Client’s order

Community analysis formed on Twitter around the bookshop’s account in order to detect the factors that may explain its followers’ following-followee reasons on this social network.

Network Outsight’s suggestion

  1. Semantic analysis added to the users’ descriptions;
  2. Following-followee networks analysis among followers. Community analysis with quantitative and qualitative approximations.

Set up goals

  1. Assess Twitter’s own community’s efficiency;
  2. Find influencers that may contribute to popularise a product or event for the client;
  3. Get to know emerging opinions and necessities in the community;
  4. Assess the efficiency of the employed digital strategy;
  5. Optimise the Community Manager’s strategy;
  6. Prepare the commercial actions on the different market niches that form the community and target them according to similarity.

Obtained results

  1. A digital strategy audit based on the report of the work made and the results obtained. It included the work carried out by the bookshop until then, plus the improvement suggestions to employ from then on;
  2. Knowledge of the characteristics of the clients that belong to different subcommunities that form the community constituted on Twitter around its brand;
  3. Definition of the proper constant communication with the client based on the knowledge obtained from him/her;
  4. Search for the proper objective audience to carry out specific special offers depending on the interests expressed on the networks;
  5. Guidance for the organization of the cultural events that the bookshop carries out as a distinguishing feature from the competition: determination of the type of events of interest for its community; decisions making on the form, frequency and content of the messages to be communicated; discovery of the market niches to be addressed depending on the content of the event, etc.;
  6. Location of profitable niches to work on in order to manage the sale of its products. For instance: thanks to the already-made work, we found a market niche that was very difficult to fins without the community analysis. It is the romantic novel readers niche. These readers do not identify themselves as romantic novel readers in the shop or offline. They tend not to talk about their readings and it is not always possible to identify them. Thanks to the free and spontaneous associations that take place on the networks, we can discover such a profitable niche. With them we can take advantage over the competition before a large group of potential book purchasers.


After obtaining the data concerning the following-followee relations among the bookshop’s followers, we employed the following analysis oriented to understanding the reasons that motivate the associations between the Twitter accounts.

1. Semantic analysis of the users’ descriptions

A first key element is to know how Twitter users describe themselves. We detect the most common words in the descriptions given by the users; their frequencies, and we delete words with very low informative content (articles, prepositions, adverbs, some verbs, etc.) in order to draw the following words cloud. The following are the 100 most common words used in t%%he descriptions of the accounts that follow the bookshop, and its size corresponds to its use frequency:

A priori everything seemed to indicate that the community was very well structured. The bookshop followers’ interests had to do exactly with what the bookshop sells. However, we wondered what followers’ particular characteristics could offer a hint on how to improve a strategy that was already good enough.

2. Following-followee analysis among the bookshop’s followers

People associate with each other on the networks according to our interests. What is known as homophilia in scientific literature is the tendency we have to relate with people similar to us, according to age, gender, cultural characteristics, etc. There are other kinds of relations on Twitter, but the most powerful is the association depending on interests and passions, making those people that get moved by the same things end up getting nearer between them, even after a heterophilic exercise addressed to know new people on the net.

On the graph above, the Twitter accounts with intense following-followee relations have been located close to each other. The bigger nodes size is, the larger the following-followee relations received within the graph are, regardless of its followers’ number, this way its power can be observed contextually. The colours have been randomly assigned with a community detection algorithm that has provided 5 communities.

3. Quantitative analysis

In order to avoid misinterpretations and observe only the relations established among followers, we delete the bookshop’s account and all those accounts that did not follow anyone but the bookshop. Behind the filters, the net is composed by 4628 nodes and 344,591 edges; a total sum of 4628 Twitter accounts that kept 344,591 following-followee relations with one another. It happens to be a very dense virtual network, a very significantly well cohesive community.

After analysing some of its formal characteristics with statistical tests, we found out that:

  1. In the network Twitter accounts are united by 2.56 separation grades. The maximum distance between the furthest nodes is only 8 grades. This means, together with its high level of density, a great cohesion in the network. It can be called ‘community’;
  2. Very few accounts are followed by lots of accounts. It is a very centralized network in terms of ‘followings’ receipt, something that implies very strong leaderships;
  3. In contrast, the network is very decentralised in terms of ‘followings’ issue. This means that the community is uniformly proactive when following the network ‘leaders’; that implies a great interest in the information that they provide;
  4. We also found out that there is a very significant number of powerful nodes in terms of intermediation that is related to the network density. This makes its power very weak in general terms. There are no nodes that may be found in key mediating positions, because the majority of them are very well connected with the network. The important leaderships of the network are opinion leaderships, no intermediation leaderships.
4. Qualitative analysis

Perhaps the most important (and interesting) part of this order consisted in observing differences between the found communities. For this, we studied separately what the characteristics of the opinion leaders of each community were, and we reached the following conclusions:

  1. Yellow subcommunity: big Spanish publishers, big bookshops and online services for readers. They concentrate the strongest leaderships in the network;
  2. Pink subcommunity: bookshops, bookshops associations and small publishers. Their leaderships are not so strong as in the previous subcommunity;
  3. Blue subcommunity: independent publishers, specialized magazines and artists’ and writers’ associations. They represent the sidestream scene;
  4. Grey subcommunity: literary bloggers, offline referential proponents and institutions from the bookshop industry. This is the most cohesive community;
  5. Orange subcommunity: literary and political celebrities, institutions and cultural services. This is the most dissonant cluster in relation to the network;
  6. Green subcommunity: publishers and children’s and teen literature consumers.

As we see, although the network was generally very dense, we were able to detect its particularities and identify the mutual association motivations of big quantities of users. We also replicated the algorithms of community detection to find more smaller groups. From this, different communities of great commercial worth emerged. For instance, within the green subcommunity we could distinguish between parents that were children’s literature consumers and youngsters that were comic and manga readers. In the yellow community we could identify readers by genre: detective novel, romantic novel, etc.

It allowed our client to organise his followers according to his interests and to design digital communication strategies oriented to providing his followers the exact type of information that they wish: informing them on new publications, events of their interest, or even using microtargeting to make campaigns on social networks.


If you wish to have more information on the applications of all this work on your company, brand or product, please, feel free to contact us:




Icons made by Freepik on licensed by CC 3.0 BY

Uso de cookies

Este sitio web utiliza cookies para que usted tenga la mejor experiencia de usuario. Si continúa navegando está dando su consentimiento para la aceptación de las mencionadas cookies y la aceptación de nuestra política de cookies, pinche el enlace para mayor información. CERRAR