First, welcome to readers from progressivebloggers.ca, where I am now an affiliate (just).
If you are coming here for the first time, let me catch you up. I’m in the middle of a chapter-by-chapter, nearly page-by-page critique of Chris Anderson’s book about "the future of business" called "The Long Tail". I’m doing it because the book is widely read, influential, enthusiastically pro-market, and also sloppy and wrong wrong wrong. If you want to see the beginning of the series, go here. If you want to see them all in descending order, go here.
I usually try for about three chapters a week, but I only got two done over the last week, so here’s a bonus teaser for the next chapter, which is Chapter 7. It’s a little thing – one sentence in the middle of the book – but it says something about the author’s cavalier attitude to facts.
Chris Anderson says:
Dell spends hundreds of millions each year on promoting its quality and
customer service, but if you Google "dell hell" you’ll get 55,000 pages
of results.[99] (I put page numbers in square brackets)
At ten results per page (the default) that’s 550,000 results. Let’s see if we can reproduce this number.
First let’s try "dell hell" in quotes, as he shows it. Over at the right of the page it says "results 1-10 of about 69,600". That’s 696 pages, not 55,000.
But Google’s "about" numbers are notoriously inaccurate. To get a more precise number, first I recommend changing the preferences to return 100 results per page (less clicking), redo the search, and then click through the pages. How many results do you end up with? A total of 693. That’s 70 pages of results, not 55,000.
Suppose he meant without the quotes. Let’s try that. The "about" number at the top right of the page is now 1,620,000 – or 162,000 pages of regular results. That’s about three times the number Anderson quotes – perhaps that’s what’s changed in the last two years. So this is where he gets his number from, I guess.
But as I said, that "about" number is notoriously unreliable. Click through the pages and how many results do you get? 753. That’s 76 pages of results at the default ten-per-page.
So Anderson claims 55,000 pages of results because it’s a big number, and a big number illustrates the point he is making, but a five minute test shows that he is overestimating by about 54,924.
So much for that claim. And unfortunately (in case you were wondering what I think) it’s typical of the book.
I don’t know the context but it seems most likely to me that when he said “55,000 pages of results” he meant “55,000 web pages use the term”. Your number of 69,600 is broadly consistent with that (I currently get 64,800).
The method of clicking through Google’s result pages to get a ‘real’ result also isn’t very useful. I think Google only ever gives you something like the ‘unique’ part of the top 1000 hits, so you can never get more than 100 “pages of results” in the sense you understand it.
Great series of posts, though, it’s nice to see this sort of detailed intellectual review of a top “non-fiction” book. That it turns out to be a fluffy collection of anecdotes with an alternatively weak central premise and an overbroad-to-the-point-of-uselessness central premise is not surprising. The Tipping Point was quite similar. Freakonomics is slightly different in that it doesn’t really pretend to be anything else than a collection of “heh, statistics sure is cool” anecdotes.
Thanks for the comment.
There’s not a lot of context to the quote from the book, so you’re not missing anything there. The statistic is thrown in to show how powerful consumer sentiment is, even for a large company like Dell.
I really hadn’t thought of reading his sentence as “55,000 web pages use the term”. I guess it’s possible, although it would be a strange phrasing.
What you say about Google not showing more than 1000 or so search results is interesting – and would make my criticism wrong, even if it wouldn’t guarantee that Anderson’s number is right. Any source for that? Or is it observation? I don’t know much about PageRank or the way google works, but it does sound plausible – I tried it with a couple of queries that would obviously return huge numbers of results and I do see what you say. Aargh. I may have to retract.
It *is* a very strange phrasing. Perhaps something like “a result of 55,000 pages” got mangled into “55,000 pages of results” somewhere along the way from the author’s brain to the printed book. Another way to think of it is as an amalgam of “55,000 pages” and “55,000 results”, each of which would be reasonable. Or if you had 55,000 instances of the phenomenon “a box with a book in it” you could reasonably talk about “55,000 boxes”, “55,000 books” or “55,000 boxes of books”.
I can’t find any official documentation at Google for their 1000 limitation but if you manually modify a url to something like this:
http://www.google.com/search?q=dell+hell&hl=is&lr=&start=1001&sa=N
You should get confirmation of the phenomenon.
In any case this is such a minor issue that it’s probably not worth training the big guns on. His point is that lots of people are unhappy about Dell and that’s a well-documented fact.
No argument there.