isabel wang

Amazon reminds me of a Möbius strip

I was telling the marketing director of a large hosting company that if I were him, I'd try to cross sell books from Amazon. If someone signs up to build an online store, don't you think they'd be interested in SEO? Or if they're getting a dedicated server, maybe they'd find the Linux Troubleshooting Bible helpful?

A few days after that conversation, I got an email from Amazon: "we noticed you purchased an ecommerce related book in the past and we thought you might be interested in an exciting new product from Amazon Services, WebStore by Amazon." Wow.

Coming soon: "we noticed you performed a search for Ruby on Rails; at EC2, you can get the entire Nginx/Apache/Mongrel/MySQL stack in one ready-to-deploy appliance." Or: "we noticed there's a disaster recovery book on your wish list. If you sign up for S3, we'll ship it out today at no cost!" Just you wait; it could happen!

So if you're one of the many people who've told me that Amazon is "not a hosting company", you might be right. At the same time, Amazon isn't a bookselling company either. Instead, it's kind of like a Möbius strip, which has one continuous outer surface. Amazon's core competency, I think, is in observing and learning from customer behavior, and using that knowledge to put whatever products people might want right on the counter. 

In contrast, after Demand Media CEO Richard Rosenblatt's HostingCon keynote, I mentioned to someone that 1&1 ought to spin off its domain registration business, merge it with Sedo, Web 2.0-enable Sedo's antiquated domain parking service, and go after the secondary domains market big time. "But 1&1 is a HOSTING COMPANY," my friend said. "They do domain registration only so that they can sell hosting." But what if customers aren't registering domains so they can buy web hosting? I'm not sure you can make a product more relevant by giving it more shelf space.

BTW, I wouldn't be surprised if Amazon got into the domains business. Check out Fabulous.com CEO Dan Warner's "Aftermarket Value Innovation" presentation (PPT), where he discusses contextual domain search. For example, if a customer searches for "chicken risotto", Fabulous will show him available domains relating to recipes, restaurants, food, etc. Now imagine Amazon taking this contextual info and auto-populating each domain with an aStore. aStores-for-domains would feature top selling products from relevant categories (dynamically updated, of course), as well as items from each visitor's wish list and search history.

Some time ago, after Google published the Efficient Frontier/AdWords for Domains case study, many domain portfolio owners complained that Google's revenue sharing algorithm doesn't adequately account for the exceptionally high ROI that type-in traffic apparently delivers. It seems aStore-for-domains might offer better incentive alignment? Remember - Amazon's auto-recommendation technology is supposedly responsible for 30% of its sales. This means aStores can extract significantly more value than pay per click ads out of each direct navigation visitor - and pass on a % of such revenues to domain owners.

Attention-as-a-service?

When I read about Google's electronic billboard patent application last year, I thought it could be combined with Mini Cooper's RFID keyfob concept for an opt-in advertising program. Just as websites track visitors via cookies, Google's billboards would communicate with passers-by's key chains (on which users would store their preferences) and use the data to serve up targeted advertising.

On websites, though, advertisers could be reasonably assured that their messages would be placed within each visitor's field of vision. Not so with real world billboards. But I just read in Wired that a company called xuuk is coming out with...

"a palm-size video camera surrounded by infrared light-emitting diodes. It can record eye contact with 15-degree accuracy at a distance of up to 33 feet. A simple glance from a passerby scores an impression."

This means Google may soon be able to sell you precisely metered amounts of attention. Quick glances might cost X per millisecond, but prolonged eye contact would set you back 5X.

BTW, speaking of billboards, Frank Schilling recently mentioned this YouTube video about  the subliminal power of ambient advertising. If it's actually true (I'm not convinced), billboards could become as big as a business as Adwords.

Imposed vs Emergent Market Research

Hmm... just thought of something kind of obvious.

I came across Andrew McAfee's idea of imposed vs emergent structure on his blog last summer; it totally made sense: once upon a time, Yahoo! tried to organize the web hierarchically, under categories selected by professional taxonomists (imposed/less efficient).  Nowadays, we've got tag clouds on Flickr and del.icio.us, which reflect the aggregated wisdom of the crowd (emergent/more efficient).

Somehow I've always thought of emergence in the context of specifically-defined communities whose participants are explicitly aware of their membership (employees of the same company, users of the same Web 2.0 service, etc) . But James Governor's long-ago post on attestation made me realize that emergence could take place across any buyer groups that a market researcher might focus on: electronics consumers, enterprise CIOs, whatever.

The idea i have in mind looks to the future of declarative metadata collection and aggregation. Much like Larry Lessig’s election week concept “emblogment” - it's an informal endorsement through a tag associated with a user that talks to that users’ preferences. It is not a poll by phone methodology (but how scientific are they anyway?)...  Users will offer public “attestation” - [take the pledge!] - this is who i am! this is what my system is. that is my configuration. these “emblogment” tags–distributed as Javascript say, could then be aggregated–and hey presto, Technorati or some other blog aggregator starts competing with IDC.

James got the idea from his friend Len, who suggested that in the future we'll all maintain personal profiles featuring our product/brand/political preferences. (BTW, Vincent Rais thinks he once came across a now-defunct dating network that consisted of such profiles.) Attestations could transform market research from estimation (ie data collection through imposed methodologies and assumptions) to measurement (aggregation of emergent data).

I can't believe that post was written more than 2 years ago. MySpace had fewer than 10 million accounts back then. YouTube didn't yet exist. Google's share price was <$200. Given all that's happened in our Everything 2.0 world, I'm disappointed that participatory attestation hasn't become more prevalent. Some signs of progress I see are...

1. Stowe Bowd's suggestion that vendors could put up  "beacons" (or standardized tags) for encouraging emblogment. (Have yet to see this picked up by any vendor)

2. Doc Searls' idea that VRM helps CRM. (VRM = vendor relationship management.) Instead of waiting to receive vendors' marketing messages, Doc says buyers should assert their preferences. As an example, he mentions a prototype system for movie reviews where users own and control their ratings. The user-ownership makes the data more valuable because it's not locked into any vendor's silo. Doesn't that sound kind of similar to Len's and James' idea?

3. TrendIQ's "Internet opinion" analysis, which measures the amount of available text on the web for any brand/product/issue you choose. This is not the same as intentional emblogment though. TrendIQ says their sentiment measurements mirrors real world purchasing and voting behavior, but how do they distinguish between real attestations and machine-generated text on splogs, parked domains, etc? As with Technorati, some amount of spam is bound to slip through the cracks. (Of course, as James pointed out, any kind of attestation system would be vulnerable to abuse as well - and it's not as if phone surveys and focus groups have perfect accuracy anyway.)

The question is, will participatory attestation have its day, or will the "tyranny of statistics" takes over? (My friend Patrick isn't too excited about my idea of predictive ecommerce,  whereby Quantcast/MIT's reality miners/other math modeling geniuses can tell you what "someone like you" is supposed to prefer.) In either case, it's about time for emergent market research - don't you think?

Digital Billboards Ad Network: This Could Work

I read about it on ClickZ through John Battelle's Searchblog. A few weeks ago Google filed a patent application for ad space allocation on a "network of electronic display devices". The system it described will allow advertisers to select keywords, limit ad display to preferred billboards, and tie their inventory system to the ad network, so that marketing messages for sold-out products are automatically removed from rotation. Ads on the billboard network might consist of video, audio, printed incentive, "interactive data transfers" and/or combinations thereof.

And earlier this week, I saw on NewTeeVee that Cisco is interested in those electronic display devices, too.  Paul  Kaputska points out that Cisco recently bought a digital signage startup; should they buy a retina scan company to round out their Minority Report-style ad system?

I'm not sure either keywords (how will they be triggered?) or retina scans are the way to go, though. Instead, the Google patent info made me think of this Engadget post from last week about Mini Cooper's RFID billboard.

MINI has begun a pilot advertising campaign in Chicago, New York, Miami, and San Francisco, which gives select Cooper   owners the chance to get an RFID keyfob in the mail.. . Users can select a custom message to be encoded on their RFID chip, and when they cruise near an overhanging MINI billboard, their particular message lights up for the world (or at least nearby motorists) to see.

What if Google (or Cisco, or whoever) sets up an opt-in program that combines RFID keyfobs and CVS' (barcode based, at least for now) ExtraCare loyalty card? Over 50 million customers are active  participants, who allow the pharmacy to track their purchasing behavior in exchange for discounts.

You would sign up for a keyfob online; its RFID chip would be encoded with demographic info, purchasing preferences, etc. Whenever you approach a Minority Report-style billboard, it'd light up with contextual advertising. And if you're interested in the offer, it will give you $10 to spend with a merchant that supports Google checkout.

The Future of Mobile Search

I read about Ms Dewey on Terry Gold's blog. You've *got to* check out this nifty interface for Microsoft's Live Search. It was developed by EVB, whose other projects include this also-must-see gallery of Mini Cooper roof graphics.

But while the site is way cool, Ms Dewey just fetches the same search results you'd get at Live.com. If only she were as interactive as V-Girl from Artificial Life...

A friend forwarded me this New York Times article (PDF) in early 2005. A former professor of artificial intelligence and neural networks, it said, has combined computerized voice synthesis, streaming video and text messaging technologies to create "Vivienne", a 3D virtual girlfriend whom you interact with through your 3G mobile phone. She speaks 6 languages and is able to converse on 35,000 different topics, "from philosophy to movies to sculpture".  She is particularly interested in banking, as Artificial Life tweaked and reused 70,000 sets of questions and answers that it developed for a Swiss bank's knowledge base. Her attention and affection cost $6/month, but the price doesn't include airtime - or the virtual flowers and chocolates she expects to receive from time to time.

Vivienne has since been replaced with a team of 5 V-Girls. They laugh, cry, flirt and argue - but keep your hands to yourself... According to their FAQ, they'll offer companionship, advice and even commitment, but sex is absolutely out of the question. Artificial Life also offers V-Boys and V-Penguins. The characters converse with customers by querying an "expert system" (some kind of natural language knowledge base software similar to what ATG has, maybe?) running on the company's servers.

So - imagine if you combined Artificial Life's expert system with Collarity's social search capability? Instead of inputting search keywords in a mobile browser, you could just make a quick call to your virtual friend. He/she/it could instantly give you relevant results filtered by your past conversations, your mobile  and desktop search history, and behavior data from "people like you"...  If Terry and I both found Ms Dewey riveting, V-Girl Search might consume every bit of everyone's attention.