Mike and Loic are wrong about Twitter search

Bob Warfield has it all right: Loic Le Meur’s call for authority-based Twitter searches is all wrong.

What is Loic’s idea? To let you do Twitter searches with results ranked according to number of followers.

You’d think I’d be all over that idea, right? After all I have a lot more followers than Loic or Arrington has.

But you’d be wrong. Ranking by # of followers is a stupid idea. Dave Winer agrees. Mike Arrington, on the other hand, plays the wrong side of the field by backing Loic’s dumb idea.

Here’s why it’s a stupid idea: everyone is gaming the number of followers. And, even if everyone weren’t, popularity on Twitter isn’t a good way to measure whether a Tweet is any good or not.

It would increase noise, not decrease it. After all, if such a system were in effect you’d see my Tweets at the top of the page, even for things that I don’t have any business being at the top of the page for.

For instance, let’s say we were talking about something in China. How about something affecting supply chain management. Who should be at the top of such a result? @liamcasey because he runs a sizeable supply chain management company in China. But, no, he won’t be at top if Loic gets his way. I would be. That’s really lame.

So, what’s a better idea? Study the metadata that really matters.

Here’s some on Twitter:

1. Number of retweets of that tweet.
2. Number of favorites of that tweet.
3. Number of inbound links to that tweet.
4. Number of clicks on an item in Twitter search.

On friendfeed there’s even more to study:

1. Number of likes of that tweet.
2. Number of comments on that tweet.
3. Amount of resharing of that tweet.
4. Clicks on each tweet.
5. Velocity of commenting and liking behavior.

On both services you should see a bias of tweets made by people you’re actually following. Who you are following is a LOT more important than who is following you. Why? Those are active choices YOU made, which should tell the system something about you and who brings you the most value. The numbers of people following you is almost totally irrelevant.

I really hope that the Twitter team doesn’t listen to the popular users on this issue.

Oh, and friendfeed, why is your search so bad?

I can’t pull much value out of the search engine. Why can’t I say “show me all tweets that include the word ‘obama’ and that have two or more likes and three or more comments?” If we had the ability to actually pull value out of friendfeed’s database this whole argument would be moot.

To Loic and Mike: since when did “authority” have anything to do with “popularity?”

Comments

  1. About Friendfeed, I think they really either need to build up their search feature, or give a 3rd party an easy way to do it. I completely agree.

  2. About Friendfeed, I think they really either need to build up their search feature, or give a 3rd party an easy way to do it. I completely agree.

  3. I made the same point over on Techcrunch following your comment there, Robert. I think that this is absolutely ridiculous. I don’t see it happening, but even the request comes across as very shortsighted, elitist, and ego-centric. I understand the point, and from his angle, it definitely would improve his experience…forsaking the rest of us.

    Thanks for trying to set the record straight, Robert.

  4. I made the same point over on Techcrunch following your comment there, Robert. I think that this is absolutely ridiculous. I don’t see it happening, but even the request comes across as very shortsighted, elitist, and ego-centric. I understand the point, and from his angle, it definitely would improve his experience…forsaking the rest of us.

    Thanks for trying to set the record straight, Robert.

  5. Loic Le Meur’s Search by Authority idea was a stupid one. What is he up to? Trying to establish ways to game Twitter for traffic, fame and fortune?

    Just leave Twitter as it is.

  6. Loic Le Meur’s Search by Authority idea was a stupid one. What is he up to? Trying to establish ways to game Twitter for traffic, fame and fortune?

    Just leave Twitter as it is.

  7. Your post is right on!
    Retweeting is a much better metric than followers.
    Focusing on the number of followers also sometimes means people don’t follow new people out of a fear of ruining their ratio of followers to followed.
    -Josh

  8. Another month, another obsession with numbers. What we are seeing is bottoms up leadership through microblogging (insert your favorite buzzword here) that is moving the attention span of the readers from the larger newspapers and blogs to the smaller community.

    Putting it another way: it is becoming a smaller world, and the relevance of the masses is to the community, not a central figure.

    And that rocks.

  9. Your post is right on!
    Retweeting is a much better metric than followers.
    Focusing on the number of followers also sometimes means people don’t follow new people out of a fear of ruining their ratio of followers to followed.
    -Josh

  10. Another month, another obsession with numbers. What we are seeing is bottoms up leadership through microblogging (insert your favorite buzzword here) that is moving the attention span of the readers from the larger newspapers and blogs to the smaller community.

    Putting it another way: it is becoming a smaller world, and the relevance of the masses is to the community, not a central figure.

    And that rocks.

  11. Robert,
    Your completely right on this. An additional note, having more followers has nothing to do with “authority”. In most cases its just “internet celebrity”. Having more followers than say, someone in China, makes your tweets about the earth quake rank higher than me?

    Robert would each twitter user eventually have a sort of page rank based on past link reputation? Link reputation summed up adds to the total of your twitter source reputation?

  12. Robert,
    Your completely right on this. An additional note, having more followers has nothing to do with “authority”. In most cases its just “internet celebrity”. Having more followers than say, someone in China, makes your tweets about the earth quake rank higher than me?

    Robert would each twitter user eventually have a sort of page rank based on past link reputation? Link reputation summed up adds to the total of your twitter source reputation?

  13. Adam: well, I expect there will be a TechMeme someday for tweets. That’s EXACTLY how TechMeme works (past behavior is best predictor of future results). I just hope it’s fairly granular, like how Google works today.

  14. agree. the number of followers should not be looked at as authority. having said that though, I would just much rather have all of those options to sort as I see fit and let me figure out what I think equals quality or authority.

    @jakrose

  15. Adam: well, I expect there will be a TechMeme someday for tweets. That’s EXACTLY how TechMeme works (past behavior is best predictor of future results). I just hope it’s fairly granular, like how Google works today.

  16. agree. the number of followers should not be looked at as authority. having said that though, I would just much rather have all of those options to sort as I see fit and let me figure out what I think equals quality or authority.

    @jakrose

  17. My understanding of what he was saying is to allow you to search by # followers, but not make it the only way to search. I do see a purpose in it… it’s like your ‘distribution rate’ in terms of numbers of readers of a newspaper. If you purely want to know how many people saw a tweet (or get close by seeing if people with a lot of followers tweeted it), this is a good way to do it.

    But, I agree with you, it should definitely not be the only way to search. There are many other factors that make a tweet influential or not.

  18. My understanding of what he was saying is to allow you to search by # followers, but not make it the only way to search. I do see a purpose in it… it’s like your ‘distribution rate’ in terms of numbers of readers of a newspaper. If you purely want to know how many people saw a tweet (or get close by seeing if people with a lot of followers tweeted it), this is a good way to do it.

    But, I agree with you, it should definitely not be the only way to search. There are many other factors that make a tweet influential or not.

  19. I think you are dead right on this one Robert, the idea of page ranking tweets based on the number of followers is a bad idea. I make a habit of actually checking out the profile of anyone that follows me before I follow back, and if it’s a spammer I will take a look at who is following them, and you’d be surprised at some of the people following spammers to inflate their numbers.

    Maybe some people should be looking at the massive scrape that was done of twitter where they looked at the communication between people and established a page rank for people:
    http://groups.google.com/group/get-theinfo/browse_thread/thread/605a00d5ddc62d72?pli=1

  20. I think you are dead right on this one Robert, the idea of page ranking tweets based on the number of followers is a bad idea. I make a habit of actually checking out the profile of anyone that follows me before I follow back, and if it’s a spammer I will take a look at who is following them, and you’d be surprised at some of the people following spammers to inflate their numbers.

    Maybe some people should be looking at the massive scrape that was done of twitter where they looked at the communication between people and established a page rank for people:
    http://groups.google.com/group/get-theinfo/browse_thread/thread/605a00d5ddc62d72?pli=1

  21. Not to mention, the argument somehow assumes that we all use twitter for the same reasons and that quantity trumps quality. Someone simply may not have the time to devote to twitter to gain tons of followers. Does that make their message less important? SM is not one-size-fits all. Let’s stop trying to make it that way.

  22. Not to mention, the argument somehow assumes that we all use twitter for the same reasons and that quantity trumps quality. Someone simply may not have the time to devote to twitter to gain tons of followers. Does that make their message less important? SM is not one-size-fits all. Let’s stop trying to make it that way.

  23. [...] Robert Scoble and Bob Warfield also believe that relevance (defined by popularity of the tweet) should be an important input in Twitter search. Robert Scoble suggest that the popularity of a tweet should be measured by a combination of retweets, favorites and inbound links for that tweet. I think search result clickthroughs are less relevant in Twitter search, as clickthroughs will only happen on tweets with links. [...]

  24. Twitter IS about noise. It is very individual how to filter it. Or vise versa how to amplify it. If noise reduction will be possible without intelligent effort it is hard to expect any intelligent results.

  25. Twitter IS about noise. It is very individual how to filter it. Or vise versa how to amplify it. If noise reduction will be possible without intelligent effort it is hard to expect any intelligent results.

  26. I agree with you on this one. The idea’s pretty much spot-on ridiculous. Popularity shouldn’t be associated with authority. And authority isn’t always the equivalent of quality. Believe it or not, there are some people out there who turn out quality content, but aren’t among the more well-known. This isn’t to say that those with more followers don’t turn out good content – it just doesn’t hold true for everyone.

  27. I agree with you on this one. The idea’s pretty much spot-on ridiculous. Popularity shouldn’t be associated with authority. And authority isn’t always the equivalent of quality. Believe it or not, there are some people out there who turn out quality content, but aren’t among the more well-known. This isn’t to say that those with more followers don’t turn out good content – it just doesn’t hold true for everyone.

  28. “popularity on Twitter isn’t a good way to measure whether a Tweet is any good or not.”

    Given but numbers are measurable results and people will always strive on numbers, whether it is to measure exposure or success. I say have the best of both worlds have great followers and have great content, can’t beat that.

  29. “popularity on Twitter isn’t a good way to measure whether a Tweet is any good or not.”

    Given but numbers are measurable results and people will always strive on numbers, whether it is to measure exposure or success. I say have the best of both worlds have great followers and have great content, can’t beat that.

  30. I like your take on this. Still… retweets, which are important, are at least partly a function of number of followers. (Some really smart tweets won’t get retweeted because they they are comments from someone who only has a couple hundred instead of several thousand followers.) Still, it is a better measure than authority on its own.

  31. I like your take on this. Still… retweets, which are important, are at least partly a function of number of followers. (Some really smart tweets won’t get retweeted because they they are comments from someone who only has a couple hundred instead of several thousand followers.) Still, it is a better measure than authority on its own.

  32. As mentioned in my follow up post immediately after the TC post

    While counting re-tweets, the re-tweets generated on original content should be counted and not on ones which are someone else’s content

    for ex: count if someone did RT @mayankdhingra “xyz” but not if someone did RT @mayankdhingra @scobleizer “xyz”. In the latter case retweet should be counted for your account not mine

    more here http://mayank.name/blog/2008/12/27/an-algo-for-twitter-authority/

  33. As mentioned in my follow up post immediately after the TC post

    While counting re-tweets, the re-tweets generated on original content should be counted and not on ones which are someone else’s content

    for ex: count if someone did RT @mayankdhingra “xyz” but not if someone did RT @mayankdhingra @scobleizer “xyz”. In the latter case retweet should be counted for your account not mine

    more here http://mayank.name/blog/2008/12/27/an-algo-for-twitter-authority/

  34. Loic’s comment is so far off the mark, one gas to wonder if the comment was made for the purposes of whipping up contaversy. Loic has seen how many clicks he gets when he is at the center of a very public debate. The problem is, that ruins his credibility, influence and personal brand long term.

  35. Loic’s comment is so far off the mark, one gas to wonder if the comment was made for the purposes of whipping up contaversy. Loic has seen how many clicks he gets when he is at the center of a very public debate. The problem is, that ruins his credibility, influence and personal brand long term.

  36. Le Meur was describing ‘influence’ (eg. follower #s) and mislabeled it ‘authority’ (eg. expertise). This is simply one of the ways one should be able to filter search results. Scoble nails another, retweet #s, which can be gamed like digg #s. An approach using AI would be ideal, but currently not possible. TweetDeck is addressing this issue far better than Twhirl, or any other product at the moment.

  37. Le Meur was describing ‘influence’ (eg. follower #s) and mislabeled it ‘authority’ (eg. expertise). This is simply one of the ways one should be able to filter search results. Scoble nails another, retweet #s, which can be gamed like digg #s. An approach using AI would be ideal, but currently not possible. TweetDeck is addressing this issue far better than Twhirl, or any other product at the moment.

  38. Robert Schoble,

    I don’t really care how they do with this authority ranking. I have no problem with belonging of the very end of the very long tail… ;) It would be great if you could have a good search engine for Twitter and FriendFeed so could find new interesting individuals.

    By the way: Thanks for mentioning a supply chain expert. I will now follow @liamcasey. I worked as a purchaser between 1989 – 1997 and I am a member of a purchasing and logistics association in Sweden, so I am interested in following what is going on in this field.

  39. Robert Schoble,

    I don’t really care how they do with this authority ranking. I have no problem with belonging of the very end of the very long tail… ;) It would be great if you could have a good search engine for Twitter and FriendFeed so could find new interesting individuals.

    By the way: Thanks for mentioning a supply chain expert. I will now follow @liamcasey. I worked as a purchaser between 1989 – 1997 and I am a member of a purchasing and logistics association in Sweden, so I am interested in following what is going on in this field.

  40. Filtering by follower numbers isn’t anyone’s be-all-end-all but providing filtering is a good idea. You know even the people who say it’s a bad idea for a follower filter would test the filter regularly, and when it helped, they’d be grateful for it.

  41. Filtering by follower numbers isn’t anyone’s be-all-end-all but providing filtering is a good idea. You know even the people who say it’s a bad idea for a follower filter would test the filter regularly, and when it helped, they’d be grateful for it.

  42. Robert:

    A couple of notes on your post. I think the spirit of this discussion is the right one. Your suggestions are excellent but probably not enough.

    Anytime you reveal how search works, it can be instantly gamed. There are many smart people in the world. I can already see some smart engineers developing a system to take advantage of your ideas. Spam is a serious full-time product problem.

    Expert search products need to solve the cold start problem for people with true authority. While many on Twitter call themselves experts, the real experts may not be tweeting. When and if they do, their opinion may get lost in all the noise. For example, David Ku (VP Engineering, Yahoo! Search) is not tweeting (at least, I have not seen him) but if he did, I am not sure everyone would appreciate his brilliance, authority or expertise as it pertains to the topic of search. Something other than the volume of X or Y must be a factor and relevancy consideration. You must also be able to motivate experts to maintain their presence.

    We can learn some excellent lessons from Yahoo! Answers and Naver from Korea. To support your point, the experts at Yahoo! Answers
    lack serious credibility. Many of the experts respond in volume but their answers lack quality.

    Twitter is sitting on a big idea. Developing a set of search algorithms
    to take advantage of this opportunity should not be rushed. I hope they hire someone with serious experience to lead them in this area.

  43. Robert:

    A couple of notes on your post. I think the spirit of this discussion is the right one. Your suggestions are excellent but probably not enough.

    Anytime you reveal how search works, it can be instantly gamed. There are many smart people in the world. I can already see some smart engineers developing a system to take advantage of your ideas. Spam is a serious full-time product problem.

    Expert search products need to solve the cold start problem for people with true authority. While many on Twitter call themselves experts, the real experts may not be tweeting. When and if they do, their opinion may get lost in all the noise. For example, David Ku (VP Engineering, Yahoo! Search) is not tweeting (at least, I have not seen him) but if he did, I am not sure everyone would appreciate his brilliance, authority or expertise as it pertains to the topic of search. Something other than the volume of X or Y must be a factor and relevancy consideration. You must also be able to motivate experts to maintain their presence.

    We can learn some excellent lessons from Yahoo! Answers and Naver from Korea. To support your point, the experts at Yahoo! Answers
    lack serious credibility. Many of the experts respond in volume but their answers lack quality.

    Twitter is sitting on a big idea. Developing a set of search algorithms
    to take advantage of this opportunity should not be rushed. I hope they hire someone with serious experience to lead them in this area.

  44. Agree with you about popularity vs importance, but then again, as geeks we ALL know that lesson first hand!

    Got one to add to your list:
    1. Number of retweets of that tweet.
    2. Number of favorites of that tweet.
    3. Number of inbound links to that tweet.
    4. Number of clicks on an item in Twitter search.
    and
    5. Number of past tweets including searched keyword
    (that would help you get a few more subject matter experts)

  45. Agree with you about popularity vs importance, but then again, as geeks we ALL know that lesson first hand!

    Got one to add to your list:
    1. Number of retweets of that tweet.
    2. Number of favorites of that tweet.
    3. Number of inbound links to that tweet.
    4. Number of clicks on an item in Twitter search.
    and
    5. Number of past tweets including searched keyword
    (that would help you get a few more subject matter experts)

  46. [...] Robert of course calls bullshit on this and says results should be based on a series of metadata results like re-tweets, how many favourites for that tweet (you mean people actually use favourites in Twitter?) and number of inbound links for that tweet (in other words – a Twitter Techmeme) [...]

  47. Completely agreed. Just because someone has a large number of followers does not make them an authoritative speaker on anything they write about. If twitter were to sort searches by “authority” then anyone with a lot of followers merely mentioning a single word will ultimately be displayed well ahead of the really smart people

    Arrington advocating this idea does not surprise me. Loic advocating it does.

  48. Completely agreed. Just because someone has a large number of followers does not make them an authoritative speaker on anything they write about. If twitter were to sort searches by “authority” then anyone with a lot of followers merely mentioning a single word will ultimately be displayed well ahead of the really smart people

    Arrington advocating this idea does not surprise me. Loic advocating it does.

  49. The gaming of followers is growing exponentially. Anyone who wishes to see it can simply record results for major Twitter users using Twinfluence and check back in a few days. The number of new users with huge numbers of followers is growing rapidly.

    Just as being a celebrity is not an indication of quality, numbers of followers is not either.

  50. The gaming of followers is growing exponentially. Anyone who wishes to see it can simply record results for major Twitter users using Twinfluence and check back in a few days. The number of new users with huge numbers of followers is growing rapidly.

    Just as being a celebrity is not an indication of quality, numbers of followers is not either.

  51. I like the fact on Google, the network determines the authority.

    There’s a lot of precision though and relevancy is king. It’s about the authority links for the relevant keywords.

  52. I like the fact on Google, the network determines the authority.

    There’s a lot of precision though and relevancy is king. It’s about the authority links for the relevant keywords.

  53. [...] Robert Scoble: “Here’s why it’s a stupid idea: everyone is gaming the number of followers. And, even if everyone weren’t, popularity on Twitter isn’t a good way to measure whether a Tweet is any good or not.” [Ok, but it is a good way of determining how loud that message was] [...]

  54. I agree that aligning # of followers does not always indicate authority. But I see little harm in the option being added to search. You know I’m a big advocate for the little guy, and I try not to let # of RSS subscribers, followers etc. measure somebody. But why not enable the Twitter search database to be very flexible and include this option? Maybe I’d use it.

  55. I agree that aligning # of followers does not always indicate authority. But I see little harm in the option being added to search. You know I’m a big advocate for the little guy, and I try not to let # of RSS subscribers, followers etc. measure somebody. But why not enable the Twitter search database to be very flexible and include this option? Maybe I’d use it.

  56. 99% agree with you. IMO Technorati ‘authority’ does not work if you are outside the valley crowd, if it even really works there. One disagreement, I use favourites to mark tweets a lot but I am one of the few I know who do. Since it has low usage – I’m not sure how useful it is.

    Great that you are blogging more again

  57. 99% agree with you. IMO Technorati ‘authority’ does not work if you are outside the valley crowd, if it even really works there. One disagreement, I use favourites to mark tweets a lot but I am one of the few I know who do. Since it has low usage – I’m not sure how useful it is.

    Great that you are blogging more again

  58. That’s a pretty basic concept you’ve used to beat this idea with Robert. After all, Google uses popularity as part of its ranking algorithm but it hasn’t resulted in you ranking for Britney Spears related searches has it. I think we have to give the folks at Twitter a modicum of common sense.

    Is it really so far fetched to see Twitter apeing Google and using the number of followers, and in turn the reputation of each of those followers as ‘part’ of a ranking algorithm? Doesn’t seem to require much creativity to find a workaround to the problem you highlighted.

  59. That’s a pretty basic concept you’ve used to beat this idea with Robert. After all, Google uses popularity as part of its ranking algorithm but it hasn’t resulted in you ranking for Britney Spears related searches has it. I think we have to give the folks at Twitter a modicum of common sense.

    Is it really so far fetched to see Twitter apeing Google and using the number of followers, and in turn the reputation of each of those followers as ‘part’ of a ranking algorithm? Doesn’t seem to require much creativity to find a workaround to the problem you highlighted.

  60. [...] Saturday afternoon Robert Scoble sent out a simple tweet letting people know he popped up on the retweetradar (much appreciated Robert!), we had been talking about use of interesting metadata, for instance retweeted information possibly being used to rank quality posts on Twitter on the comments of his blog posts about a better Twitter Search. [...]

  61. [...] Robert Scoble:「なぜこれがバカげた考えかというと、みんなフォロワーの数を競っているから。そうでないとしても、Twitterでの人気度はつぶやきの良否を判断するには適していない。」[それはそうだが、メッセージの声の大きさを知るには良い方法だ] [...]

  62. Does this option to sort results the way I prefer, lessen your Twitter experience (again, built in as an option)?
    What’s the fuss?

  63. Does this option to sort results the way I prefer, lessen your Twitter experience (again, built in as an option)?
    What’s the fuss?

  64. Thank you Scoble, for sticking up for little people, or, ummm, rabbits like us. That’s why you are smarter than the rest. People just want to game a system, or perpetuate a status quo. The net is to be organic, and information not related to power.

  65. Thank you Scoble, for sticking up for little people, or, ummm, rabbits like us. That’s why you are smarter than the rest. People just want to game a system, or perpetuate a status quo. The net is to be organic, and information not related to power.

  66. [...] email marketing. But I don’t come here to question the people scammed, rather to damn the notion of authority in toto. Given that we are attempting to build systems that don’t reinforce old power structures, which is [...]