Online Trends Analysis - itnig friday


Published on

Almost everybody knows what an internet meme is. But what most people don't know is that there's a science behind memes and internet trends. What's the definition of meme? How does it propagate and why? How and where can we find trending memes? What are the key properties for a successful meme? And most important... where to find them?

Published in: Technology, Business

Online Trends Analysis - itnig friday

  1. 1. Online trends Analysis Francesc Gomez-Morales @francescgo #itnigfridays  
  2. 2. WHO  AM  I?  
  3. 3. Analy/cal  Heart  &  Mashup  Mind   2007   2008   2009   2010   2011   2012   2013  
  5. 5. Web  is  about  measurement  
  6. 6. Too  many  things  to  measure   Gatorade  Mission  Control  
  7. 7. A  framework  proposal     (from  a  marke/ng  perspec/ve)   hBp://  
  8. 8. Web  Analysis  Framework  (Marke/ng  perspec/ve)   On-­‐site   Owned  Media   Web  Page   “Web   Analysis”   Google   Analy/cs  /   Kissmetrics  /   CrazyEgg  (…)   Social  Media   and  other   Public  Profiles   “Social  Media   Analysis”   Facebook   insights  /   SocialBro  /  (…)   Off-­‐site   Compe/tors  owned   media   Web  Page   Web   Compe//ve   Intelligence   Alexa  /   Compete  /   Builtwith   Social  Media   and  other   Public  Profiles   Social  Media   Compe//ve   Intelligence   SocialBakers  /   DKS  Social   Smart   Consumers   media   Web  Pages   and  Social   Media  Profiles   Web   Monitoring  or   netnography   Radian6  /   Brandchats  /   Websays  /   Google  Trends   Search  Engines   Organic   SERP   Monitoring   SEOmoz  /  Link   Assistant  /   Raventools   Paid   PPC   Campaigns   Monitoring   Semrush  /   Wordstream  
  9. 9. Let’s  focus  on  trends   hBp://  
  10. 10. What  the  trend?   A  trend  (fad)  is  any  form  of  behavior  that     develops  among  a  large  popula/on  and     is  collec/vely  followed  with  enthusiasm     for  some  period     (generally  as  a  result  of  the  behavior's     being  perceived  as  novel  in  some  way)   Sociology  in  a  Changing  World   By  William  Kornblum,  Carolyn  D.  (COL)  Smith   Page  213  
  12. 12. Thank  you,  Tim  Berners-­‐Lee   Massive  adopYon  of  online  services  has  contributed  to  the     unique  and  unprecedented  possibility  of     register  and  analyze  micro-­‐trends     in  almost-­‐real-­‐Yme   hep://  
  13. 13. Key  technological  innova/ons  (IMHO)   hep://   Rich  Internet  Applica/ons  (RIA)   ex.  Use  of  javascript   Web-­‐oriented  Arquitecture   (WOA)   Apache  Hadoop  +  MapReduce  +   Google  File  System  
  14. 14. The  biggest  Social  Graph     in  human  history   One  graph  to  rule  them  all!  
  15. 15. Visualize  your  own  graph   heps://  
  17. 17. People  and  trends   Diffusion  of  Innova/ons,  5th  Edi/on   Everee  M.  Rogers   2003  
  18. 18. Social  Technographics   •  Trends  are  generated  by   people’s  ac/vity   •  Social  Technographics  classifies   people  according  to  how  they   use  social  technologies   •  Forrester's  Social   Technographics  data  classifies   consumers  into  seven   overlapping  levels  of  social   technology  par/cipa/on   People  and  trends    >  Social  Technographics   hep://  
  19. 19. People  and  trends    >  Social  Technographics   Forrester  Ladder  
  20. 20. Differences  between  countries   People  and  trends    >  Social  Technographics  
  21. 21. Different  between  Ages   People  and  trends    >  Social  Technographics  
  22. 22. all  animals  are  equal…   People  and  trends    >  Influence   …  but  some  animals  are  more  equal  than  others  
  23. 23. klout   People  and  trends    >  Influence   hep://  
  24. 24. klout   People  and  trends    >  Influence   hep://­‐media  
  25. 25. SOURCES  OF  TRENDS  
  26. 26. Sources  of  trends   •  Every  single  ac/vity  on  the  web  is  registered   and  could  be  analyzed   •  Different  ac/vi/es,  different  sources  of  trends   •  The  more  “passive”  behaviour,  the  easiest   analysis.  Analyze  Visits  is  simplier  than  analyze   blog  posts.  
  27. 27. Sources  of  trends:     classifica/on  by  user  ac/vity    Creators    Conversa/onalists   Cri/cs   Collectors   Joiners   Espectators   Inac/ves  Forrester  Ladder  
  28. 28. Ranking  of  the  main  social  networking  sites   Sources  of  trends  >  Visits/Pageviews  of  websites/content    
  29. 29. Top  sites  in  spain  -­‐  Sept  14,  2013   Sources  of  trends  >  Visits/Pageviews  of  websites/content    
  30. 30. Top  sites  by  niche   hep://   Sources  of  trends  >  Visits/Pageviews  of  websites/content    
  31. 31. heps://   Sources  of  trends  >  Visits/Pageviews  of  websites/content    
  32. 32. Youtube  Consump/on  in  Spain   Sources  of  trends  >  Visits/Pageviews  of  websites/content     hep://­‐sta/s/cs/country/spain/  
  33. 33. hep://     Sources  of  trends  >  Visits/Pageviews  of  websites/content    
  34. 34. hep://     Sources  of  trends  >  Visits/Pageviews  of  websites/content    
  35. 35. hep://     Sources  of  trends  >  Use  of  keywords  in  search  engines  
  36. 36. Sources  of  trends  >  Use  of  keywords  in  search  engines   hep://    
  37. 37. Sources  of  trends  >  Use  of  keywords  in  search  engines  
  38. 38. Sources  of  trends  >  Use  of  keywords  in  search  engines  
  39. 39. hep://     Sources  of  trends  >  Installa/on  of  mobile  apps    
  40. 40. Sources  of  trends  >  Installa/on  of  mobile  apps    
  41. 41. hep://     Sources  of  trends  >  Installa/on  of  mobile  apps    
  42. 42. No se puede mostrar la imagen. Puede que su equipo no tenga suficiente memoria para abrir la imagen o que ésta esté dañada. Reinicie el equipo y, a continuación, abra el archivo de nuevo. Si sigue apareciendo la x roja, puede que tenga que borrar la imagen e insertarla de nuevo. Sources  of  trends  >  Installa/on  of  mobile  apps     hep://      
  43. 43. hep://     Sources  of  trends  >  Use  of  social  games  or  social  apps  
  44. 44. hep://     Sources  of  trends  >  Use  of  social  games  or  social  apps  
  45. 45. Top10  stories  of  the  day  (From  May  24,  2007  to  May  23,  2008)   Sources  of  trends  >  News  aggregators  ac/vity   hep://    
  46. 46. hep://­‐pages/celebri/es/spain/     Sources  of  trends  >  Social  support  of  profiles   Celebri/es  in  Facebook  (Spain)  
  47. 47. hep://­‐pages/celebri/es/spain/     Sources  of  trends  >  Social  support  of  profiles   Poli/cs  in  Twieer  (Spain)  
  48. 48. hep://­‐pages/7703918541-­‐internet-­‐explorer     Sources  of  trends  >  Social  support  of  profiles  
  49. 49. hep://     Sources  of  trends  >  Social  support  of  profiles  
  50. 50. hep://­‐/me-­‐most-­‐favorited-­‐tweets     Sources  of  trends  >  Support  and/or  Difussion  of  content  
  51. 51. hep://­‐/me-­‐most-­‐retweeted-­‐tweets     Sources  of  trends  >  Support  and/or  Difussion  of  content  
  52. 52. Sources  of  trends  >  Support  and/or  Difussion  of  content   hep://    
  53. 53. Sources  of  trends  >  Support  and/or  Difussion  of  content   hep://    
  54. 54. Sources  of  trends  >  Find  Places  and  Check-­‐in   hep://    
  55. 55. Sources  of  trends  >  Use  of  keywords  in  Twieer   hep://    
  56. 56. Sources  of  trends  >  Use  of  keywords  in  Twieer   hep://    
  57. 57. Sources  of  trends  >  Use  of  keywords  in  Twieer   hep://    
  58. 58. The  business  of  trends   heps://­‐the-­‐value-­‐of-­‐promoted-­‐trends  
  59. 59. Sources  of  trends  >  Contribu/ons  to  open  source  projects  or  wikis   hep://    
  60. 60. Sources  of  trends  >  Contribu/ons  to  open  source  projects  or  wikis   heps://    
  61. 61. Sources  of  trends  >  Contribu/ons  to  open  source  projects  or  wikis   heps://­‐frequency    
  62. 62. Sources  of  trends  >  Contribu/ons  to  open  source  projects  or  wikis   heps://­‐card    
  63. 63. Sources  of  trends  >  Contribu/ons  to  open  source  projects  or  wikis   hep://     Top  100  most  controversial  ar/cles  in  English  Wikipedia  
  64. 64. Sources  of  trends  >  Contribu/ons  to  open  source  projects  or  wikis   hep://     Top  100  most  controversial  ar/cles  in  English  Wikipedia  
  65. 65. Sources  of  trends  >  Contribu/ons  to  open  source  projects  or  wikis   hep://     Number  of  words  of  english  geolocated  Wikipedia  ar/cles  
  66. 66. Sources  of  trends  >Adop/on  of  technology   hep://    
  67. 67. Sources  of  trends  >Adop/on  of  technology   hep://     Top  10k  websites   Top  1M  websites  
  68. 68. Sources  of  trends  >Adop/on  of  technology   heps://­‐Like    
  69. 69. TREND  DYNAMICS  
  70. 70. Meme/cs   Cultural  traits  are  transmieed  from  person  to   person,  similarly  to  genes  or  viruses.       Cultural  evolu/on  therefore  can  be  understood   through  the  same  basic  mechanisms  of   reproduc/on,  spread,  varia/on,  and  natural   selec/on  that  underlie  biological  evolu/on.   Trend  Dynamics  >  Meme/cs   Cultural  EvoluYon  and  MemeYcs   Francis  Heylighen  &  Klaas  Chielens   hep://­‐Springer.pdf  
  71. 71. Meme   Cultural  replicator;  a  unit  of  imita/on  or   communica/on     Trend  Dynamics  >  Meme/cs  >  Meme   The  Selfish  Gene  (2nd  ediYon)   Dawkins,  R.     (1989)  Oxford  University  Press.   1. Longevity   2. Fecundity   3. Copying-­‐Fidelity  
  72. 72. Memeplex   Collec/on  of  mutually   suppor/ng  memes,   which  tend  to  replicate   together   Trend  Dynamics  >  Meme/cs  >  Memeplex   Cultural  EvoluYon  and  MemeYcs   Francis  Heylighen  &  Klaas  Chielens   hep://­‐Springer.pdf  
  73. 73. Memes  and  trends   Trend  Dynamics  >  Meme/cs   Memes   Trends  
  74. 74. Meme  Dynamics   Trend  Dynamics  >  Meme/cs   Assimila/on   • No/ce   • Understand   • Acceptance   Reten/on   • longevity   Expression   • Speech   • Text   • …   Transmission   • Internet   • Mass  Media  
  75. 75. Meme  Dynamics   Trend  Dynamics  >  Meme/cs   Assimila/on   Reten/on   Expression   Transmission   Meme  Fitness  
  76. 76. The  anatomy  of  a  (Twieer)  trend   Main  takeaways:   •  the  resonance  of  the  content  with  the  users  of  the  social  network  plays  a   major  role  in  causing  trends   •  a  majority  of  the  content  propagated  to  cause  trends  arise  from  tradi/onal   media  sources   Trend  Dynamics  >  Twieer  Trending  Topics  >  The  anatomy  of  a  trend  
  77. 77. The  way  the  trends  dissipate     fits  a  normal  distribu/on   16.32  million  tweets  on  3361  different  topics  over  40  days  in  Sep-­‐Oct  2010   Trend  Dynamics  >  Twieer  Trending  Topics  >  The  anatomy  of  a  trend  
  78. 78. The  decay  factor  of  a  trend  is  1/t   Trend  Dynamics  >  Twieer  Trending  Topics  >  The  anatomy  of  a  trend   Example       A  trending  topic  with  10.000  tweets/hour   How  many  tweets  will  have  axer  5  hours?     N(t=0)  =  10.000  tweets   t  =  5  hours     N(t=5)  =  N(t=0)·∙[1/t]  =   =  10.000  ·∙  1/5  =  2000  tweets  
  79. 79. Almost  every  trending  topic  is  short   Trend  Dynamics  >  Twieer  Trending  Topics  >  The  anatomy  of  a  trend  
  80. 80. Strong  rela/on  between  influencers   and  trending  topics   Rela/on  between  authors  and  ac/vity:   •  No  propagaYon,  no  trend     Correla/on  between  number  of  unique  authors   with  the  dura/on  (0.8)   •  No  people  retweeYng,  no  trend   The  number  of  retweets  for  a  topic  correlates   very  strongly  (0.96)  with  the  trend  dura/on   •  Many  influencers,  longer  trends   Nega/ve  correla/on  of  −0.19  between  the   domina/on-­‐ra/o  of  a  topic  to  its  trending   dura/on     (The  domina/on-­‐ra/o  for  a  topic  can  be  defined  as  the  frac/on  of  the  retweets  of   that  topic  that  can  be  aeributed  to  the  largest  contribu/ng  user  for  that  topic)   Trend  Dynamics  >  Twieer  Trending  Topics  >  The  anatomy  of  a  trend  
  81. 81. Twieer  doesn’t  generate  news,     but  does  filter  them   Social  media,  far  from   being  an  alternate  source   of  news,  func/ons  more   as  a  filter  and  an  amplifier   for  interes/ng  news  from   tradi/onal  media   Trend  Dynamics  >  Twieer  Trending  Topics  >  The  anatomy  of  a  trend  
  82. 82. (Twieer)  Informa/on  diffusion   Main  takeaways:   •  the  resonance  of  the  content  with  the  users  of  the  social  network  plays  a   major  role  in  causing  trends   •  a  majority  of  the  content  propagated  to  cause  trends  arise  from  tradi/onal   media  sources   Trend  Dynamics  >  Twieer  Trending  Topics  >  Informa/on  Diffusion  
  83. 83. Very  few  people  follow   or  are  followed  by  many  users   15  million  URLs  exchanged  among  2.7  million  users  over  a  300  hour  period   Both  the  in-­‐degree   (followers)  and  the  out-­‐ degree  (followee)   distribu/ons  have  /ght   log-­‐normal  fits   Trend  Dynamics  >  Twieer  Trending  Topics  >  Informa/on  Diffusion  
  84. 84. Twieer  is  a  small  world   15  million  URLs  exchanged  among  2.7  million  users  over  a  300  hour  period   Twieer  graph  is  a   small  world  with  a   mean  directed  path   length  of  3.61   Trend  Dynamics  >  Twieer  Trending  Topics  >  Informa/on  Diffusion  
  85. 85. The  majority  of  users  has  low  ac/vity   Trend  Dynamics  >  Twieer  Trending  Topics  >  Informa/on  Diffusion  
  86. 86. The  majority  of  URLs     have  no  resonance   If  your  URL  appears  in   10  different  users,   then  it's  more  popular   than  99.9%  of  the  rest   of  URL  tweeted   Trend  Dynamics  >  Twieer  Trending  Topics  >  Informa/on  Diffusion  
  87. 87. Only  a  very  few  tweets  has  several   cascades  of  retweets   RT-­‐cascades     (the  user  men/ons  the   source)     F-­‐cascades     (the  user  men/ons  the   source  and  is  its  follower)     Each  cascade  consists  of  one   or  more  subcascades.  The   number  of  subcascades   per  cascade  is  power-­‐law   distributed   Trend  Dynamics  >  Twieer  Trending  Topics  >  Informa/on  Diffusion  
  88. 88. How  big  are  subcascades?   For  each  cascade   the  subcascades  not   only  vary  in  number,   but  also  in  size.   Trend  Dynamics  >  Twieer  Trending  Topics  >  Informa/on  Diffusion  
  89. 89. If  your  tweet  is  interes/ng  is  more   likely  that  I  will  follow  you   The  distances  to  the  root  are   short,  even  when  compared   with  the  already  short  average   path  length  in  the  follower   graph.       One  hypothesis  explaining  this   data  could  be  that  when  a  user   receives  some  interes/ng  URL   along  an  path  longer  than  1,   then  that  user  is  very  likely  to   start  following  the  original   source  of  the  URL   Trend  Dynamics  >  Twieer  Trending  Topics  >  Informa/on  Diffusion  
  90. 90. The  more  users,     the  bigger  the  subcascades   The  URLs  tweeted  by  the  highly   connected  users  reach  large   audiences  and  are  likely  to  be   (re)tweeted  by  their  followers.   However,  the  causality  is  likely   to  be  bidirec/onal:  the  users’s   URLs  are  tweeted  more   because  they  have  many   folllowers,  but  also  they  have   accumulated  many  followers   because  what  they  tweet  tends   to  be  interes/ng  and  viral.   Trend  Dynamics  >  Twieer  Trending  Topics  >  Informa/on  Diffusion  
  91. 91. How  long  takes  to  make  a  RT?   The  diffusion  delay  taken  across   all  the  (u,  i)  pairs  in  the  social   graph  is  log-­‐normally   distributed  with  a  median  of  50   minutes   Trend  Dynamics  >  Twieer  Trending  Topics  >  Informa/on  Diffusion  
  93. 93. Trending  Topic  Classifica/on   Trend  Parametriza/on   TwiBer  Trending  Topic  ClassificaYon     Kathy  Lee,  Diana  Palse/a,  Ramanathan  Narayanan,  Md.  Mostofa  Ali  Patwary,  Ankit  Agrawal,  and  Alok  Choudhary     2011  11th  IEEE  Interna/onal  Conference  on  Data  Mining  Workshops     768  trending   topics  selected   randomly  from   the  +23000   collected   between  2010   and  2011  
  94. 94. Classifica/on  Tecniques   Trend  Parametriza/on   Comparison  of   classifica/on   accuracy  using   different   classifiers  for   network-­‐based   classifica/on     TwiBer  Trending  Topic  ClassificaYon     Kathy  Lee,  Diana  Palse/a,  Ramanathan  Narayanan,  Md.  Mostofa  Ali  Patwary,  Ankit  Agrawal,  and  Alok  Choudhary     2011  11th  IEEE  Interna/onal  Conference  on  Data  Mining  Workshops    
  95. 95. Classifica/on  Tecniques   Trend  Parametriza/on   Comparison  of   classifica/on   accuracy  using   different   classifiers  for   text-­‐based   classifica/on     TwiBer  Trending  Topic  ClassificaYon     Kathy  Lee,  Diana  Palse/a,  Ramanathan  Narayanan,  Md.  Mostofa  Ali  Patwary,  Ankit  Agrawal,  and  Alok  Choudhary     2011  11th  IEEE  Interna/onal  Conference  on  Data  Mining  Workshops     NBM  =  Naive  Bayes  Mul/nomial   SVM  =  Support  Vectors  Machines     (number  of  tweets,  top  frequent   terms)  
  96. 96.   “Online  Trends  Analysis”     by  Francesc  Gómez  Morales     is  licensed  under  a     Crea/ve  Commons  Aeribu/on-­‐ NonCommercial-­‐ShareAlike  3.0  Unported   License