Päätin (/sain aikaiseksi) vihdoin julkaista graduni Verkkoja ja innovaatioita – sosiaalisen verkoston analyysi organisaation viestinnän tutkimuksen apuna. Gradu on vuodelta 2009 ja tein siinä kyselytutkimuksen erään elintarviketeollisuuden jalostajalla. Vastaajia oli nelisenkymmentä.

Organisaation tiedonkulun verkosto.

Tutkimuskysymykset (joihin vastaukset löytyvät gradusta):

  1. Minkälainen on Yrityksen sisäisen kokonaisverkoston  sosiaalisen verkoston rakenne?
    a.    Muodostuuko Yrityksen neuvonhaun ja tiedonkulun verkostoon keskittymiä, mielipidejohtajia?
    b.    Ovatko mahdolliset keskittymät korkealla organisaation hierarkiassa?
  2. Kenen toimijan kautta tieto innovaatioista tulee verkostoon sisälle?
    a.    Muodostuuko tiedonkulun verkostoon yksi vain useampi silta (mielipidevälittäjä) eri ryhmien välille Yrityksen ulko- ja sisäpuolelle?
  3. Jos Yrityksen sosiaaliseen verkostoon muodostuu mielipidejohtajia ja mielipidevälittäjiä, ovatko nämä samoja henkilöitä?
    a.    Jos eivät ole samoja henkilöitä, ovatko mielipidejohtajat ja -välittäjät yhteydessä toisiinsa?

Tutkimusta oli mukava tehdä, ja samalla polulla tässä ollaan edelleen. Tosin polku on vienyt vahvasti yrittämiseen ja sosiaalisen median analyysiin. Tekstissä on liitteenä suomennettua verkostoanalyysin sanastoa sivulta 93 lähtien, ja mallit verkostokyselylomakkeesta.

FIFA World Cup 2014, the biggest sporting event in four years (sorry Olympics) is starting today. The tournament holds 736 players from 32 countries. When the players are not playing for their national teams, they play in 301 different clubs. Players from different national teams meet in these clubs. For example, Manchester United has players from 9 different national teams. This means that players in the World Cup who play in Manchester United know players from at least eight different national teams. Why is this important? If two players belong to the same team (national or club), they have a social connection. Using social network analysis we can analyze and visualize this connection (examples from UEFA 2012 and World Cup 2010 similar networks). So, here’s the social network of FIFA World Cup 2014:


In the picture above is the player-to player-connections (here’s a pdf with a better resolution). The size of a players name represents the total number of other players he shares a club with. The top players, Antonio Valencia, Javier Hernández, Julian Green, Shinji Kagawa, Robin van Persie, Nani, Arjen Robben, Mario Mandžukić, Patrice Evra, Xherdan Shaqiri, Daniel Van Buyten, Dante and Javi Martínez have all 13 club mates in the tournament. The color of the player is determined by a computer program that detects clusters. Most of the clusters are the same as the national teams, but we notice that in the middle, the line between Spain, France, Brazil etc. start to become blurry. To get a better picture of the most connected players, below is the core of the network: top 10 % of the players.

worldcup_players_coreWe can also visualize the connections between different national teams and the clubs.


A line between a club and team is formed when a player plays in both of them (pdf). The more a national team has players from a specific team, the thicker the line. For example the German team has seven players from Bayern Munich and Spain has seven players from Barcelona. The most diverse teams are Algeria and Nigeria whose players come all from different clubs. To clarify the situation, below is a picture of the teams and clubs that have at least two common players. Four countries have at least two players from Napoli.

worldcup_players_core_v3Most interesting finding for me is that all of the players are connected one way or another. On average, two players have less than three steps between them. It would be interesting to extend this analysis to the players’ previous clubs and see how the social network of past five years would look like.

EDIT: For those that are more of DIY type network analysts, here are the network files (GraphML): player-player network and team-club network.

Twitter: jattipaa


Data from Wikipedia. Visualizations with Gephi.

A couple of years ago – and still today – any unexplained phenomenon that included social media would simply be named “viral”. Virality means that idea/news/meme starts to spread without the help of the original source. This spreading happens by the infectious nature of the idea or by the influence of those who have come to contact to the viral agent. This has been both the dream and the nightmare of the marketing and PR departments. Last week the nightmare scenario occurred to the The Copyright Information and Anti-Piracy Centre in Finland (CIAPC).

In the spring 2012 the CIAPC contacted a Finnish man claiming copyright violations in a P2P network. Long story short: After some threatening emails by the CIAPC and the man’s refusal to pay a compensation the police performed a search and seizure operation. The seizure part included a 9-year old girl’s Winnie the Pooh labeled laptop. Last week, the man described the situation in a Facebook post (original in Finnish) that suddenly started spread through the Facebook’s share-function. Soon the main stream media caught on (first online article in Finnish) and by the end of the next day in even made its way to international media (TorrentFreak and TechCrunch). The result: a PR disaster for the CIAPC.

The original post has at the moment 2 655 Facebook shares. The video below is the dynamic network of the public sharing of the post. The data was gathered by Mikael Rekola using the 99analytics.com  social media analytics platform. Each line represents a share of post, not the post views.

The video visualizes how the original post was shared and re-shared in the first 48 hours of its publication. Largest portion of the shares happened before middle part of the video within the first 24 hours. Actually, 97 % of the public shares happened during the 48 hour period. Those users whose shares got the most re-shares joined the game early.  This means that that the time is of the essence if the PR departments wish to react to these events.

When we look at the time scale of the spreading, we notice a sharp increase in the shared posts between 8 and 11 A.M..  The cascade started from several sources, including politicians  Dan Koivulaakso and Jyrki Kasvi,  who shared the post around 9 A.M. The first news story increased the shares of the original post. Before that, most of the shares were re-shares. The role of main stream media in facilitating these viral phenomenon cannot be forgotten.

The picture below that holds all of the shares. Size of the node represents the amount of shares the user caused and the brightness of the color of the node represents the amount of comments and likes the share received.

Almost all of the shares became directly from the original post (73 %). Less than 4 % of the shares reached some sort of virality i.e. spread beyond the first sharer.The longest chain of shares expanded for 5 steps. ( In Facebook, users are in average 4 steps from one another). We also checked if a friend relationship is present in the shares. Of the public shares, only 14 % of sharing happened between friends. But in the second degree shares, friend relationship was present in 43 % of the shares. The long sharing chains are actually quite rare (pdf) and awareness doesn’t require sharing: for every share there are tens or hundreds of share views.

The power of main stream media to spread a message is unparalleled: the power of social media comes from the power of amplification and raising awareness.

Twitter: jattipaa



EDIT: added some details on what the lines between the nodes represent.

When we look in to our past, we see different communities: schools, companies, teams etc. Most important thing in these communities is that we meet people; we network. One big community, UEFA European Championship 2012 is nearing its final: Spain vs. Italy.

Like all of us, the players in these teams have a past of different communities. In this case, the most interesting communities are football clubs. Like in my previous post of all of the teams, I ventured a guess: if two players have played in the same club, they share a connection (in this post they share a connection if they have played in the same club in the same year). Combine all those connections and you get a social network of the players.

Size of the players name represents the betweenness of the player: how important the player is to the experience flow of the team. We see that Pepe Reina is the most important player from the network perspective. Second and third are Christian Maggio and Álvaro Arbeloa. We see that most of the players are connected. Note: I only took in to account the prime league clubs.

The network structure of the both teams are similar in their density: both utilize about 20 % of all possible ties. Spain, however form almost twice the amount of cliques. This translates as more tightly connected groups. Data from team Spain from here and from team Italy here. Interesting is that Thiago Motta and Mario Balotelli are more connected to the Spanish team than to the Italian team. This may translate to better knowledge of the opposing team. The Spanish team does however have more players that have played together for several years. In other organizations this usually leads to better performance, so based on social network analysis, my money would be on Spain.

twitter: jattipaa


facebook: Verkostoanatomia

Visualized with Gephi.

The UEFA European Championship 2012 has some interesting statistics. For example its 16 teams have 368 players from 137 clubs. This means that in average each club has almost 3 of its players on the tournament. But averages are a poor tool when you can use social network analysis.

Like in 2010 World Cup the teams, clubs and players form a social network. In short, if a two players belong to the same team (country or club), they have a connection and this connection can be visualized. For similar visualizations, check out this viz from a Finnish newspaper Helsingin Sanomat (I also got the raw data used to do this analysis from the site).

The picture below is all the players, teams and clubs in the same network. The size of the node represents the amount of connections the node has.

Bayern München is the best represented club: it has 13 players in the tournament.

The next picture has only the clubs and teams. They are connected through players. The size of the node represents the “importance” (or eigenvector centrality) of the node. This means that the players from the Netherlands team are best connected throughout the network. Surprisingly Sweden is second. As the number of common players increase, so does the thickness of the line between the club and the team. They also get closer they get on the map. For example Spain and Portugal both have lot of players from Real Madrid and these two countries end up close to each other on the map. From the clubs’ perspective the players from Arsenal, Bayern München and Manchester City are central to the network.

To take the analysis even further here is the network of all the players in the tournament. A connections here means a shared team or a club membership. The size of the node represent the importance (or a good position) of the player. The color represents the “community” of players; a clique of sorts.

Franck Ribéry and Arjen Robben are the most well positioned players in the network. Every player can reach every other player in max 5 steps. A more clearer picture emerges after we remove all the players that play with less than 27 other players (22 from their team + 5 from the club). A sort of crème de la crème of players.

For you SNA enthusiasts here is the .net file of the players, teams and all of the actors. Visualized with Gephi.

twitter: jattipaa


facebook: Verkostoanatomia

Verkostoanalyysilla voidaan selvittää paljon monimutkaisia ja piileviä kuvioita. Yksi esimerkki on yritysten hallitusten väliset verkostot. Maailmalla tästä on ihan hienoja sovelluksia kuten They Rule.

Tällaisten verkostojen tekeminen ei välttämättä ole vaikeaa, se on vain työlästä. Tässä on ohjeet siihen, miten itse voit visualisoida esim. hallitusten välisiä verkostoja. Esimerkkinä käytän Kauppalehdessä 26.3.2012 tekemääni analyysiä valtionyhtiöiden hallitusten verkostoista, mutta menetelmää voi käyttää muissakin verkostoissa.

twitter: jattipaa


facebook: http://www.facebook.com/pages/Verkostoanatomia/189756439160

EDIT: English summary as the Google’s translation makes me look like an illiterate git. Mapping the Facebook friendship network and visualizing it with Gephi. The size of the node represents the amount of friends the MP has amongst other MP’s and the color represents the party.

Nyt kun presidentinvaalit ovat ohitse ja sosiaalisen median vaikutukset poliittiseen kampanjointiin on punnittu, voimme siirtyä suomalaisen arjen olennaisimpaan peruskysymykseen: kenellä eduskunnassa on eniten kavereita?

Usealla kansanedustajalla on joko henkilökohtainen profiili tai oma sivunsa, jonka kautta he pitävät yhteyttä äänestäjiinsä. Näistä profiileista ja sivuista raportoinut on esimerkiksi Verkkouutiset. Itseäni kiinnostaa pelkkien sivujen olemassaolon lisäksi se, miten kansanedustajat ovat toistensa kavereita. Jos tiedät kahden henkilön Facebook-id:n, voit kysyä Facebookilta, ovatko nämä kaksi kavereita keskenään. Tämän tiedon avulla voin hakea kaikkien Facebookissa olevien kansanedustajien kaverisuhteet toisiinsa.

Kuvassa pallon koko vastaa kavereiden lukumäärää ja väri puoluetta. Viiva edustajien välillä tarkoittaa kaverisuhdetta. Voimme todeta, että Arto Satonen (kok) voitti eduskunnan kaverikisan 124:llä kaverillaan. Miapetra Kumpula-Natri (sd) oli kakkonen 111 ja Merja Kyllönen (vas) täpärä kolmonen 110:llä kaverillaan. Jos ei muuta voi sanoa, niin nämä kolme ovat kansanedustajien keskuudessa ainakin aktiivisimpia kaverikutsujen lähettäjiä.

Keskimäärin kavereita edustajilla oli 42 kappaletta ja kaveriverkoston halkaisija (eli kuinka monella askeleella kaksi kauimmaista edustajaa tavoittavat toisensa) oli neljä askelta. Puoleilla keskimäärin kavereita: sd 56, kok 50, kesk 42, vihr 42, vas 40, ps 30, kd 25, r 20. Siis RKP, mars kavereita kyselemään!

Verkostoanalyytikolle mielenkiintoista on edellämainittujen lisäksi puolueiden väliset kaverisuhteet. Kuvassa olivat kansanedustajat sijoittuneet selkeästi kaveriporukoihin. Mitä keskemmällä henkilö on, sitä useammassa kaveriporukassa tällä on kavereita. Kaveriporukat näyttävät muodostuvan puoluerajojen mukaisesti.  Silmiin pomppaa RKP:n ja Perussuomalaisten kaverisuhteiden puute.  Alla olevassa kuvassa on visualisoitu puolueiden välisten kaverisuhteiden lukumäärä (viivan paksuus ja luku vastaavat kaverisuhteiden lukumäärää).

Määrällisesti demareiden ja kokoomuksen välillä on eniten kaverisuhteita. RKP ja Perussuomalaiset eivät ole kovinkaan läheisissä väleissä keskenään, mutta Vihreiden ja KD:nkään välillä ei liiemmin kaveripyyntöjä ole lähetelty.

Kiitos Teemo Tebestille, joka auttoi kaivamaan kansanedustajien yhteystietoja. Jos joku haluaa itse kokeilla kansanedustajien kaverisuhteiden visualisoimista, tässä tiedosto. Verkostot on visualisoitu Gephillä, ja apua visualisointiin voi hakea näiden ohjeiden loppupäästä.

Jos kiinnostaa tietää enemmän, mitä Facebookissa ja sosiaalisessa mediassa tapahtuu, käykää kirjautumassa Sometrik-palvelun betaversioon (maksuton eikä vaadi mitään asennuksia).

twitter: jattipaa


facebook: http://www.facebook.com/pages/Verkostoanatomia/189756439160

Edit: I’ve jumped the gun: the Markkinointiviestinnän viikot #-tag was changed. This is the new version. Note self: check the #-tag beforehand…

Everything seems to happen sporadically . This is also true in Finland as the Mindtrek event in Tampere and – aptly named two day event – Markkinointiviestinnän viikko (Week of marketing communications) in Helsinki coincide in the same week. Luckily one is able to follow both via twitter with the #tags #mindtrek and #2011mvv   (+  #mvv2011). Moreover we can visualize the conversations happening in both. Do they overlap, which event activates more people and conversation and who orchestrates the communication etc? In this picture I have combined the tweets of the two events (at 2.20 PM).

The two events seem to overlap but two distinctive groups are formed: Mindtrek on the left and MVV on the right. There are total 446 tweeters combined.  The size of the node represents the bridgespanning role of the person: the bigger the node the between the two groups tweets the person is. We see that   @vesilola, @pauliinamakela, @mindtrek_ and @arimarjamaki ‘s tweets connect the two events.

Looking at the two events separately we can detect the most mentioned people. The size of the node represents the amount of tweets received and the color of the node the persons activity in mentioning others.

In Mindtrek there are at the moment 187 different tweeters. @mindtrek_@vesilola are the most mentioned while @mindtrek is the most active.

In Markkinointiviestinnän viikot there are at the moment 289 different tweeters. @vesilola, @socialdistrict and @greenpeacesuomi are the most mentioned while @pauliinamakela and @eisoma the most active. The overall structure seems more dense than Mindtrek’s so there seems to be more going on in Helsinki.

This is a a snapshot of the situations. If you want to do similar analysis yourself instructions can be found in these slides. The tweets were gathered with NodeXL and visualized with Gephi.



Visualizing Twitter networks can make sense in the constant stream of tweets. We can detect different communities or active Tweeters. I have made a quick hands on guide for twitter hashtag visualization using Gephi and NodeXL:


twitter: jattipaa


facebook: http://www.facebook.com/pages/Verkostoanatomia/189756439160

I found out of the horrible Norway attacks an hour after the bombings in twitter. The shootings and extent of the murders were revealed a while later. Usually I gather the tweets for network visualization with NodeXL‘s script, but this time I offer only the dataset for researchers.

The data is 24 hours of tweets gathered every 5 minutes containing the word “oslo” from 4.35 PM (CET) Friday 22nd July. I cannot guarantee that every tweet is in this dataset. The file format is GraphML (you can use text editor to find the term “Tooltip” where the content of the tweet is).


Let me know if you can make use of the data.

twitter: jattipaa