Saturday, June 02, 2007
Peter Hershberg offers a nice reminder to Jason Calacanis' Mahalo that there might have been a few other search plays in history that used an "energetic group of Guides." In addition to About.com and Ask Jeeves (nice catch there, Peter and Danny), I'd cite, er, Go Guides, this list of mostly-defunct "Expert Sites," and of course the former Zeal community. Not to mention Squidoo. Yahoo Answers. Digg, et al. People, people, everywhere; and in some cases, unique technology to leverage their contributions.
Humans do scale, of course, in the sense that there are billions of user behavior decisions every month, and a smaller universe of editorial judgments being made. From PageRank to Wikipedia to Usenet to Slashdot to the Yahoo Directory, search engines, vertical communities, and widely-based human-edited web plays [see Traffick article "Are These Verticals Too Horizontal? The Slow Death of Mega-Guide Sites] have always tried to leverage editorial judgments, communities of meaning, and the value of expertise and passions.
I think a really good question for any startup to ask today has to be: in spirit, how much more sophisticated or useful is your plan than Jerry Yang's collection of favorite links in 1994?
Mahalo will become part of a huge trend that's been ongoing since the dawn of web search tools and web directories. Is it any better or even as good as the many mentioned above? If it turns out to be, it'll be because it reached a critical mass of users, and found better ways to ferret out the problem of "smuggled spam" -- the gradual deterioration in editorial standards that happens to unrigorously-edited web properties in a world that respects editorial integrity so little that Pay-Per-Post is seen as mainstream. If all goes well, it will succeed in some verticals only because of their uncommon quality.
And that's no different than many that have come before. An open web platform allows quality stuff to get out there, whether or not it's found on a particular secondary layer that purports to do a better job of sorting it. Mahalo might become as well known as LookSmart, or The Drudge Report. Either way, as Hershberg argues, it's going to be scrapping hard over 1% market share.
If there's any takeaway from the launch of Mahalo, it's a reminder that without any humans at all exercising editorial judgments but also judgments on how to structure the look and feel of results pages, you get a jumbled mess in response to a search query. Google and other leading search engines a combination of user experience producers and algorithmic methods of determining query intent. One of the best things about companies like AOL and MSN (and in its unique way, Yahoo) for the mainstream user, was always the sense of "consumer editorial responsibility" on common queries. Mahalo is a reminder to these companies that they should be actively recruiting editorial personnel, and continuing to heavily produce the portions of useful search query result pages on popular queries that are family friendly, consumer friendly, educational, useful, etc. More packaged answer sets, less jumble and clutter, is a great way to stand out in the subset of society that prefers these.
Huge opportunity: such results sets are also mobile-friendly. More on this later...
Perhaps then, when Ask allowed the "algorithm to kill Jeeves," it zigged when it should have zagged. If you're going to be scrapping over 1% market share, with the potential for growth if you hit a real nerve out there, why not go out in style? If I were Mahalo, in addition to working on technological innovation and community-building, I'd use at least some of the lavish funding to attract marquee writers and journalists for high-profile editorial oversight. This would give them a chance to beat the NYT-owned About.com at its own game.
Labels: ask jeeves, mahalo, odp, search engines, yahoo
Tuesday, May 01, 2007
Thanks to Per Koch of Pandia for letting me share them. Per and I go way back, virtually speaking. Our meeting at the frenetic SES New York was all too rushed.
Labels: marketing, search engines
Sunday, April 15, 2007
Is AOL important? Well, not in search, certainly; especially not outside the United States and Canada. Quite a few numbers have been bouncing around lately in terms of search share and online advertising generally.
For those who take a non-North-American, yet search-centric perspective, the following share numbers reported by Richard Zwicky at Enquisite should interest you:
Search Market Share (Global, Excluding Canada & US):
Google Images 6.9%
That's not to say that the US market isn't the largest and most important, but a wakeup call nonetheless. Enquisite isn't a service whose numbers get reported along with Hitwise and Netratings every month, but the sample size isn't small, either.
That global Microsoft footprint appears to be a myth, also. Surprising that their many partnerships and advantages haven't translated into search engine usage.
Labels: metrics, search engines
Wednesday, March 28, 2007
Barry at SEL reports: MSN UK put a prefill in a search box, touting The Apprentice on BBC. Ouch!
Lately I've been in on quite a few, shall we say, monetization strategy meetings. Those newest to the 'net often think you can "stick in" advertising not just in ad slots, but by performing a takeover of common navigation devices - let's make this search turn into a search for a certain vendor's products... you name it. I think users disagree!
Then again, I wonder how far you can push this - and so do many - as evidenced by the MSN experiment.
For example, on the "example queries" published below the search box on a site like Yelp, what if it said, instead of:
"Example: Tacos, hair salon, English pub"
it said: "Example: cafe, Anchor Bar, physiotherapy"
And on a major vertical search site, would an example query using a national advertiser, in that little line, like "Walgreens," be way out of line? What's your take? That's one of the topics I hope Danny and I discuss today in the Daily Searchcast at 11:30 ET.
Labels: search engines, usability
Wednesday, March 21, 2007
- I get a bit unsettled when someone goes up a level of generality in their domain naming. Like, "we managed to snag a more generic domain name than you'd expect - so look at us!! and be prepared to be thrilled for us in six years when we flip that sucker for $1,650!!" The Keele Street Christian Church proudly displays their domain name, www.keelestreet.ca, in the front window. Hey, shouldn't that be keelestreetchristianchurch.ca, or keelestreetchurch.ca? At the very least, make it geographically informative, like keele-and-annette.ca, by putting the cross street in there.
- Topica, the email campaign management service, has a beautiful graphic on its home page. A huge 3-d "2007." Beautiful. Folks, it's March 21.
- It isn't very hard to find great Toronto restaurant reviews, if you know the main sites (Chowhound, Toronto Life, etc.) that purvey reviews. These will generally come up easily in a search. And if you're a good searcher and know the type of food you want, or the neighborhood name (Queen West, for example), you can even kind of do a themed or geographic search, in a way. But as for Google Local and the other blue-chip "local" and "map" based search engines, currently they aren't doing a very thorough job of aggregating this information. No doubt in a couple of years this problem will be solved and the maps and results will be intuitively and comprehensively populated - but for now, knowledge of your favorite sources (Toronto Life, NOW Magazine, Chowhound, etc.) and how they divvy up the city or leverage the community for recommendations, is the only path to foodie satisfaction. In case you were wondering, for dinner tomorrow night I'll be at one of: Parsi Restaurant, Jules, or Czehoski's. Vote now!
Labels: domain names, local search, search engines
Sunday, March 04, 2007
Tom Foreski in SiliconValleyWatcher lists all the ways that we the people are expected to help search engines do their jobs - so search technology isn't pulling its weight as compared with the human element.
Eye-opening at first, but rather than a blow-by-blow response, why not sum it up this way:
More detailed response:
- Should you do extra work to label your content or install sitemaps? Hmm, only if you want to be found.
- We're talking about publishers, not people. Therefore, purveyors of information (and/or products and offers) in a hugely competitive, open environment. Tom's article, for example, sports ads for conferences as well as Edelman, the world's largest PR firm. Seems like a tag or two might be a decent tradeoff for the exposure. You don't expect the engine to actually write the content for you, so what's a bit of extra metadata between friends? As for "people," it's the users that are getting a good deal out of the extra work you might do to label your content
In short, the claim that "people should just find me" is a bit like building an all-graphics site and hoping people will find you when they search for "guitar pick." Or sitting on your back porch strumming "Galveston" and praying you'll be invited onto American Idol.
- Tags or labels are indispensable when it comes to some kinds of content, such as videos or photos
- If something is useful or popular enough, depending on the community, third-party tagging can be helpful. What's the incentive to do this? Interesting question. What's my incentive to type this sentence? But yes I think there is a huge bunch of unlabeled stuff that probably will stay unlabeled because there is no incentive to label it. That doesn't mean search engines aren't going to try to "organize it and make it universally accessible."
- The article's general tone seems to suggest that the search engines are stingy about "sending their robots around." Far from it! Even relatively unpopular sites are spidered frequently nowadays.
- Search engines have advanced in many ways over the past few years. One of them is their sheer storage capacity. Index size is a huge challenge, which brings us to:
- The claim that corporate search engines are doing a better job of letting publishers take the lazy way out is a bit odd. It's a much smaller dataset, so stuff is much easier to find. But I'll grant that it is interesting that some of these technologies are quite good at recognizing industry-specific patterns, and autocategorizing content -- no user tagging required. But that's a whole internal debate in the info retrieval field. I'm sure some companies use human categorization!
- Search is a bit like matchmaking, and the meaning of what "search" is has expanded. Take the emerging field of local search. Now add the premise that "refine is the new search" (I don't think it really is on its own, but users definitely want to be able to "drill down" to get exactly what they want by telling the search engine). And hey, why not toss in the idea of geolocation & mapping. So I'm a user and I'm looking for a hardware store, let's say. Let's say I also want to find a hardware store that sells a certain brand of doorbell. I'd prefer it be within 20 minutes driving distance. And I want to find one that is "open 24 hrs." (just for argument's sake). None of that is ever going to be findable without a huge amount of research, unless of course the "publisher" (hardware store owners) is willing to upload their information in a structured format. By uploading that info, buyer and seller connect more easily. By not uploading it, you choose "not to be on the map." It's your choice.
- Things like Google Base are arguably research projects to help Google find out what are some common categorization schemas in a given industry - or a whole category, like brick and mortar retail. (If "open 24 hrs." is a common one, then maybe it'll come up more often in search and navigation databases as a yes/no item down the road, let's say.)
Labels: galveston, google base, local search, mapping, metadata, search engines, sitemaps
View Posts by Category