The BBC Uncovers Image Search Algorithm

Nov 26
2009

We can become a bit smug when it comes to the BBC. We generally view its level of journalistic integrity to be a bit above that of its cable TV counterparts. Last night, however, those of us involved in SEO were surprised to note that even the Beeb’s esteemed reporters aren’t immune to poor research. As is always the case when you notice something untrue reported as fact, you wonder how many facts you hear on a daily (hourly?) basis that are woefully under-researched.

The BBC news report I was watching was about the Michelle Obama / Google Images incident. A crudely Photoshopped, offensive image of the First Lady was ranking atop Google images for her name. In an explanation of how such a thing could occur, Rory Cellan-Jones, the BBC’s Technology Correspondent said:

Google doesn’t decide what comes top when you search for a word or an image. That’s determined by a complex formula. But it basically boils down to the fact that the more people click on a certain site, the higher up the list it comes.

An audio version of this part of the report is available here. For a short time, British readers can view the entire segment on iPlayer (between minutes 14:45 and 17:10). At the end of the piece, Cellan-Jones says again:

For now, the offensive picture of Michelle Obama has disappeared from Google’s search results, but if web users find it elsewhere and click on it, then it will rise up the search engines list once again.

Incidentally, my good friend Ciarán Norris was providing an accurate description of how it happened on Radio 5 at the same time (1hr, 26min in).

And it was Ciarán who figured out why the Beeb most likely said such a thing. A report on their news website stated that “the search engine’s results get to the top based on popularity, not because of any ranking system by people”, a statement apparently given to them by David Vise. There is nothing particularly untrue about that, but the BBC have misinterpreted “popularity”, taking it to mean clicks, not links. No one bothered to check out Vise’s statement or make sure they’d understood him properly. Thus, it was reported to the nation that it was users clicking on the offensive picture of Michelle Obama that pushed the picture to the top of Google’s rankings.

Of course, there may be some ounce of truth to the clicks idea, if you believe that Google closely monitors click-through and bounce rates. However, not once in the piece were links–the currency of SEO–mentioned. Taking into account that click-through and bounce rates are highly likely to be very small ranking factors, there is no way even a small amount of research would have backed up the statements made in the report.

What we’ve learned, we already knew: journalists need stories to go to press nowish and don’t have much time to put together stories to feed the public their daily news. The BBC found a quote from an expert; it was just a little misunderstood. However, recognising such mistakes certainly makes me wonder what else is reported to us as simple fact that is actually quite badly misguided.

21 Responses to “The BBC Uncovers Image Search Algorithm”

  1. Yura says:

    Umm, you still trust the mass media? Trying to figure out what *and why* they are misrepresenting is a pet hobby of mine.

  2. Geir Ellefsen says:

    BBC busted by SEO.. that’s funny :)

  3. Gareth James says:

    I’m suprised at that coming from the Beebs ‘technology correspondent’, would expect from a non tech journalist.

  4. Jane says:

    @Yura Quite :P I’m not sure they’d have any reason in particular to misreport that, besides prompting people to spend the rest of the night clicking on their own blogs…

    @Gareth It surprised me too, especially given how many people–especially in tech–have a pretty good understanding of SEO now.

  5. Sam says:

    I saw this report and thought the very same thing Jane! Glad you’ve blogged it – a lot of people will really take what they heard as the truth. How many people out there are probably now trying to visit and re-visit their own site images? ;)

  6. Jane says:

    Sam, I totally have the mental image of amateur bloggers and site owners, Britain-wide, re-clicking their own URLs over and over again.

  7. Aidan Beanland says:

    If the BBC “journalist” has performed the same search you have in your screenshot and clicked on the “official” Google result explaining pigeonrank it could have been a lot more embarrassing…

    http://www.google.com/technology/pigeonrank.html

  8. Jeff says:

    Nice call Jane.
    Although we find it funny, how many people now think they have that secret truth?
    Speed and a slight misunderstanding can often lead to some pretty wild statements.
    Thanks for sharing.

  9. Graeme says:

    I have a category on my blog called “wrong” for this sort of thing, that I happen to notice: http://pietersz.co.uk/category/media/wrong

    One thing I have noticed is the media (except specialists like the FT) are terrible at covering anything I know about (especially finance and investment). I have to assume that they are equally bad at everything else, if only I knew enough to spot it

  10. Jane says:

    Graeme,

    The other thing, besides SEO, that I know a fair bit about is swimming, and I agree that I see a lot of misinformation from the mainstream press about it. It’s never maliciously intended, but just poorly researched or misunderstood.

    This isn’t to say I’ve never written about something that I know less about than an expert, although this example would have been so incredibly easy to get right.

  11. Tilbud lån says:

    The algorithm is always changing and maybe there will be different ranking factors tomorrow.

  12. malcolm coles says:

    Maybe they listened to Radio 5? Matt Brittin, Managing Director, Google UK was on the Jeremy Vine show a couple of months back and said the number of visitors influenced the results. Apparently (there was a brief Twitter stir at the time).

    And when they show images blended into the web results, they probably measure clickthroughs on those, no?

    None of which is to take away from your main point, which I agree with!

  13. Jane says:

    @Malcolm Interesting, thanks for the link. I agree that images included in the regular SERPs with Universal search (i.e. when you see thumbnails in the Web rankings) are likely more susceptible to being influenced by click through data. I was slightly leery of saying this at all, given the fact that image search has never made as much sense to us as regular search, but yep–still a lot of discrepancies in the report :\

  14. randfish says:

    Nice catch Jane :-) I’d also add that Google themselves has said that CTR is a very noisy, easily game-able signal (just think of what you could do with Mechanical Turk and a couple hundred dollars) and thus doesn’t positively influence rankings (at least, not directly). I find that logic pretty compelling, and I think it’s sad that media doesn’t have more responsibility or accountability for accuracy on these types of issues.

  15. Luke Jones says:

    That is hilarious and somewhat worrying. How can we be sure that the majority of the BBC’s stories have any integrity whatsoever?! I was watching a part on BBC News 24 about internet security and SSL certificates. The ‘expert’ was stating falsehoods and giving inaccurate information to the public. I emailed in to give them the correct information – not even a response or a peep from it on the show.

    Ridiculous.

  16. jamesq says:

    Have to agree with Luke. I’d missed the whole Michelle Obama image story but surprised at the lack of controls to ensure information broadcast is correct.

  17. Toby says:

    Solid pot and definitely a nice find. Amazingly this is a misconception I come across with a lot of my friends and who are interested in rankings. Personally I think may people trust the internet to much without doing so old fashioned research and validation.

  18. joelchrist says:

    It will not permanent. At any time methodology may change.

  19. Kim says:

    Well, think of it this way: The people who believe this nonsense will spend more time searching for their own business. This will affect their personalized rankings, which they are likely completely ignorant of, so they’ll think it’s actually working.

    Ultimately this will buy the rest of us some time to do some real SEO work and out-rank our competitors.

    The downside is that it will make our job of explaining how SEO really works to clients, because they always want to take shortcuts.

  20. amarjee01 says:

    Some great links there, thanks – “making the stats fit the argument” is an extremely common practice, not just by the media but by professionals as well. It’s a little scary to think about how biased we all our towards accepting information that supports our presumptions, and refusing information that does not (belief-confirmation bias). We are all much more biased than we realize.

    [user name edited, as visitor came from a Y! Site Explorer result for a competitor. I'm not helping with your SEO, sorry ;)]

  21. Popularity comes from clicks. Funny CTR SEO! | SEOteky Network Inc., Philippines says:

    [...] might want to see this post of Jane Copland about BBC messing up with SEO, particularly on image optimization. I’d also recommend that you read this algorithm coined by Google called PigeonRank technology [...]

Leave a Reply

*