Bing, Yahoo! ads on Michael Jackson results on Google. Click to see full version.

Seen over the weekend: ads for bing and Yahoo! on Google search results for “Michael Jackson”.

The bing ad led to bing’s xRank page for Michael Jackson. The Yahoo! ad bizarrely led to a Yahoo! shopping results page for Michael posters, most of which had nothing to do with Jackson.

The bing ad appeared every time I reloaded the page. The other ads were much less frequent.

Update: AOL has joined the fray with ads for AOL Music, including a pitch to download a “Michael Jackson tribute toolbar” on the landing page.

The new news feed is like watching CNN during a breaking news event: you can watch for hours and hours and only get two or three bits of interesting information amid the endless blather.

It’s a giant step backward and as more people get on Facebook and become more active it’s going to become worse.

Among the issues I have with it:

• Feed items from frequent users drown out feed items from infrequent users.
• Friends seem to be treated equally.
• It rewards spammy applications, such as the quiz applications that seem to pop up every day and apps like PicDoodle.
• It doesn’t eliminate duplicate items. If 10 people post the same item, it’ll be inserted multiple times. As much as I love the Twitter parody by current.tv, I don’t need to see it anymore. This is made worse by the use of URL shorteners that obfuscate the item you’re clicking on.

One of my design philosophies is that you shouldn’t make users do work that computers can do better. Filtering and priortizing is high on that list. The old Facebook news feed algorithm and the current highlights section provided some level of this.

Here are some of the factors to consider when prioritizing:

• Degree of interaction. Items from people I interact with regularly should be prioritized higher.
• Number of friends involved. The greater the number of friends involved, the higher the priority.
• Number of times shared. The more of my friends that have shared it, the more likely it is to be of interest.
• Location. I’m more interested in things happening near me than on the other side of the country. Facebook will need to become aware of locations that are being embedded by applications such as Brightkite.
• Posting frequency. If someone rarely posts on Facebook, the odds are good that when they do post, it’s something important.
• Application usage. If I use the same app, I’m more likely to be interested in the content that the app generates.
• Topic similarity. If the item is about a topic that I frequently post about, it should get a boost.
• Been there, done that. If I’ve seen it already, it should be downweighted.

The ideal feed would adapt to visit frequency. Someone who visits every five minutes would see a feed very similar to today’s feed. Someone who visits once a week, would see a “best of” from the week.

Some of these things are harder to do than others, but any sort of this filtering is better than what I see today.Any algorithm will undoubtedly miss something that I care about, but the current endless river of unfiltered content ensures that.

Besides, there’s a trick I use when I want to bring someone’s attention to something: instead of hoping that they see it on my feed, I message them directly.

More on: Facebook

Like a recent piece from Henry Blodget, mass inflator of the Web 1.0 bubble. He is at it again with a piece on Facebook being a Google Killer. He points to RBC Capital Markets analyst Ross Sandler’s “analysis” of Facebook’s incredible growth and comScore data on entries and exits.

This is the kind of incessant hyping that inflated the housing bubble we’re all suffering through now — assuming that extreme rates of growth will continue.

The 1427% growth cited for Facebook starts from an insignificant base. With Google’s 468MM uniques in 2006, the only way for Google to have grown 1427% would be to reach every man, woman and child on earth. And it certainly couldn’t sustain that growth into the future, even if a lot of couples got really busy really fast.

Blodget also points to comScore’s entry/exit data to bolster his case. Here, he falsely equates correlation with causation. comScore’s entry/exit report doesn’t necessarily mean that site A drove traffic to site B. It just means that after someone went to site A, they went to site B.

If you go from Facebook to Google, it counts as an exit from Facebook and an entry to Google. It doesn’t matter whether you clicked on a link in Facebook to go to Google or not. You just happend to do those two things. Given that a lot of people use both Google and Facebook, any big site will show up on both entry/exit reports for any site.

Blodget says:

Fully 19% of Google sessions now come from Facebook, up from 9% a year ago.  At the very least, this will likely give Facebook the leverage to negotiate a sweet referral deal at some point.

Nope. Those people are going to Google anyway, without any prodding from Facebook. Google would be stupid to pay for that traffic.

comScore’s entry/exit report is one of the most useless reports they generate and really difficult to interpret. The only real curiosity in the Facebook data is this: 6MM people go to Friendster after they go to Facebook?

Yet another issue with RBC’s graph is that it doesn’t take into account duplicated reach. The combination of Google and Facebook is not 99% of worldwide uniques, because there is a high degree of overlap between the two sites. RBC analysts evidently don’t know how to use the unduplicated reach feature of comScore’s reporting tools.

That’s three huge flaws in one report. Sadly, that’s not uncommon. Analysts and journalists frequently ignore methodology while chasing killer headlines.

Thanks to @carolalene for the pointer on the comic.

In this part, I’ll focus on why the conversational nature of Twitter makes searching it effectively a hard problem.

Consider this exchange:

@CherylHaas: Celebrating my newly purchased iPhone. w00t!!! No longer a Luddite. App suggestions, please?
@rakeshlobster: yelp and shazam and Facebook

This is how people interact on Twitter. Partly because we’re lazy, partly because a lot of the interaction is done from mobile devices where typing is hard and partly because of the 140 character limit on tweets.

Between these two tweets, we have an answer to the query “iPhone app”. But Twitter Search treats these tweets independently. As a result, if you search for “iPhone app”, you’d get Cheryl’s question. Not very helpful.

If you search for “shazam,” you’ll get back my response. But there’s no context for it. The meaning of my response is lost without the context of Cheryl’s question. The question could have been “what apps are causing your iPhone to crash?” This happens in ordinary conversation on Twitter; when people are slow at responding and I get a “@rakeshlobster yes,” I’ll sometimes have forgotten the context.

This problem could be alleviated if Twitter presented threaded conversations. But then Google could just as easily index the conversation, as it does with Yahoo! Answers.

Another issue is that people don’t write for Twitter the way they write for search engines. Compare my tweet above with this post I wrote on my favorite iPhone applications. That was written with searchability in mind. There’s also a lot of shorthand on Twitter. @maryvale shortened “Nikon D80″ to “D80″ in her tweet discussing my last blog post.

That may change if searching Twitter takes off, but it would also change the nature of Twitter. I’ve been experimenting with adding more keywords in my tweets. For example, when I dropped my laptop, I originally wrote:

“laptop hinge broken. argh. it’s pretty, sleek and light. and extremely delicate.”

But then I added in the “toshiba portege r500 is”. It’s more searchable, but it makes the conversation sound stilted and robotic.

Another challenge with searching Twitter for information is that a lot of the value in Twitter is not in the tweets, but in what the tweets point too. With the extensive of URL shorteners like TinyURL and bit.ly, even the minimal keywords are lost.

Beyond the content difficulties in search, there are the related issues of search order and authority.

The results that you get back are sorted chronologically and are highly dependent on when you search. Although the “best” answer for a search can fluctuate over time (one of my criticisms of Google is that its algorithms don’t do enough to counter the effects of Web rot), for most searches it doesn’t vary dramatically over the course of a day or a week. A notable exception would be queries like “what’s a good party at SXSW right now?”

As with asking questions of the Twitterverse, searching Twitter doesn’t provide any guidance as to whose answers are better than others. Searching Twitter is in someways like stepping back 15 years in search technology, before search engines widely used off-page clues and link authority to rank results.

Some suggestions have revolved around developing authority rankings based on number of followers, number of tweets, etc. The problem with that is that no one person is an authority on everything. A search result from Om Malik (@Om) on telecom should be ranked much higher than a result from Om on migration patterns of birds in Africa. Review sites like Amazon and Yelp have devoted a lot of energy to helping people determine which results are valuable. Twitter will have to develop something similar.

Despite today’s issues, the immense amount of data that Twitter and Facebook are collecting could be used to build a better, more spam-resistant search engine. The marriage of search and social networks has the potential to get us better and more credible answers, while also increasing our connections to our friends.

More on: Twitter, Google

Disclosure: I worked with several members of Twitter’s search team at AOL Search. While I don’t believe in the current hype in the blogosphere about Twitter as a Google killer with the current technology, the guys I know are very smart and I look forward to seeing what they do next.

There are two scenarios that have been talked about with Twitter and search: using Twitter to ask questions of the Twitterverse and using Twitter search to search Tweets.

In the first scenario, you send out a Tweet looking for information. An example is a Tweet by Om Malik on Feb. 2 at 7:02 p.m. “suggestion for great Indian restaurant in or around Palo alto. needs to be authentic”. Within a few minutes, he got a bunch of responses. (I recommended Amber India in Mountain View, which was a frequent recommendation.) By 8:19 p.m. Om was “eating at amber India in mountain view.”

Wow! Send out a query and you can get answers from real people right away. Who wouldn’t want that?

Not so fast.

This reminds me of latenight TV commercials for miracle diet drugs. You see pictures of people who have: lost 75 pounds in 8 weeks! lost 10 pounds overnight! gone from a size 24 to a size 6 while eating cake and sitting on the couch! You usually see a line of fine print that says “results not typical.”

That’s exactly the case here. You get atypical results when you have tens of thousands of followers as Om (23,000+) and Battelle (11,000+) do.

I posted a Tweet yesterday “Looking for good wineries in napa. Focus on ambiance and red wines.” To give the Tweet extra chance of success, I posted it using Twinkle, an app that adds a location layer to Twitter. Given the nature of my query, my friend network and location in the Bay Area, I expected success. At 200+ followers, my follower count is well above the median for Twitter.

Three hours later, I got one response from a friend. By that time I’d already decided which wineries I was going to visit. Hardly a Google killer.

Suppose for a moment that I just picked a tough query. What if I’d gotten a dozen responses?

Then the problem becomes how I decide which of those responses are better than the others. Many queries have qualitative components: “What’s a good winery in Napa?” “Is the Nikon D80 a good camera?” The value of the answer depends on my needs as well as the expertise of the answerer.

With a typical search result, you have a number of clues as to quality of the answerer. If something appears on the first page of Google, presumably a lot of people have found that resource valuable. If I get a page from dpreview, I can see that they’ve reviewed hundreds of other cameras, so they probably know what they’re talking about.

With Twitter answers, I get limited information about the source and limited content.

In some cases, this is OK. I did a Tweet a while ago wondering if it was sunny at the Beach Chalet in San Francisco. You’d have to be a jerk to lie about the answers, there’s not much expertise required to answer the question and the answer fits within 140 characters.

But a query like “Is the Nikon D80 a good camera?” is tougher. If @maryvale says “yes, absolutely” then that’s all I need. I know her, I love her photography and I know she knows a lot about cameras. That doesn’t hold true for most of my other followers. And it certainly doesn’t hold true for people I don’t know at all. Someone may say “D80 is a piece of crap” because they would never consider anything less than the $2,000+ D700 or because they aren’t very technically savvy. A Tweeted answer doesn’t provide that context.

Part 2: Challenges of searching Twitter

More on: Twitter, Google

Step 1: Figure out if the Borders near my house is open. I called 1-800-555-TELL using my car’s Bluetooth to get the phone number. Tellme connected me to my neighborhood Borders. “Thank you for calling Borders… for our store hours and locations press 1.” FAIL. With more people relying on cell phones and increasing legislation requiring hands-free systems while driving, Touch Tone-only systems need to go away. After 2 1/2 minutes and three full loops of the menu I was finally connected to the store.

Step 2: Arrive at the store and look for a guidebook. No one was behind the information desk. Two computer terminals allowed me to search for a book. The search results include books that are only available online, aren’t yet published and a few that are “likely available in store” in a seemingly random order. FAIL. Gee, wouldn’t you want to sort the books that I can walk out with first? Otherwise, why am I here? The screen says my book is in section “B020,” with no indication of where that it is.

Step 3: I notice on my way out of the store a book in the bargain bin on Sonoma wineries. I figure this is the perfect opportunity to try out SnapTell, an iPhone comparison shopping application. Take a picture of a book, CD or DVD and SnapTell shows you how much that item sells for online. (It’s like Shazam for shopping.)

My first picture wasn’t good enough; I got an error message. Second time was a charm, despite taking a picture of a book that was too wide to fit in the frame. I could scroll through a list of prices from online merchants.

While I was impressed with the image recognition, the data quality needs work. The results included older editions of the book; the 1 cent price shown on the summary screen was for an older edition. Even when the current edition was shown pricing didn’t correlate to the merchant. Clicking on the $1.99 link pictured shown showed a price at the merchant of $6.99.

Data errors like this aren’t unusal in large databases, but I suspect will improve over time.

SnapTell is an interesting tool for research. And with access to location information, they could build a nice database of what people are searching for and from where.

That’s assuming that stores like Borders improve service to the point that I’ve got other reasons to come in than “I can’t wait for shipping.”

Disclosure: I work for Tellme, a company that makes speech recognition systems for many large companies. I have a bias against poorly implemented telephone systems.

We launched one of those today at Tellme. Dash users who call 1-800-555-TELL or 1-800-CALL-411 and do a business search will have the business listing sent automatically to their Dash device. The listing arrives almost instantaneously. It looks something like this when it arrives:

The Dash screen showing a new address sent from Tellme.

A few clicks and your Dash will give you turn-by-turn directions to the business you selected. This solves a real need: using the power of speech to make entering businesses much easier.

We’re doing lots of exciting things in the automotive space. I’ll share them as they launch.

More about this on the Dash blog.

I love the Eagles’ Hotel California.

I flew United from SeaTac today after a full day meeting in Kirkland.

Lipitor has been recalled by the FDA.

California will be hit by a giant earthquake according to USGS.

The Princeville is my favorite hotel in Kauai.

The Eagles played the Jets last night.

DoubleClick is one of the largest display ad networks.

April 15 is tax day, make sure you get to the post office.

CNN, AOL, Mapquest and Entertainment are all part of the same company.

Jimbo Wales is a co-founder of Wikipedia.

• Why Lane Hartwell Popped the ‘Bubble’ Video (WIRED) – The hilarious video by the Richter Scales parodying the Web 2.0 bubble to the tune of “We Didn’t Start the Fire” disappeared from the Web after photographer Lane Hartwell filed a takedown request with YouTube. A picture she took of Valleywag’s Owen Thomas was used for a fraction of a second in the parody video. Although she’s gotten a lot of flack for it, it’s hard to fault her for protecting her rights as a photographer.
  The issue also brings up the challenge that the Web and amateurs pose for people like Hartwell. It’s easy to steal commercial content and it’s easy to find freely usable “good enough” amateur content. When flickr makes it so easy to find pictures that you are free to use, why go through the hassle of improperly using a commercial image? (The picture in this post is a Creative Commons image from Telstar Logistics.)
• Amazon Ordered to End Free Delivery on Books in France (IDG News Service) – The French high court rules that Amazon is selling books too cheaply because free shipping constitutes an illegal discount. Under French law, booksellers can’t discount a book by more than 5% off the list price. (Discounts of 50% on bestsellers aren’t uncommon in the United States.) via Dave Smith
• Search: 2010 – A Review (WebProNews) – A look at the future of search with Marissa Mayer from Google, Larry Cornett from Yahoo!, Justin Osmer from Microsoft and Daniel Read from Ask. More of the usual stuff. Usability consultant Jakob Nielsen speaks of moving from “relevance” to “usefulness” to evaluate search. Good luck measuring that. via Jim Simmons
• Dodd Challenges Google to Provide Leadership in the Digital World (WIRED) – Presidential Noshot Chris Dodd speaks at the Google campus about providing leadership in the new information driven economy. He chastens Google for their approach to China and encourages them to stand up to governments (including our own) when they seek to trample the rights of their citizens.

• RSS feeds. You can subscribe to feeds of all reviews near you. If you find a reviewer you like, you can stay up-to-date on his or her reviews.
• A “first reviewed by” designation to highlight contributors who are the first to review a place.
• Attribute drill down. You can narrow your search using filters such as “family friendly,” “casual” or “elegant.”

It’s been a few months since I last checked in on Yahoo! Local. Overall, it’s a huge improvement. It has a ways to go before catching category leader Yelp. (The metric being by my subjective opinion of product quality.)

Yelp has had the first two features for at least a year.

Among the local players, Yelp has had the best incentive system for contributors. Its “First to Review” designation is one of many things that Yelp does to encourage frequent participation. An “Elite” system rewards frequent contributors with a badge on their profile and invitations to parties. The front page of the site highlights a review of the day. Featured Yelpers also appear on the home page.

It may sound corny, but such incentives are important to keeping people engaged. Most social systems have some sort of perk system, including ODP’s edit-alls and metas and the Wikipedia cabal.

Although Yahoo’s design is more visually appealing than it used to be, it’s still cluttered.

Unlike Yelp, the map scrolls off the search results page, making it hard to see where results 3-10 are located unless you have a very large screen.

Getting reviews is more work than it should be. Yahoo! breaks its 69 reviews for The Italian Store across 29 pages, 3 at a time. Yelp shows all 42 of its reviews on one page, making it very easy to scan.

Then there’s the ads. I’m all for ads — I work in the Web space and like to get paid — when they’re relevant. The ads on Yahoo! Local are anything but. Here is an example of the ads that appeared above the listings for restaurants:

The top two ads are for services that compete with Yahoo! Local. Ads on the side (not shown) pitched “Watch mouth-watering videos of Oklahoma’s best restaurants” and one from Target offered “Find restaurant online. Shop & Save at Target.com Today.” (I’ll admit to clicking through on the Oklahoma ad just to see what would constitute a mouth-watering video of Okahoma restaurants. Unfortunately, they linked it to a video of a bad rendition of Rudolph the Red-Nosed Reindeer.)

I understand that local advertisers are scarce, especially outside the Bay Area. But Yelp takes the right approach.

More on: local search, yahoo, yelp

Disclosure: I used to work on local products for AOL.