Google Search today is repeating the same mistakes I remember from the late 90s.
For the first decade of the web, there were a handful of search engines competing, rising and falling in popularity. The best were altavista, and fast.
One thing that was noticeable back then was that bad search engines (and search engines that 'jumped the shark' and became bad) generally did so in similar ways:
a) they included paid results, or devoted too much real-estate to advertising
b) when they failed to find results, they tried to trick the user by showing related results (eg: omitting or substituting terms)
c) they avoided 'logical and' for search terms, in favor of 'logical or', making it difficult for users to search with precision.
The people at Google surely believe their recent changes have nothing to do with all that. Far as I'm concerned, aside from the extra millions of dollars they've spent on AI research, it's the same old story. Nobody needs a somewhat smarter version of AskJeeves.
Google is a victim of its own success and of the increasing global accessibility of networked communications. If 90% of the American search market were split between 5 companies of roughly equal popularity, the ROI of gaming ranking to trick any one of those implementations would be much lower. There would be fewer people who could make a living by faking relevance/quality signals for junk.
Right now the best paying job many people with unspecialized skills can get is "tricking people into clicking things they shouldn't." Google is sorely taxed trying to keep up with the antics of a million people whose career is trying to game Google. Early Google was better because the people really desperate for money couldn't even afford to get online. That was a glaring inequity that doubled as a crude spam filter [1]. I think about this every time a real live person telephones me on behalf of "Windows Support."
[1] This is a large part of what I miss when I'm pining for the early Web. Practically everyone publishing online then had to be either more affluent than average or cleverer than average to get into Club Web. People contributing on the early Web were almost all financially situated independent of what they were contributing, so Web participation was almost all done out of passion rather than financial desperation. Authors didn't worry about how to get paid for what they wrote online and readers didn't worry about how to support their favorite sites either. People in the club were understood to have other means of sustenance. If you didn't, you wouldn't be in the club in the first place!
First: the problem isn't websites made by poor people. My biggest waste of time is ordinary quality but utterly irrelevant results.
Second: very often the problem isn't like before where websites included valuable search terms in white text on white background but rather that Google includes results that never included the search terms at all.
These are entirely Googles fault I think for over-optimizing for quantity instead of quality, not the fault of black hat SEO.
Yeah, being successful has its problems, as everyone wants to game your search engine, but what has that to do with the changes they made to make their search almost impossible to refine.
Operators are basically useless, they may or may not be respected, I can't select results from discussions like there was before, etc.
So basically google decides what to do with my imput, I have no saying on what I really want, so I have to deal with lots of crap.
I'm really one of those lazy users, and I find myself using other search engines more and more, just because this reasons, and despite their "normal" results being worse than google.
This is specially bad in Spanish. There aren't many alternatives in Spanish and sometimes drives me crazy. To the point that I looked into crawling myself some sites and put some OS search solution on top (I'm poor and It's a lot of work, so I didn't finally).
> Nobody needs a somewhat smarter version of AskJeeves.
That's exactly what I need. It's not right for everything, but asking questions is the natural form of information seeking for a human. Being able to do that well is a huge value add.
Yeah the current state of the tech is (in my opinion) just a voice command line. You have to know the right keywords to say, and in the right order, to get it to do what you want. It's largely an exercise in guess, test, and revise.
Interestingly, it didn’t just boil down to a Quora/StackOverflow model; it wasn’t a “wisdom of crowds” thing. Instead, your question really was used as a search query — but instead of searching a pool of documents, it would search a pool of experts, matching you with an expert who knows about similar things†, then facilitating contact with them (and forwarding them your initial query/question to start off the conversation, like a Helpdesk system.)
† Not sure how they did this part — for academic experts, they could “just” fulltext-index their corpus of published journal papers, to build up a “knowledge fingerprint” of the expert. Not sure what they would do for people in industry without a stream of publications, though.
Sadly, Google bought them, shut down the Aardvark product, and probably just put the engineers on regular SRE code-slinging tasks. It almost seems like Google felt threatened. And — hint hint — nothing’s stopping anyone from building something like this again :)
it blew everything else away pretty much immediately.
Right, I forgot people actually believe that.
Um, they were a little better, maybe noticeably so by one out of a thousand people. But wow, that's not why people switch search engines in droves.
The real reason, aside from their gift for self-promotion (I first heard about them in a science glossy, which was rare for a web 'company'), is that they had a cute, zany name, and didn't do the three things I mentioned.
Google was far more popular because of its spartan design, than its quality, regardless of how people mythologize the company now.
Notably, at a the time, a lot of those of us who were "power users" tended to stick with Altavista longer in many cases because Google was at first not so much better that they could compete with highly precise use of Altavista's search operators.
Google was much more noticeably better for non-technical who didn't know how to improve their results and didn't care to learn, but at the time that didn't translate to a massive numerical advantage in terms of users.
Yep, I also stuck with AltaVista for a couple years (or maybe Fast? Long time ago) because the I didn't find Google's results significantly better. I don't think it was the operators. In the late 00's Google's operators were the +traditional +plus +to +and +terms. They didn't introduce them immediately? Boy I was pissed when Google extracted that into 'verbatim mode'
Google's results were significantly better than other search engines, but only for a while. Soon SEO caught up with it using link farms and other forms of link spam. Then it was basically the same as old engines which were already spammed to death using other methods.
That period of better result quality, combined with other factors you mention, was long enough that most of Google's competition went away. So when the quality went down there was no-one to switch back to.
do you have evidence to support this position? they certainly weren't the only search engine with a cute, zany name. trying to recall exactly why i switched. i don't entirely recall, but i remember that at the time i used a rotation of multiple engines, usually starting with metacrawler and then branching to other engines that didn't index. after switching to google, that rotation immediately ended. my recollection is that results really were that much better.
i certainly appreciated google's design minimalism, but it certainly wasn't make-or-break, seeing as i was willing to go through multiple sites on every search.
i wouldn't have cared about what you termed 'tricking the user', which feels like a skewed characterization. i just interpreted such things as 'no results', the same way i do now.
that said, i can't really support my position either. i'm curious if anyone has tried to measure the quality of search over time.. you always see a lot of opinions about it here and elsewhere but it seems very much non-obvious to me, and hugely multivariate.
Unfortunately for my argument, my claim isn't founded on much beyond my gut feelings at the time. It would be easier to find articles that invalidate it. Certainly the ancient magazine article I mentioned (was it Discover? SciAm?) was a breathless story about boy genius graduate students and their breath-through page rank algorithm etc.
The people around me (maybe bloggers, too?) seemed as, or more, interested in Google's simple 'noncommercial' design and use of the word 'googol', than in the accuracy of Google's results.
On the plus side for my take, there's likely evidence that search users didn't consciously value search algorithms like they do now. In the 90s, people were mostly after 'total number of pages indexed'
Agreed. The first search I made on Google (~1999?), probably for something tech-related, turned up several porn results on the first page. Back to AltaVista, DMOZ, and webrings for at least another year…
No, it was better because it was much faster and results were much better. I remember those days, we (me, friends and family) were all using Altavista or lycos before, do you think we all switched to Google in a blink just because of it's spartan design?
I'm perfectly willing to check Bing or DDG or anything else I can find when google insists on misunderstanding my query. Google had exact word match, symbol for symbol, it had reliable logical operators, it had reliable string query. None of that is nearly as reliable as it was 10-12 years ago, and searching for some mildly controversial topics will get google practically yelling its opinion at you, rather than just searching for text matches.
> Far as I'm concerned, aside from the extra millions of dollars they've spent on AI research, it's the same old story.
It's interesting because it seems like the UX equivalent of "burning down furniture to heat the house" -- how does this kind of thing become so institutionalized at companies?
Is this merely the natural end season of the corporate life cycle where after innovation and growth the now engorged and dying corpse must be parted off and sold by the pound? There's something so uncomfortably Darwinian to me about that. But I suppose that's also why it's common -- it works.
> It's interesting because it seems like the UX equivalent of "burning down furniture to heat the house" -- how does this kind of thing become so institutionalized at companies?
You're in charge of revenue for a division. You give an estimate of $X for the current quarter and $Y for the next quarter; your boss changes your estimate to $1.5X and pushes it up the chain. Now there's two weeks left in the quarter and projections are that you'll only reach $1.1X, so your boss pushes you to stick more ads and make them bigger 'just for two weeks', but also reminds you that your revenue target for next quarter is $1.5Y, so maybe you should keep the big ads.
Yes, it would be. But if they charge for it people would not want to be served ads, and they would need to charge a lot of money for that to make sense for them.
For the first decade of the web, there were a handful of search engines competing, rising and falling in popularity. The best were altavista, and fast.
One thing that was noticeable back then was that bad search engines (and search engines that 'jumped the shark' and became bad) generally did so in similar ways:
a) they included paid results, or devoted too much real-estate to advertising
b) when they failed to find results, they tried to trick the user by showing related results (eg: omitting or substituting terms)
c) they avoided 'logical and' for search terms, in favor of 'logical or', making it difficult for users to search with precision.
The people at Google surely believe their recent changes have nothing to do with all that. Far as I'm concerned, aside from the extra millions of dollars they've spent on AI research, it's the same old story. Nobody needs a somewhat smarter version of AskJeeves.