your clue here is to know your business, and stick to it.

A Clue...to Internet Commerce is a weekly publication of @Have Modem, Will Travel. It's sent free to a qualified e-mail list. Like Netscape Navigator, it carries a list price -- $49 per year. (Unlike Netscape, we don't expect you to pay it.)Subscribers can receive either a .txt file or an .htm file. The .htm version features links which become active when online with a browser, or an e-mail package like Netscape 3.0. (Let us know which you prefer.) To take your name off the list, simply write REMOVE as the subject, or content, of a message replying to this post. To request your free copy, write us at Dana Blankenhorn@worldnet.att.net. We're on the Web at www.tbass.com/clue and www.ppn.org/clue.

A Clue...to Internet Commerce -- Copyright @Have Modem, Will Travel and Dana Blankenhorn, 1997