July 31, 2005
What do Golf, Startups, and Slots have in common?
They all conform to one of the fundamental rules of human behavior conditioning - provide random and variable rewards to reinforce repeated behavior.
Ivan Pavlov in a series of seminal experiments in the 1900's established all of this and now the gambling industry uses the same principles to "hook" and "keep" people. As Alan Krigman wrote in this post:
Nearly everybody knows of the experiments by Ivan Pavlov in the early 1900s. Pavlov started by ringing a bell and presenting food simultaneously. When he stopped offering the food, the dogs still salivated at the sound of the bell. Later, to keep the dogs responding to the bell, he reinforced the behavior by accompanying it with food at random, but frequently enough that the stimulus was not futile too often. B F Skinner extended this work during the 1930s using conditioning to train animals such as pigeons and rats to perform various tasks. Tom Creed, a psychology professor at the College of St Benedict at St John's University, says that slot machines utilize similar conditioning processes to produce player behavior characteristics desired by the casinos.
Like slots, Golf and Startups provide the same conditioning response (though the only good news is that unlike slots in which the odds are conditionally designed for you to lose, Golf and startups provide opportunities for truly skilled players to hit the jackpot). How often is it that you play golf and go through 16 holes of hell but remember that one great shot in hole 17 in which you drive within 6 feet of the hole and one putt it for a birdie - in my case very often - and it is that one shot that makes you want to come out and play golf again.
In Silicon Valley, startups have some of the same conditioning allure of the slots. Slots are designed to be "loud" and "blaring" when there is a huge payout so that all the other players around the slot realize that someone has won and thus get reinforced that they could also win. This motivates the other slot players to play more often. The same thing happens with startups - think Google. In Slots, the machines are designed to have many "close" outcomes to trick players into thinking that they almost hit the jackpot. The same thing happens in startups (not by design but just by their nature) when the startup is close to being acquired or when they get a major deal. Slots make periodic payouts that condition players to keep playing. Startups get acquired or go public in quick enough time frames that they resemble "payouts" to the employees and thus motivate them to start or join another startup.
Now please don't mistake this post to mean that startups are a "con" like slots. Far from it. Startups offer a truly unique opportunity for people to pursue their vision to create fantastic products/services, delight customers and make everyone wealthy. Unlike slots which are destructive for everyone (except the casinos of course), Startups actually benefit society. I draw the parallel only to show why startups are addictive to some folks. They share the same conditioning response with slots and golf. Hopefully my wife now understands why I still play that god forsaken game of golf.
May 22, 2004
Ars Technica links to an interesting research project (KnowItAll)at the U of Washington aimed at developing a search tool for extracting large collections of facts from the web. If you wanted to know about used car dealers in San Francisco, for instance, the tool would generate a list of dealers for you. The search engine uses an awareness of syntactic patterns to extract information from phrases such as "used-car dealers including XXX". DARPA and Google are reported to be funding the project.
From the abstract of an overview paper on the project's website:
Manually querying search engines in order to accumulate a large body of factual information is a tedious, error-prone process of piecemeal search. Search engines retrieve and rank potentially relevant documents for human perusal, but do not extract facts, assess confidence, or use information from multiple documents. This paper introduces KNOWITALL, a system that aims to automate the tedious process of extracting large collections of facts from the web in an autonomous, domain-independent, and scalable manner.
December 11, 2003
eBay to capitalize on online auction data
Two recent stories (Reuters, Washington Post) report that eBay is considering licensing its online auction data. The articles note that Ebay has been quietly experimenting with about 20 customers who pay $10K annually for access to the data and that it plans to focus on its data business in the coming year.
Example applications (from the articles):
* PGA.com, on Friday launched a used golf equipment value guide based on raw data from eBay's golf club and accessory sales.
* Intuit Inc.'s Its Deductible division, depends on eBay information to help users of its TurboTax tax preparation software to figure the fair market value of their donated items -- from used jeans to hard-to-value gadgets like cell phones.
* Andale Inc. uses eBay data in its Sales Analyzer software that helps eBay sellers price their items most effectively. "Andale Inc., which uses eBay's data in a research service it started selling over the summer for $2.95 a month. Andale Research lets people compare pricing trends on specific items, such as a 10-gigabyte Apple iPod music player or the Viewsonic P810, a 21-inch CRT computer monitor. Typing the Viewsonic model number into Andale's research box tells you that its average selling price has dropped in half since January, to $100, while the number of units offered each month has fallen from 549 to 52. At the same time, the value of 10-gigabyte iPods has held relatively stable on eBay, averaging $237. Tens of thousands of people are paying Andale to use the new research service, said chief executive Munjal Shah. The data show how many items sold at different price levels, correlated with days of the week, time of day and various merchandising factors."
* tapping eBay's auction prices to help appraisers value insured items when they're lost or stolen
The Post article places this eBay initiative in the broader context of Internet companies (such as Amazon and Google with their web services APIs) licensing their databases to developers and customers as a means of tapping into innovation and reaching more customers. Munjal Shah (CEO of Andale) draws an interesting analogy:
Shah thinks we are witnessing the birth of data analytic services that will piggyback on eBay much the way Bloomberg News and other data providers piggybacked on the New York Stock Exchange. But we're just starting to set the kind of standards needed to allow smart retail data analysis, Shah said.
"It's still like the New York Stock Exchange was back in 1910 or even earlier," he said. "Imagine a stock market with none of the tools we use today to trade -- no tracking tools, no P/E ratios, not even stock symbols. Both from a buyer and a seller standpoint, today we are all trading blind."
Umair points out the issues with using eBay auction prices (correcting for winner's curse, etc.) - assuming for a moment that such effects can be corrected for or somehow normalized for, there is potentially a tremendous amount of value that can be unlocked through exposing eBay's transaction data (ideally through open web services APIs made available to developers at no/low cost as Amazon and Google are doing). This would open up opportunities for data analytics software companies that could mine this information. The eBay transaction data would be valuable not just to buyers and sellers on eBay but, independently, to marketing departments in a range of industries (in terms of pricing, demand forecasting, trend forecasting, etc. etc.). Fertile territory.
October 27, 2003
Camera phones as bar-code scanners
Sean Neville writes about an interesting camera phone application - the idea is to take a snapshot of the barcode, perform OCR on the image to extract the UPC, do a query on Amazon/Froogle and return product info and pricing.
In the category of interesting possibilities that open up at the intersection of technology trends/developments: in this case (1) ubiquitous wireless data access, (2) networked cameras, (3) OCR, (4) Amazon web services. As Neville eloquently puts it, this is about using a camera phone as a bar code scanner "to transform all physical goods into mere floor demos." Also file this under the category of innovative instances of technology reuse.
Elsewhere there was an interesting discussion a while back re: "the digital camera as a data-gathering device":
Because digital photography is so damn convenient, image capture becomes convenient for a variety [of] recording activities that hitherto were simply too clumsy to do with [a] film-based camera.
According to their press release, NeoMedia have combined one winning technology: Amazon's Associate program (which lets business access the Amazon catalog database), and one failed technology: the CueCat (where bar codes in advertisements would lead people to websites) into one interesting m-commerce application.
The shopper works by using your cameraphone to take a picture of a book's ISBN number (by using the bar code). Then using a proprietary application, the picture is sent to NeoMedia, who will use the bar code to determine the ISBN number and send you Amazon's price for that book. No mention is made of whether you will be able to purchase the book from your handset.
October 25, 2003
Amazon's Search Inside The Book
I've been waiting for this for a while. Amazon announced today that you can now do full text searches of nearly 120,000 books in the Amazon catalog.
Starting today, you can find books at Amazon.com based on every word inside them, not just on matches to author or title keywords. Search Inside the Book -- the name of this feature -- searches the complete inside text of more than 120,000 books -- all 33 million pages of them.
Wired has an article on this announcement that touches on a couple of the implications of this move from Amazon. It also ranges wider - drawing connections to ambitious attempts like Brewster Kahle's Internet Archive and the Gutenberg Project and commenting on why this move is strategic for Amazon:
This shifts power away from the people who own finite sets of copyrighted material and toward the people who offer access to information about where this material can be found. Information about books, not ownership of copyrights, becomes a new center of power...Amazon's Search Inside the Book is not an ebook project. It is merely a catalog. But a decade of Internet history proves that the catalog is exactly what you want to own.
The contents of books may be the only publicly accessible data set with the potential to match Google's Web index both for size and utility. Search Inside the Book makes Amazon the sole guide to tens and ultimately hundreds of millions of pages of information. And while Google's business is vulnerable to any competitor that builds a better search engine, Amazon's book archive is the product of negotiated contracts with hundreds of publishers. Amazon has cornered the market on information that was once hidden away in books. The burden of the physical – the fact that the database Amazon uses is linked into a complex system involving real things – gives it a stunning, if perhaps temporary, advantage.
October 18, 2003
Just found out about this cool new service via WIFLblog.
JungleScan.com is a free service that
- Scans products on Amazon.com to record their sales ranking over time
- Allows you to start or participate in discussions about any website or Amazon.com item
Authors and journalists take note. Here is a way to track the Amazon sales ranking of a book or product over time. One can follow the ranking of your own novel, or CD, or you can collectively track the rise and fall of an idea, or group of items. I still haven't figured out why Amazon itself does not offer this service since they could do it so much better (all that data, and world-class skill in interface design), but in the alternative this cool free website does a good job.
For instance, to see how Clayton Christensen's new book (which I just ordered) is doing, see here.
October 14, 2003
Gizmodo links to an interesting Business Week article on passive radar systems. These radar systems can detect and monitor moving objects (cars, planes, etc.) over hundreds of miles using the reflections of cellular, FM and TV broadcasts off their sides.
Passive or transmission-less radar has the following advantages over conventional radar systems: (1) it is impossible to detect, (2) the radar device can be very cheap since it does not need to incorporate a transmitter, (3) it does not require a license to transmit, (4) it can leverage the high-power, wide-coverage radio transmissions that are already occuring (FM, TV, cellular) as well as the network of already-deployed base-stations and towers.
Applications include surveillance, air- and ground-traffic monitoring, ground-based air defence and airborne applications. The article cites Celldar and Lockheed Martin's Silent Sentry, among others.
Three quotes from the article:
Despite Celldar's military potential, Lloyd predicts the first applications will come in the civilian sector. He says transportation officials are eager to use Celldar to monitor road traffic because it would avoid the expense of installing either sensors in roads or TV cameras overhead. And police cars equipped with Celldar could follow a car driven by a suspected crook or terrorist from a safe distance, without danger of being seen.
Lockheed-Martin's system is dubbed Silent Sentry. Last fall, in a demo for the U.S. Air Force, a third-generation Silent Sentry radar tracked all the air traffic over Washington, D.C., by picking up FM and TV echoes. Because FM and TV transmissions are more powerful than their cell-phone cousins, Silent Sentry can detect planes as far away as 135 miles, roughly 10 times the reach of an individual cell-phone tower.
John D. Sahr, a University of Washington electrical engineer, [...] has operated a passive-radar system unshrouded by military secrecy. It harnesses an FM station's signals to study particles in the ionosphere -- the top layer of the atmosphere, over 300 miles up. Sahr decided to go with passive radar, he says, "because it's incredibly cheap" -- $20,000 vs. $25 million for a comparable active system. "You could probably do an amateur system for under $5,000," Sahr adds. A system for small airports might cost as little as $15,000. That's important because of the 5,280 public airports in the U.S., only about 300 currently have radar.
October 10, 2003
Word of mouth: quantifying buzz
Interesting paper that quantitatively examines word of mouth (WOM) and its relation to future sales.
"[The] success of a product is related to the WOM that it generates. In fact, it is commonly believed that WOM directly impacts sales. It might affect awareness in some cases, or preferences in others. On the other hand, WOM may simply serve as a leading indicator of a product's success."
Whether word-of-mouth is a driver of sales or merely a leading indicator, it is clear that a firm introducing a new product ought to try to measure it. This paper examines two quantitative measures of word-of-mouth: volume and dispersion. Volume is simply the quantity of buzz (e.g, how many Usenet posts with “Blair Witch Project” on the subject line). Dispersion is a measure of how this volume of buzz is dispersed among disparate communities – for instance, if Resident Evil gets talked about only on rec.arts.erotica.milla.jovovich it has low dispersion whereas if it gets picked up by a variety of different user communities it might have a higher dispersion. The measure of dispersion is the familar entropy formula with p_i= proportion of posts on this show that appear in newsgroup i. The paper examines the hypotheses that higher volume and dispersion are correlated to higher future sales of the product.
In the context of this paper, the product category is new TV shows and "sales" are measured by Nielsen’s Ratings. It examines the correlation between word-of-mouth about the show in the online community (Usenet) and Nielsen’s Ratings for the show in subsequent weeks. The key findings are that (1) volume is not consistently associated with higher future sales, (2) higher dispersion is correlated to higher future sales, (3) the effect of dispersion decreases over time (which implies that WOM is particularly important early on in the product lifecycle).
October 07, 2003
Netscreen acquires Neoteris for $245MM
Neoteris is a market leader in the SSL-based VPN space which is projected to grow to $600MM by 2006 (Infonetics Research report). SSL-based VPNs offer a client-free solution to secure remote access by terminating http sessions over SSL at a centrally-located SSL-based VPN appliance. SSL-based VPN solutions are expected to grow to become the dominant mode for secure remote access, the primary advantage over IPSec being the "client-free" nature of the solution. IPSec is expected to continue to be used for secure site-to-site applications (such as connecting branch offices).
October 06, 2003
Book Watch Plus
Book Watch Plus is a cool service that I discovered a while back. It uses the web services APIs of Amazon, Google and weblogs.com to create a powerful new application. It tracks all blogs that have changed in the last several hours, looks for links to Amazon books in the blog postings and creates a running Top Ten lists of books that are getting a lot of buzz in the blogging community. Using the Amazon and Google web services APIs, it pulls down product information, related news, etc.
Weblogs.com publishes a list of weblogs that have been updated in the last three hours - this list is published as XML and is queryable using XML-RPC or SOAP. Amazon's web services API allows for detailed product information to be pulled down using XML/HTTP or SOAP. Google allows searches to be performed using a SOAP API.