TAM Ireland - opinion piece€¦ · COLOURTEXT)FORCRM ) COLOURTEXT)FORSOCIAL)INSIGHT...
Transcript of TAM Ireland - opinion piece€¦ · COLOURTEXT)FORCRM ) COLOURTEXT)FORSOCIAL)INSIGHT...
COLOURTEXT FOR CRM
COLOURTEXT FOR SOCIAL INSIGHT
COLOURTEXT FOR MARKET RESEARCH
!
!
!
TAM Ireland Project Execu@ve Summary !!!Author: Jason Brownlee !Date: 20th June 2014 !Contact: [email protected] ! +44 (0) 7970 626532 !!!
Colourtext Ltd | 48 Stramongate, Kendal, Cumbria LA9 4BD | 01768 881321 | Company Number 8248347
COLOURTEXT FOR CRM
COLOURTEXT FOR SOCIAL INSIGHT
COLOURTEXT FOR MARKET RESEARCH
Introduc@on to Colourtext !Colourtext bridges the gap between classic market research and the world of data analy9cs. We are leaders in Social Insight analysis and work with a range of blue chip brands and media companies including Samsung, Bloomberg, Fox TV, The Radio Adver9sing Bureau and TwiFer. !Colourtext has been at the forefront of the ‘Display Media + Social’ debate for some years. In 2010 we undertook a project (as Dollywagon) for the UK RAB called "The Online Mul9plier", which won Gold at the UK MediaWeek awards. More recently we have worked closely with TwiFer and contributed towards their successful 'promoted tweet' media proposi9on. !!Project objec@ves TAM Ireland, the official TV ra9ngs body for Ireland, commissioned Colourtext to undertake a project that would inves9gate the role played by TV in s9mula9ng social media discussion and other online behaviours. In par9cular, TAM Ireland wished to understand how TV ads drive brand-‐related online search and website visits and if TV successfully s9mulates discussion of brands in social media contexts. TAM Ireland asked Colourtext to focus on the following key objec9ves: !• Correlate the incidence of TV advert broadcasts with the frequency of visits to branded websites • Correlate the incidence of TV advert broadcasts with the frequency of brand men9ons on TwiFer • Is there a correla9on between official TAM Ireland TV ra9ng data and the frequency of TV show men9ons on
TwiFer? • Total number of Irish TwiFer users that tweet about TV content • Total propor9on of Irish TwiFer traffic that men9ons or relates to TV content • Total propor9on of Irish TwiFer traffic that men9ons or relates to TV content during the Evening Peak • Iden9fy the Top 10 TV shows and the main TV content genres that s9mulate most TwiFer men9ons • Iden9fy the most prolific Irish users that talk about TV content on TwiFer !We set out to implement an ambi9ous Big Data media research methodology. Our aim was to collect as many of the Tweets as possible (not just TV-‐related tweets) that were issued by Irish residents over the en9re month of April 2014. The numbers associated with this project are massive. !We used the TwiFer data hose (managed by Datasib) and TwiFer's public API (Applica9on Programming Interface) to collect an ini9al 25 million tweets, which added up to a heby 11GB of data. TAM Ireland provided complete logs of all TV shows and ad spots transmiFed across all Irish TV channels during the survey period. In addi9on, similarweb.com provided visitor site traffic data for 300+ Irish websites. !!� 2 !
COLOURTEXT FOR CRM
COLOURTEXT FOR SOCIAL INSIGHT
COLOURTEXT FOR MARKET RESEARCH
Summary of findings !In April 2014 Colourtext collected and analysed every tweet that could be posi9vely iden9fied as coming out of the Republic of Ireland, on behalf of TAM Ireland, the country’s official TV ra9ngs agency. To our knowledge it's the biggest and most comprehensive independent study its kind. The study’s findings are important for understanding other advanced consumer markets and media cultures like the US and UK. !!Methodology !Ini9ally a total of 25 million tweets were collected for the month of April. However, defining a tweet as 'Irish" via data that TwiFer makes publicly available is not a simple process. TwiFer users are not required to state their home loca9on -‐ to do so is purely voluntary. TwiFer itself provides no standard loca9on data fields for users to complete, meaning we get a bewildering variety of place descrip9ons for personal domiciles. For instance, it's fascina9ng to see how many people in France describe their home loca9on as "Voldemort's bed"!
!Our task was to capture data from TwiFer users who could be reasonably located within the Republic of Ireland -‐ this obviously implies filtering out users from Northern Ireland. We therefore adopted a series of rules that were coded into a data cleansing process, which defined our Republic of Ireland 'survey sample area'. This cut
� 3 !
COLOURTEXT FOR CRM
COLOURTEXT FOR SOCIAL INSIGHT
COLOURTEXT FOR MARKET RESEARCH
our ini9al haul of 25m tweets down to 15.6m messages authored by 170,000 ac@ve Twiaer users who could be reasonably located as resident within the Irish Republic. !It’s important to note our system detected significant numbers of Irish UIDs that could not contribute data to the project. Around 29% of the total Irish UIDs we detected were not ac9ve during April i.e. authored no tweets whatsoever. A further 8% of Irish UIDs were ‘protected accounts’ i.e. only selected followers can view their tweets, meaning all their messages are essen9ally ‘private’.
!We also discovered a significant number of tweets wriFen in languages other than English. This represented around 4% of all tweets authored in the RoI, with Spanish as the most popular non-‐English language and perhaps reflec9ng the increasingly mul9-‐cultural nature of Irish society. !!!
� 4 !
COLOURTEXT FOR CRM
COLOURTEXT FOR SOCIAL INSIGHT
COLOURTEXT FOR MARKET RESEARCH
General behaviours !Something we really wanted to discover was the main topics of conversa9on on TwiFer. To answer this ques9on Colourtext pushed a random 6% sample of tweets (1 million messages) through its general seman9c analysis framework. This revealed the Top 12 main discussion themes within the data -‐ in descending order of prevalence we found: !1. Sport 2. Bodies -‐ parts and aFributes 3. Food & ea9ng 4. TV 5. Music 6. Clothing & accessories 7. Brands 8. Medical 9. Money 10. Sleep & sleeping 11. Drink & drinking 12. Poli9cs !This reveals a kind of "hierarchy of needs" at a social level for the Irish na9on – sport, bodies, food, TV and music -‐ these are the topics around which we converse on a social level with others. Eugenia Siapera, chair of social media studies MA at Dublin City University, commented that determining a strong link between a social media outlet such as TwiFer and television made sense. !“What is being tweeted is what we deal with and talk about in our everyday lives and to expect TwiFer to be any different is counter-‐intui9ve. While poli9cs and social issues are also important they are overshadowed in social media by lighter, less serious maFers." !Tweets are oben wriFen when someone is alone, which oben makes it feel like an in9mate kind of communica9on, even though you're wide-‐cas9ng to a large group of friends and followers. It's therefore not surprising that people oben share comments about their own bodies or those of other people. The whole range of body-‐related tweets will cover anything from big ones, small ones, nice ones, smelly ones, aches, pains, scratches, itches and wishes. !This in9macy also includes frequent discussion about food and ea9ng, which oben reflects modern lifestyle paFerns. People talk about the food they're ea9ng or are looking forward to ea9ng. They also moan about diets and share their guilt over indulging in tasty treats. !!!
� 5 !
COLOURTEXT FOR CRM
COLOURTEXT FOR SOCIAL INSIGHT
COLOURTEXT FOR MARKET RESEARCH
TV & Twiaer !• 43% of Irish TwiFer users in our sample issued at least one tweet that referred to TV content during April
2014 • Sport accounts for 31% of all TV related tweets • 10pm is the 9me that twee9ng about TV peaks each day during the week. • The weekend peak in TV twee9ng happens earlier in the evening around 8pm
!The study's results reveal that paFerns of peak TV viewing and the 9mes when we use social media overlap very closely. It’s therefore not surprising that TV viewing and social media use oben go hand in hand with each other. !We also learnt that viewers using social media to read other people's live comments about a show, and some9mes contribute a comment of one's own, adds to the fun and sense of engagement an audience feels with a program. We expect this 'second screen' dynamic to play a bigger role in TV formats in the future. !The study also confirmed that the size of a TV show's official audience ra9ngs does not provide a reliable guide to the amount of interest it can generate on TwiFer. Moreover, some of the biggest trending TV shows on TwiFer feature preFy low down in the official TV ra9ngs league table. !If we consider the Top 100 ra9ng TV shows transmiFed in Ireland during April 2014 (represen9ng 24 different TV show franchises), there appears to be a rather weak rela9onship between the size of a show’s official audience ra9ngs and the number of tweets it generates (correla9on = 0.28). However, for the highest ra9ng shows (e.g. Top 25) we found a more solid rela9onship between audience levels and tweet volumes (correla9on = 0.6)
Colourtext Ltd | 48 Stramongate, Kendal, Cumbria LA9 4BD | 01768 881321 | Company Number 8248347
COLOURTEXT FOR CRM
COLOURTEXT FOR SOCIAL INSIGHT
COLOURTEXT FOR MARKET RESEARCH
!The most tweeted about shows on Irish TV in April 2014 were:
!The launch of the new WWE Wrestlemania season, followed by the death of Lucy Beale in Eastenders, also generated the highest peak of Irish TV tweets in a 5 minute period. This neatly reflects our finding that two different types of TV show generate the largest volume of tweets. !The first type are popular high-‐frequency formats like chat shows, reality entertainment and soaps. Tonight with Vincent Browne and The Late Late Show (both chat shows), Fair City (a soap) and The Voice of Ireland all generate a big reac9on on TwiFer !The second, and actually the most highly tweeted shows, are rela9vely niche 'passion' franchises like Game of Thrones and WWE Wrestlemania. A vibrant social media fan culture has grown up around these shows, reflec9ng deep passion and commitment. Previous studies by Colourtext and others have suggested such shows can be poor ra9ngs performers, but their long term content franchise value can be massive and oben lies beyond the scope of conven9onal adver9sing revenue streams. This study backs up those findings. !To illustrate, Star Trek's early TV ra9ngs in the 1960's were so poor that CBS threatened to pull the show aber the first series. However, an unprecedented leFer-‐wri9ng campaign by fans of the show stayed its execu9on. But aber just the third series Star Trek was axed, yet we know that's not where the Star Trek story ended. If we look back to the 60s, what was the beFer predictor of Star Trek's mul9 billion dollar franchise value -‐ the poor TV ra9ngs data or hearvelt leFers from its commiFed fan base? !The fascina9ng thing about modern social networks is they bring commiFed and passionate fan cultures to the fore. Colourtext's work suggests social data can be mined to seek out the next fan franchise that could deliver Star Trek-‐sized financial returns. !� 7 !
Rank TV ShowTwiaer men@ons in April 2014 Rank TV Show
Twiaer men@ons in April 2014
1 game of thrones 20714 8 britain's got talent 5923
2 wwe wrestlemania 17872 9 prime 9me 4563
3 tonight with vincent 10566 10how met your mother 4117
4 the voice of ireland 7624 11 sky sports news 3744
5 eastenders 7170 12 the masters 3618
6 the late late show 6321 13 rte news 3377
7 made in chelsea 6283 14 hollyoaks 3326
COLOURTEXT FOR CRM
COLOURTEXT FOR SOCIAL INSIGHT
COLOURTEXT FOR MARKET RESEARCH
!Rela@onship between TV adver@sing and social web behaviour !Colourtext used website traffic data provided by similarweb.com in combina9on with TAM Ireland TV schedule data to undertake a Pearson correla9on analysis to see if the transmission of TV ads (or a rise in the number of TV ads transmiFed) was reflected by growth in visitor traffic to the websites of adver9ser brands. In total we examined data for 318 branded TV campaigns booked by 108 separate adver9sers. We grouped these into 21 dis9nct product categories. !We found 4 adver9ser categories that exhibited a significant correla9on between the transmission of branded TV adverts and a growth in visitor traffic to a brand's website, with a further 4 categories exhibi9ng a milder, but s9ll significant, correla9ve effect. In descending order of strong correla9ve effect we find Retail (high street & online), Insurance, U9li9es and Confec9onary. These are the categories that seem to enjoy the most success in using TV adver9sing to drive visitor traffic to their websites. The bulk of website visits generated by TV ads for these categories seems to occur on the same day as ad transmission. !
!!We also undertook a Pearson correla9on analysis to see if the transmission of TV ads (or a rise in the number of TV ads transmiFed) was reflected by growth in men9ons for a brand on TwiFer. Correla9ng TwiFer brand men9ons with branded TV ads is inherently difficult. For instance, brands with names like ‘Sure’ or ‘Always’ are difficult to track because of their generic usage e.g. can you be 'sure' a keyword men9on is 'always' about Sure or Always? This ruled out a cateogry-‐wide analysis in favour of a brand specific analysis. !We therefore chose a small basket of 7 brands with less ambiguous names and high numbers of TwiFer men9ons for this test. We found significant correla9ons for TwiFer brand men9ons occurring on the same day as ad transmission for 3 brands. We also found 3 cases where brand men9ons peaked significantly 24hrs aber ad transmission. This happened with the new movie release ‘Divergent’ (correla9on = 0.7) and for Nevlix (correla9on = 0.4). In both cases we find people discussing their experience of the content, sugges9ng the TV ads s9mulated consump9on followed by social commentary. !
Adver@ser category Same day 24hr offset
Retail high st online 0.705 0.598
Finance insurance 0.608 0.268
U9li9es 0.567 0.279
Confec9onary 0.487 0.351
FMCG food snacks 0.476 0.321
Charity 0.468 0.319
Travel 0.447 0.419
� 8 !
COLOURTEXT FOR CRM
COLOURTEXT FOR SOCIAL INSIGHT
COLOURTEXT FOR MARKET RESEARCH
!
Colourtext Ltd | 48 Stramongate, Kendal, Cumbria LA9 4BD | 01768 881321 | Company Number 8248347
Brand Twitter brand mentions
Correlated with TV ads on day of transmission
Correlated with TV ads 24hrs after
transmission
Tesco 48,447 0.41 0.32
Netflix 8,647 0.29 0.40
Divergent 6,786 0.47 0.70
Heineken 6,392 0.33 0.38
Coke 4,544 0.31 0.32
Mercedes 2,131 0.28 0.12
Lindt 1,666 0.37 0.31
COLOURTEXT FOR CRM
COLOURTEXT FOR SOCIAL INSIGHT
COLOURTEXT FOR MARKET RESEARCH
� 10 !