Two government bodies, Innovate UK and UK Trade & Investment, host an annual conference showcasing some of the best new ideas in tech from across the UK. Events like this are an important part of how governments support innovation. We analysed Twitter data from the recent 2015 conference to measure the event’s global reach and networks between delegates. 

Imagine a future where our virtual selves network, shaking digital hands in cyberspace. The customary coffee and biscuits, mere simulations. But this is not the future, this is already happening (ok, not the food and drinks bit). We connect on LinkedIn rather than exchanging business cards. We can follow each other on Twitter without ever meeting face to face.

Nevertheless, face to face events continue to play a very important role in facilitating connections between people. They are more personal, resulting in closer relationships. They also perhaps allow a degree of serendipity that technology cannot yet offer- introducing us to people and topics we wouldn’t otherwise have come across. Though events continue to be mostly in person affairs, we see the increasing role of the online world in them. As we will show, conferences are playing out against a background of digital activity on social media. This digital footprint allows us to better understand these interactions.

Nesta has developed an approach to analysing the impact of events using the social media data generated at them, discussed in our report The Net Effect published last year (1). In this post, we use Twitter to understand the geographical distribution of people’s engagement with a major conference and measure the network of the social connections between delegates – illustrating how data and technology are likely to be increasingly central to the future of events.

To do this, we look at the UK’s Innovation and Trade agencies’ (Innovate UK and UKTI) annual conference, Innovate 2015, which showcases UK innovation. The conference took place November 9th and 10th in London’s Old Billingsgate fish market and brought together over 3000 delegates, ranging from foreign investors to government representatives to startup founders. The diversity of participants and topics covered, from agri-tech innovation to investing in Brazil, makes Innovate 2015 an interesting conference to analyse.

Mapping engagement: Putting all tweets on a map

Mapping the geographical locations of people who tweeted about Innovate 2015 can tell us a lot about the reach and influence of the conference. Are all tweeters London-based, in close geographical proximity to the conference venue, or do we see a wider audience- both within the UK and beyond?

We collected all tweets that used the conference’s official hashtag #innovate2015 in the days around and during the conference (2). This allowed us to capture Twitter users engaging with the event’s hashtag but who weren’t physically at the conference- engagements particularly indicative of impact and reach beyond the conference hall. We retrieved 6,795 tweets (3), of which 3,591 were retweets, created by 1,896 individual tweeters. Assessing how many of these individuals were actually present at the conference would involve comparing this list to a complete list of Twitter handles of registered participants. Because a full list of attendees’ Twitter handles was not available, we were unable to determine with certainty who was physically at the conference, and who wasn’t.

An analysis that could be done was looking at where those tweeting were based. The self-reported locations in the tweeters’ profiles (see below for an example) were used as a proxy for geographical location. Though not all locations were usable – we filtered out results that were too generic (“Europe”), or not actual locations (“Hiding on the internet”) – we still obtained 348 different mappable locations, for 62 per cent of our tweets. This provided a much larger sample of location information than geo-tagged tweets (which record a user’s exact location at the time of tweeting) as very few people enable that feature (4).

Screen Shot 2016-02-01 at 14.42.51

The text in the red box shows a user’s self-reported location.

In the maps below, we have mapped all the Innovate UK conference tweets and retweets, aggregated by locations (e.g all tweets in a city or region), separating out the UK (which accounted for 75 per cent of tweets) from the rest of the world. The size of the bubbles for each location reflect the combined number of tweets and retweets (as total tweets) generated by people based there; the bigger the area of the bubble, the more tweets. Color shades similarly reflect the number of tweets: the darker pink the bubble, the more tweets generated in that location. Hovering over a bubble reveals the location’s name and the number of associated tweets. Areas with a particularly high density of bubbles can be magnified by hovering over.

Map of tweets generated in the UK

Nesta… Mapping the geographical spread of tweets

All maps created using Google Charts.

Click image for live view
Click image for live view

Our data shows that Innovate 2015 reached a UK-wide audience. Though the highest proportion of tweeters still reported London as their homebase, with Oxford, Cambridge and Swindon (where Innovate UK and the UK research councils are based) proportionally not far behind, we see tweet activity all over the country, with Edinburgh, Manchester, Birmingham and the Highlands some other noteworthy clusters.

It is important to note that the map does not filter out people who tweeted a lot. Some of the unexpectedly large bubbles in smaller locations may be generated by a small number of very active tweeters. The rationale for including these individuals is that their activity shows engagement and potentially helped the conference reach new audiences on Twitter.

Map of tweets generated outside of the UK

Click iamge for live view
Click iamge for live view

This map uses a mercator projection, which is why Greenland appears comparatively  large.

Measuring global reach

An important goal of the Innovate conference is to attract an international  audience and so generate additional foreign investment in the UK. Though we are unlikely via Twitter  to pick up any of the deals agreed at the conference, we can still see which country’s tweeters engaged the most.

The United States was the country responsible for most tweets, and we saw sizable clusters of conversation in Italy, India and South Africa as well. It is important to note that all these countries were the subject of dedicated sessions on doing business in that country. Tweet volumes are not a perfect measure for foreign engagement: not every country uses Twitter (e.g. China) and, conversely, some have a disproportionately high density of Twitter users (which likely in part explains the United States’ performance). In the Twitter results, we also saw evidence of overseas companies sending representatives from their UK branch- accounting for a significant number of mostly London-based tweets.

Conferences are about engagement, but they are also about the social networks between people, which we look at it in the next section.

Visualising the event’s social network  

Older than the phone system, older than the roads, social networks are perhaps the earliest form of infrastructure, providing their members with access to information and resources. It’s therefore understandable that we use conferences to help stimulate these networks. The development of social media has brought a new dimension to networks by formalising them and creating new platforms for interaction and debate. This has meant that it is possible to obtain data on their structures that was previously very difficult to collect.

To study the social network among delegates at the conference, in collaboration with Innovate UK, we obtained the Twitter handles of 322 event delegates before the event (these were a mixture of institutional and personal accounts). We then reconstructed the social network between these attendees the day before and the day after the event. The network after the event consisted of 2,320 individual following connections (i.e. where one user follows another in the network), a 6.7% increase from the start of the conference. It is important to flag that this is based on a sample of around 10% of all delegates. The mapping analysis in the previous section shows there likely was a significantly larger number of delegates on Twitter at the event. The number of possible connections that can be formed at an event scales more than proportionately with the number of delegates, so this should not be taken as representative of the impact of the event as a whole.

Nevertheless, it is possible to use this data to learn something about the social network between delegates. The main network of following connections after the event is shown below (This omits , participants unconnected to anyone in the main following group). The nodes correspond to the Twitter accounts of the event attendees we have information on. The lines represent one of the accounts following the other on Twitter. Innovate UK is the most connected participant at the event, and so is the most prominent node in the network. The network is however cluttered and very hard to interpret. The ball of wool effect typical of such diagrams is much in evidence.


Note: Networks in post produced with Gephi.

To make the underlying network structure clearer we do the following:

  1. Focus on the reciprocal connectionsWe only map the reciprocal connections that existed between delegates on Twitter. That is, connections where one person follows another who also follows them back. This is arguably a more meaningful measure of connectivity than unreciprocated connections, for example an audience member following a keynote speaker who doesn’t follow them back. In the graph these reciprocal connections are now represented by a single line between delegates.
  2.  Omit individuals that are less connected in the networkIn social networks some people are less well connected, and so more    peripheral, to the overall network than others. Omitting them makes the diagram clearer, without losing too much useful information. We therefore remove people with only one reciprocal connection in the previous network. We also scale the size of the nodes by the number of reciprocal connections that a delegate had.
  3.  Highlight groups of people that are more closely connected to each otherWe run a clustering algorithm on the network which identifies groups of people that are more closely connected among each other than would be expected on the basis of chance, and colours them the same colour. This helps bring out different communities that exist within the network (5).

The resulting social network from the sample of delegates at the conference then becomes clearer, as shown in the annotated chart below. For privacy reasons  we have not identified any individuals, but highlighted some of the major public sector organisations involved and, in general terms, identified some of the communities in the network. Innovate UK is at the centre of the chart, surrounded by several distinct communities of delegates.


  • The top left-hand corner shows a cluster of  higher education and research institutions who are themselves more closely connected to each other.
  • The lower right-hand side shows Innovate UK’s connections to UKTI and the Catapults. The Catapults are a network  of independent not-for-profit centres, each focused on a specific area of rapid innovation, set up by Innovate UK to connect businesses with the UK’s research and academic communities.
  • The left-hand and right-hand sides of the chart show private businesses and venture capitalists.

This is a small proportion of the event network, and a smaller proportion still of Innovate UK’s wider network, but it does show that as the organiser of the conference, and the UK’s public innovation agency, this is broadly where one would expect Innovate UK  to be  positioned in the network, connected to the private, public and third sectors (6). This kind of social network analysis, and more detailed quantitative investigation, is likely to become more important in informing the curation of conferences, and for organisations thinking about their customers and audiences.


From people’s interactions on Twitter it is clear that the Innovate UK conference engaged a wide audience across the UK and around the world. From the analysis of the reciprocal connections on Twitter among  a sample of around 10% delegates that registered their Twitter profile before the event  we can see that Innovate UK sits at the centre of the network of conference delegates, between the public, private and higher education sectors. Strategically, the position we would expect it to occupy as the UK’s public sector innovation agency.

The growth of social media use at conferences has allowed us to undertake this analysis. This highlights a trend set to become more influential: the integration of online and offline in our interaction at events. The engagement of people on Twitter around the world reflects the range of delegates attending in person, but also suggests remote online engagement internationally. The Twitter connections formed at the conference between delegates may well have involved face to face networking, but in some cases will have been virtually mediated by Twitter.

The boundaries of conferences may blur further with the wider convergence of social networking and virtual reality. A trend that can be seen with Facebook’s acquisition of the VR platform Oculus Rift, and which has arguably been going on for several years with gamification and Massive Multiplayer Online Games (MMOGs).This is likely to be reflected in the  growing use of augmented and virtual reality in conferences (7). All of which will generate more data about events, in addition to that increasingly generated by event apps and sensor  devices. The conferences of the future will look for ways to combine data and our online and offline worlds to create environments that maximise their impact: Meet Space.


  1. Bakhshi, H., Davies, J., Mateos-Garcia, J. (2015), ‘The Net Effect:Using social media to understand the impact of a conference on social networks’.
  2. This was done using the Twitter Applications Programming Interface (API).
  3. In the days before the conference, two US events both used the same #innovate2015 hashtag. Cleaning out tweets that referred to these other events was a largely manual process. By looking at tweets’ locations and references to specific topics and speakers in the text of a tweet, we were able to identify and remove 276 of these tweets. Fortunately the other #innovate2015 events focused on very different topics and did not generate a lot of social media chatter. This does show the importance of picking an event with a relatively unique hashtag when doing this type of analysis.
  4. We  only collected recovered 19 geo-tagged tweets. Geo-tagged tweets also do not provide information about a tweeter’s usual location (indeed, all 19 tweets were sent from within the conference center).  
  5. The analysis was undertaken using the graph modularity algorithm in gephi. Broadly speaking, the algorithm partitions the graph’s nodes into a set of communities (i.e. each community being a group of nodes) that aims to maximise modularity of the graph. The modularity being the number of connections that fall within communities for a given partition relative to the expected number of connections within the communities if connections were generated at random. A community partition characterised by higher modularity is thus less likely to have arisen by chance.​
  6. Etzkowitz, H. (2003),’Innovation in Innovation: The Triple Helix of University-Industry-Government Relations’, Social Science Information
  7. These can for example involve monitoring the distribution of people at a conference through levels of event wifi usage or issuing badges to delegates that register meetings when they tap their  badges together. For a flavour of some of some of things currently being discussed in event technology see the recent conference programme for Event Tech live 2015.

We would like to thank Natalie Waugh and Dan Hodges of Innovate UK for their help on this project.

This blog was originally published on Nesta. Read the original blog

John Davies is an economic research fellow on the creative and digital economy at Nesta. You can follow him on Twitter at: @johnardavies  and Katja Bego is a data scientist in Nesta’s technology futures team. You can follow her on twitter: