More Truth About Twitter Friday, August 14th, 2009 Revisited If Twitter Was 100 People with some interesting new data Previous Next Tweet Pin It Tags — pop-culturesocial mediaweb Credits — concept & design: David mccandless sources: Sysomos: Inside Twitter, Sysomos: Twitter's Most Active Users, Pear Analytics Twitter Study Related What Does China Censor Online? How Much Do Music Artists Earn Online? Million Lines of Code Books and Store Show Comments ( ) lepht If there’s only 100 twitter users how can five of them have more than 100 followers? david I’m thinking that the text around the image is ‘outside’ the metaphor of 100 people – they’re like labels. So they reference the ‘real world’, not the “if twitter was 100″ world. Jacob The “if twitter were 100 people” graphic is wrong. There should be overlap between the blue and purple groups, and the green group should only be thirty users, not fifty users. See the original post with stats. I think it is important to vet visualizations for accuracy, no matter how pretty they are. http://thompson-bender.com Chris S. Cornell Yes, you are certainly right. This graphic flat out misrepresents the figures in the original post. The person responsible for doing this work really needs to understand statistics and basic math, and the editor needs to double check. http://neoicity.blogspot.com/ Abraham Si tan sÃ³lo asÃ fueran las clases de estadÃtica. Blubb The first picture in this post is misleading. Mapping a data set of percentages on a graphic of 100 people creates the illusion that these numbers are mutually exclusive – for example that people who have more than 100 followers and people who create 75% of the tweets have no intersection. People graphics should never be used to represent a set of data that has intersections. david Yes I’m struggling with this aspect. 48% of people who have 100 followers are also “loudmouths”. Any ideas on how to represent this while also keeping the visual metaphor? Thanks! D (source: http://www.sysomos.com/insidetwitter/mostactiveusers) http://www.proactiveseo.com.au ProactiveSEO Interesting stats. So the 80-20 rule also applies here? i.e. 20% twitters generate 80% of all the tweets? http://www.heyamaretto.com Diane Guercio I was writing a post on what happens when you take the “social” out of social media when I saw your graphic representation of twitter stats. This image of a network of bots tweeting bots keeps flitting through my mind. http://spiral-scratch.blogspot.com Liz How do you get a gender breakdown? Twitter doesn’t ask for any demographic information when you create an account so gender is something that I guess you inferred. Or do you have an inside source? http://mylifeisnotveryinteresting.posterous.com alex van de sande David: you can use striped colors for intersections, or you can group them in different ways – colors for people with 100< followers and a bigger space to separate loudmouths. Also, I think it would be really interesting if you explore more the "bots" stats by putting a robot sillouette instead of a man/woman figure. Something discrete, but a nice way to integrate because, as the others, I bet there are tons of bots in every group david @alex great idea on the stripes. Thanks! Those bots are ‘robots’ BTW – i think the image is too small to distinguish though Alex Torrance How about the loudmouths just have open mouths? That way you can still colour-code the >100 followers. http://jacksoncouse.com Jackson Couse but how useful is Twitter? Do people enjoy babble? RossEBurton Maybe you could map the overlap by “striping” multiple colors on the same person symbol. For example, half of the purple “100 or more” people could be striped purple and blue to represent they are also loudmouths. Of course, you’d then have to decrease the number of solid blue loudmouths. Not totally sure this is visually accurate or as visually easy to read though. Just a suggestion. Also, I’m having trouble with the 100 or more followers thing. It’s the only number that refers only to the “real world” and not the 100 people twitter world (Everything else is text or a percent of total). Maybe you could use a percent for this one too? Like, instead of saying “more than 100 followers” say something like “more than 0.00001% of the total users as followers.” This isn’t as informative though, so I dunno. http://www.wealthesteem.org Paul Zagoridis @Blubb I see and agree with your point except the people metaphor is very engaging plus the intersection sets are interesting independent of the big picture. Given the Blue and Purple people are subsets of the Grey, I think the graphic gets the message across. If loudmouths AND popular was greater than 10% there could be another color. Alternatively a note stating that there may be an intersection there would be good. I haven’t looked at the underlying data, but intersections are notoriously absent from reported data. http://www.thisplaceisastranger.com Jacob Estes If you are getting upset over how simple the twitter world has been made with these graphs, then you are missing the point. This isn’t being presented to the National Committee to Reform and Regulate the Internet. Let it go. Calm down. jw Is the implied gender distribution of your people within your subsets accurate? All subsets have somewhat even gender distribution but the uncategorized (grey) holds more females? Jolyonwagg1 That seems about right.Just joined Twitter to see what all the fuss was about? 80% sheer drivel, 20% intelligent. Everyone sticks to each other like metal and magnets? \jolyonwagg1 Interesting ? send in a moderate comment and does not get shown? Why bother? Plurkbuddy er, how were the colors chosen? TumThustytall I came across which completely elaborately describes not far from various credit card processors. [url=http://henpt.myftpsite.net/torch-lighter.php ]cl [/url] Conditioner: Wet the hair. component ratio Yes, we experience made mistakes sometimes non-standard due to the decades, but can any other provinces in two centuries have acted in as clear-headed, and decent a construct as us ALL THINGS CONSIDERED. Steve Should there be a “1 person is Stephen Fry”? http://creativemediaworkshop.com Eddie Mercer This is pretty dope, puts the tweet about it in the top 8%. I’m retweeting it, haha. http://twitter.com/YogaChicky YogaChicky I love your blog! Came here via Mashable’s post about your Hierarchy of Digital Distractions. I’m super-visually oriented myself, so your stuff is like… I dunno, chocolate for me :) http://www.gtmsrl.com/new/index_ita.php pareti divisorie davvero interessante!! http://WieberArt.com grace wieber Thanks for the breakdown ~ especially the peak hours. Very informative and extremely helpful. http://businesseshome.net Margaret I recently came across your blog and have been reading along. I thought I would leave my first comment. I don’t know what to say except that I have enjoyed reading. Nice blog. I will keep visiting this blog very often. Margaret http://businesseshome.net pkalina I disagree with the assertion that 40% of tweets are babble. Behind each tweet is background information which is understood by the originator of the tweet and by some of his or her followers. Anyone who lacks that context does not have enough information to judge whether a tweet is babble or, for example, a clever inside joke. This is explained further at http://pkalina.posterous.com/wide-tweets-or-local Mark Campion Love your work here. What software do you use to generate a 100 people chart from a raw dataset? Is there free software that’ll do this? http://alt-tag.com Tom The intersection could also be communicated by changing the background behind the people. Some Purple People in the box, some Green People in the box. Easy intersection, and less confusing than stripes. http://www.MountainMarketingGroup.net kathy bowling I would like to see another breakdown, to reflect how many of the tweeps are tweeting on behalf of someone else…not robots, but real people. How can we measure how much actual interaction is happening vs pr? http://www.nashvilleseo.info NashvilleSEO I would love to see more data on the 20 dead and 50 lazy. What differentiates them? What could be gleaned by a timeline graph (when people abandon or become engaged)? How many of those accounts were set up from IPs in India? ;) http://www.socksandpuppets.com Ahdok I think the dead people might be removed accounts, rather than simply very inactive ones? – It’s possible to delete your twitter account. harry @ lepht maybe some people follow more than one person http://www.zoombits.co.uk/christmas-gifts gifts for men Great defence of and twitter and fascinating history of its development. Will circulate this to colleagues who trot out the â€˜why should I care what youâ€™re doing right nowâ€™ line. Thanks. http://ur7s.com/tournaments/george7s Barney An interesting insight into Twitter, would be good to go into a bit more detail, but i will be tweeting the news tweets i want RTing at the suggested time to see if there is any improvement. What is the source of this information?? Phil There is no way to distinguish babble from other tweets without detailed knowledge of both the sender and recipients, self promotion could also be chatty and/or retweetable.There will be people who haven’t tweeted in the last week but have been/will be active at other times. Together with the other problems mentioned with the presentation of the information this ends up fairly useless. Paul Wilson If you have not tweeted in the last week it is likely that you are far from “lazy”. You are probably too busy to look at Twitter or Tweet. What a ridiculous set of visuals and statements to go with the dubious interpretation of the data. It just shows how you can justify things with mis representation – I may show this to my students, as a warning against sloppy work. Fez @NashvilleSEO and Ahdok – I think the ‘dead’ accounts, which are described as empty, are the accounts where people have signed up out of curiosity and never actually tweeted. This is what distinguishes them from the ‘lazy’ accounts. Fox I have had a Twitter account since it first became known about but I never Tweet and only follow a local council. I have followed some big names in the past but 90-95% of the stuff on there is pure crap. I also do not allow anybody to follow me despite the best efforts of the leeches that like to attach themselves to anybody regardless of who they are and without even knowing them. At the end of the day Twitter is a huge network of accounts most (well over 90%) of which are dead, dormant or held by people who don’t see the point. I cannot see it lasting for long. http://www.edutechblog.com Sebastian Waack @david The first image is puzzling to me mixing up to many aspects (tweets, followers, bots …) while leaving to many people unexplained; what are the grey people doing the whole twitter day long? Maybe you could visualize as an underlying aspect the tweets per day from 0, 1, 2-9 and 10+ approximately corresponding to 85 “dead” and “lazy”, 6 “regular”, 8 “active” and 1 “loudmouth”. That would leave no grey people and also avoid the dilemma of how one can have more than 100 followers in a world of 100 people. Anyway, thank you very much for your inspiring visualizations! Achim Kemmerling Why not give half of the loudmouths, popular people a mixed color to show the intersection? (And, of course, reduce the total number to some 7 or 8). Best AK Andre I really like the infographic! I do agree with lepht’s comment ” If there’s only 100 twitter users how can five of them have more than 100 followers?” though. Don’t quite understand it. http://www.wordcurl.com Brett This is a great collection of stats but I wonder if or how much it has changed since this was posted. I also wonder about the methodology of categorizing something as either ‘inane’ or ‘chatty’. Based on my own website’s metrics, peak tweeting hours are between noon and 2PM EST. I have a few other unique metrics on my website if you are interested in this sort of thing: .