There is a growing feeling that Google's search doesn't really work any more - that, not to put too fine a point on it, compared to ten or five or even two years ago Google sucks. Like Arrington, when I used Google for the first time back in 2000 it seemed like magic and I eulogised to others just as if I was sharing a magic trick with them. If two people who knew the Google secret discovered they were talking to one who didn't they shared a happy smile. It was really that good and that impressive and to this day I remember the first ever search my colleague Brendan encouraged me to carry out - "how many turkeys does Bernard Mathews have?" - and the first result - a page from his company's website telling me he had about 13 million of the things. Magic.
Today, what once was magic just looks like a list of content farms, links to Google's own commercial sub-sites and whoever has spent the most on SEO. (They used to be cool but now they're too commercial? Well, maybe sometimes the cliche is true.)
Three things have changed at Google since the days their search results seemed like magic. Two are very well known.
The third change is less widely discussed (because it's complicated) and less understood (ditto) but more crucial to the perceived success or failure of Google's search results, and is essentially the secret of Google's popular success that was cracked by Chris "Long Tail" Anderson back in 2005. You should follow the link and read his short, excellent post on the subject now.
Google is a probabilistic system. Any algorithmic search engine is a probabilistic system - not a source of definitive, regulated authority but merely a system that is probably right. Google has (had, if you like) an excellent algorithm for determining the relevance of web-pages to user queries and throwing out the best one. Unlike its directory-based predecessors it didn't rely on a people to filter the web, it just threw up a match that its algorithm though was probably the one you wanted. A lot of the time it worked - so often, in fact, that it seemed like magic. But you were still just relying on the probability, rather than the certainty, that the algorithm had thrown up the best page each time.
As Chris Anderson wrote, people aren't wired to like (believe, trust) probabilistic results. We want there to be someone behind the curtain, for there to be a source of authority and wisdom responsible for the quality of the answers. Google solved that problem by being so damned good that we personalised it, we treated it like it was that necessary source of authority. "JFGI", we'd say - an exhortation to fools who still asked general knowledge questions when they had a computer within reach, a statement of our trust and reliance on Google. Sure, it was just an algorithm, just a probabilistic system, but it looked like magic and it felt like a source of authority and so we believed. Just fucking Google it. Trust the box. It knows.
Deteriorating quality of results is one thing, and sure it makes Google look less impressive than it did. Taken on its own, we'd have let it go (there's nothing better, we're used to Google, experiments like Bing and Wolfram Alpha and Mahalo have turned out to be frauds or jokes or failures). But combine it with the emergence of social search, and suddenly Google looks a lot less like a source of definitive authority and more like a far away room full of computers spitting out probabilities. Ask Twitter, ask Facebook, some people even ask Quora, and you get actual answers back, not probabilistic results of potentially relevant content pages. You get people you know, or at worst people you can check up on, answering the question you asked.
It's by comparison to the quality and the authority of the answers we get back from social tools like Twitter and Facebook that Google's search now seems to suck. Quoth my friend and former colleague Marc Cooper:
"I’m going to a work party on Thursday and the theme is ‘Mad Men’. So, I wear a suit, I know that much, but what else?
Ask Google, and I get spam, fancy dress rental firms, and recipes.
Ask Facebook/Twitter, and I get advice from friends on slicking my hair to the side, chain smoking, wearing a thin tie or dicky bow, a hanky in my top pocket, and golf shoes.
So now I know what to wear. And I know who to thank. (But they’re not getting a picture)
Social is more useful, and trustworthy, than search engines."
Of course it is. For a decade Google pulled a brilliant trick of making a search engine so good that we trusted it as if it was a real source of authority rather than the probabilistic engine it really is. But the economist's question is "compared to what?" Ten years ago, we compared Google (and Google's results) to the answers we could get from Yahoo, Alta Vista, Lycos. Today we compare poorer Google results to the answers we can get from people we actually know on Twitter and Facebook.
There was never anybody behind the curtain. Google never pretended there was - we pretended for them. Now we know better because we have better comparisons, and so Google's search results suddenly seem dramatically worse than they were.
Comments