My second post (of two) is up over at the BingTweets site, part of an FM partnership with Microsoft. In it I describe my frustration with search as it relates to helping me make a complicated decision: How to possibly buy a classic car. From it:
So first, how would I like to decide about my quest to buy a classic car? Well, ideally, I’d have a search application that could automate and process the tedious back and forth required to truly understand what the market looks like. After all, if I’m looking for classic Camaro or Porsche convertibles from the mid to late 1960s, there are only so many of them for sale, and they can be categorized by any number of important variables – price, model, region, color, features, etc. And while a number of sites do a fair job with a portion of the market, I don’t trust any of them to give me a general overview of what’s really out there. That’s where an intelligent search agent can really help.
But the next step is the harder one. I am not “smart” about how to buy a classic car. I don’t know enough to buy one with confidence. I don’t know what to ask about. I don’t know if it’s good or bad that an engine, electrical system, or transmission is original or rebuilt. I don’t know how one model does versus another in resale value, or insurance cost or…well, you get the picture. There’s a lot to consider, and I don’t know how to value everything. The world of classic cars is complex, like most major decisions. In short, there’s no easy way to decide in this case (unless, of course, I could just buy the most expensive one. That usually guarantees you’ve gotten what the market thinks you paid for it. Not an option for most of us).
So what do I need? I need help from a human being – someone I trust who has command of the classic car domain *and* has my best interests at heart. But given that I don’t have a spare Uncle who happens to be a classic car nut, what am I to do?
16 thoughts on “Search Frustration: It’s Still Hit Or Miss On Complex Decisions”
How about using a car site instead of a search engine?
Keyword search is not equipped to answer complex questions. It’s just a pointing mechanism. Use it that way and stop being so demanding!
Well, the next best thing is to find like minded folks who have been where I am about to venture. Wouldn’t it be great to learn from folks who have already done exactly what I’m in the process of doing – a sort of gallery of saved research? As far as I know, no such gallery exists (the idea is at the core of the Memex), but there are an awful lot of question answering services that are attempting to at least build a foundation toward that goal (ie Answers.com, Yahoo! Answers, etc.)
A “true” decision engine as you are describing, I imagine is the next step in search. Engineers have somewhat pushed the limits with keyword search, as you noted, all that is missing is context.
I think “real-time” search is one step closer to that kind of search engine. Rather than accessing saved researches, their will be a large amount of data, both hard facts, and common sense information (thanks to social networks) that is “live” allowing engineers to build engines that can not only understand better, but seek to understand better. There will be little need to index.
Entering “Classic Car” into a search box is like asking a Magic 8 Ball.
The problem is that your intent and your requirements can’t be deduced from two words.
But if you put in a few more words you will get better results.
The bigger question though is what makes you believe that the information you need is accessible on the web?
Yes, good analysis. But it is working fine. I dont understand.
I am like you i don’t know so much about sport cars, so in order to educate myself i ask to Google the answe i search. Is work.c
You write: …as soon as you go a bit down the tail – like my example for classic cars – search becomes a pivot point for an ongoing and often taxing decision process. The opportunity, I think, is to figure out a way to support that process down the tail – saving us time, clicks, and frustration along the way. I see two paths toward that goal: one is creating applications on top of “ten blue links” which help me organize and aggregate the knowledge I process while pursuing a search query
Let me remind you of a January 2007 (almost 3 years ago now) comment thread on your blog. Here are a few of the key comments from that thread, though I would recommend reading the whole thing:
Explicit relevance feedback is a known way, a known “application” for helping the user self-personalize their way down into the long tail of information they might not have found any other way.
See also some of my comments from May and September of 2008:
You also write: There’s not a definitive answer that scales to everyone who might have the question. What’s right for me is not right for you or the next fellow
Information retrieval researchers are well aware of this, and (as I mentioned above) started developing techniques over thirty years ago that still aren’t being used on the web to address this problem.
My question to you, John, is: Why not?
My tentative answer is: Monetization.
When you are deep-engaging in your “classic cars” learning session, are you in a frame of mind to click an advertisement for a classic car, a used classic car engine, or whatever they want to sell you? Probably not. You’re trying to learn about what makes a good classic car. Not actually yet buy one.
Actually buying the car, rather than learning about how to buy, is a different set of searches that you’ll be doing, at a different stage in the process.
It’s true you might click a few of the ads, to give yourself an idea about how much various cars or engines cost. But from the advertiser’s perspective, that’s a wasted, zero-ROI click, something that the advertiser will wise-up to. But again, because you’re in the early, learning stage of your information need, advertising is pretty worthless to you — it’s not really relevant to your task.
Let’s suppose this information gathering stage lasts days or even weeks, and you do hundreds and hundreds of queries or query-like interactive applications built on search. How much does that cost the search engine, in terms of processors, researchers, engineers, electricity, bandwidth, etc? A lot. Much more than a single, home page lookup search.
And at the end of the day, when you finally learn everything that you need to learn, in order to begin the purchase phase of your task (buying the car), you might start clicking ads again. But how much money is the search engine going to make on the couple of ads that you click? A few bucks?
I can’t help but conclude that the reason why we don’t see more of this decision-oriented search on the web is because of advertising-supported business models. If you make 97% of your income on search advertising because you’ve trained users to believe that search is a “fact lookup” engine, does it make sense to start to train users to think of search as a learning and exploration engine? No. Because then they’ll use the engine a lot more, relative to the amount of ads that they click. And that’s bad business.
Bing can get away with it for a while, because they’re heavily subsidized by Office and Windows. But Google? Will they ever go this route? They haven’t shown any significant signs of it over the past decade. I have to imagine that it’s because of their business model. It’s good for advertisers, but bad for searchers.
Or am I completely wrong in my analysis? I’ve been seeking an answer to this for years now, and have never seen a convincing counter argument.
Yes JG, you are thinking true and you say good idea.
Why do I have the strange feeling that this post is a prelude to a piece about Aardvark?
A bit off topic but I know this is also an area of interest to you and since today coincides with the start of the O’Reilly Gov 2.0 event in D.C. (wish I could’ve been there!)…
Comment from George Dyson on “Economics Is Not a Natural Science” By Douglas Rushkoff
“How to best transcend the current economic mess? Put Jeff Bezos, Pierre Omidyar, Elon Musk, Tim O’Reilly, Larry Page, Sergey Brin, Nathan Myhrvold, and Danny Hillis in a room somewhere and don’t let them out until they have framed a new, massively-distributed financial system, founded on sound, open, peer-to-peer principles, from the start. And don’t call it a bank. Launch a new financial medium that is as open, scale-free, universally accessible, self-improving, and non-proprietary as the Internet, and leave the 13th century behind.
In essence, I agree with the piece and the comment.
I also believe that the Chagora structure and design is very close to what they are talking about… (it can function with standard and/or newly created currencies whether localized or not. Especially when combined with methods of geographical localization with scaled anonymity.
The practical microtransaction in all areas is essential for proper scaling of civilization and its the political microtransaction (networked citizen lobbying) that is the trigger.
P.S. PayPal is a bad model for civilization development.
(Chagora is essentially scalable speech and association)
If I’m an idiot I’d like to find out soon since things are very tight. I’d like at least a chance to present my case and don’t know where else to go.
If you’re looking for serious information about classic cars, then why not go where the classic car enthusiasts are? CR4: The Engineer’s Place for News and Discussion aggregates content from Hemmings Motor News, CarDomain, and MOTORZ TV. CR4 provides search capabilities, of course, but you can also ask a question using the Automotive Forum. (Even better, a real human being will answer it.)
Here’s a link to CR4’s Automotive Section.
The people who own the information you seek, which is probably born from general research as well as individual first hand experience, can profit modestly by selling it. Why would you expect to find it for free? Because information wants to be free?
Sometimes, too, the information you seek is available in places that seem off-the-beaten path (but hopefully not too far below the fold).
If you were looking for information about the 1967 Shelby GT500, for example, a lens builder named ThomasC can help.
No, Thomas C isn’t an optometrist. Rather, he’s a master of Squidoo – a community website that allows users to create pages (called lenses) for subjects of interest.
Squidoo was founded in part by Seth Godin, author of marketing tracts such as “Purple Cow” and “Meatball Sundae”.
This is a similar problem to a lot of knowledge work. Complex knowledge work does not have a right answer. Instead it’s judged on various signals. See: Evaluating Performance of Concept Workers
And the answer is that you need to do some basic research to figure out some questions and to figure out where to find people, and the reach out to people with expertise. The beauty is that it’s incredibly easy to get to people with significantly greater expertise and they are often quite willing to help. See also: Networks and Communities
Of course, most of us learned (in school) that talking to others is cheating. So, there’s a lot of learning to be done around these work literacies.
I am so far behind in this bingtweets thing….partly because I don’t understand the power of it or the reason to use it in the first place….I wish someone would explain why people use twitter in more detail and give me an actual example of how it benefits them….I understand facebook and myspace but this really has me frustrated.
The real issue here is that this is structured data, which is hard to scrape from a crawl. If the sellers published the information in structured form via a data API, you might have a chance of building the query you want.