Social Insight

Twitter’s verdict on Lorde: impressive new singer-songwriter

Leave a Comment

Lorde is a new singer-songwriter from New Zealand, just 16 and she’s ignited a fire on Twitter.

Colourtext algorithms analysed 1 million tweets about Lorde to find the substance behind the hype, and see if it can last.

This video brings our analysis to life and illustrates the kind of insights that lie buried within social data.

To learn more about how Colourtext analyses social data from sources like Facebook, Tumblr, blogs and Twitter, please get in touch with

Sentiment analysis: Machine and man unite to assess brand value

Leave a Comment


Colourtext’s sister company (and forerunner) has been featured in a new Financial Times article about sentiment analysis.

Our CEO Jason Brownlee was interviewed by FT journalist Michael Dempsey to discuss the challenges of analysing social web feedback about brands. The key focus of the article is on the strengths and limitations of sentiment analysis.

You can read the article in full here on the FT’s website. But if you don’t have time or get frustrated by the FT’s paywall, here’s a brief excerpt:

“[Colourtext] will embark on a mission to inspect the marketplace around a product or brand as defined by social media.

[Colourtext’s] software produces a report that is then further analysed by a human. “You have to understand the dividing line between the computer and the material,” says Mr Brownlee. “Sarcasm does not compute.”

After logging mentions in a huge mound of tweets, his software breaks the content down into words that relate to elements of opinion.

This material is then subject to the human touch, with [Colourtext’s] analyst team using their own qualitative research system to herd comments into areas indicating a joke is being made or dissatisfaction expressed.

“I don’t trust a machine to tell me how people feel,” says Mr Brownlee, “but I do trust it to tell me what they are talking about.””

To learn more about how Colourtext analyses Twitter data, and what we can reveal about your brand, please email

Twitter vs TV Ratings – when media research and social analytics collide

Leave a Comment

Breaking-Bad 570

The recent ‘Breaking Bad’ finale ranked as the most-tweeted show in the US, but it didn’t even break into the top 10 of the official TV ratings list. What gives?

Our friend Matt Hart (@kiwihart) at Innovation Pipeline sent through a great article from Wired Magazine about the relationship between levels of Tweeting around TV shows and official Nielsen TV ratings. If you get time check it out – it’s the best we’ve read on this issue to date.

If you don’t have time here’s a low-down: TV ratings giant Nielsen has discovered a statistically significant link between tweet levels and audience ratings, but the correlation is a long way from perfect.

In short, shows like NCIS get massive official TV ratings but generate very few tweets. On the other hand, the Breaking Bad finale ranked as the most tweeted show in the US but didn’t even break into the top 10 of the official TV ratings list.

Factors like differences in the age profile of a TV show audience compared to that of Twitter’s user base can explain why tweet levels and TV ratings don’t often align closely.

However, Colourtext’s semantic analysis of Twitter data suggests there may actually be a firmer relationship between tweet levels and the content franchise value of a TV show.

Our best shot at defining a ‘content franchise’ goes like this. In the 1960s early seasons of Star Trek achieved mediocre TV ratings. Only the discovery of a vocal cult audience following preserved the show, and eventually led to its development as a multi-billion dollar TV, movie and merchandising content franchise.

The key point is there was nothing in Star Trek’s early TV ratings to predict the long term success and dollar-value of the show as a content franchise. Only the passion of a committed niche audience could have betrayed Star Trek’s future potential.

Which is where things might get interesting with social platforms like Twitter. We now get instant and continuous ‘mega-qualitative’ feedback on TV shows from Twitter. The problem, of course, is reading through the deluge of feedback.

Colourtext’s mantra of “Sarcasm does not compute” has shaped our approach to this challenge. We’ve developed a careful process for semantically analysing Twitter messages about TV shows and other brands, which can reveal a variety of insights, including:

  • – The nature and degree of impact generated by specific on-screen talent or dramatic characters
  • – Brand franchise DNA (the show’s character & innate appeal)
  • – Levels of anticipation or prediction about the outcome of specific plot lines
  • – Levels of engagement or antipathy with specific plot lines
  • – The quality and variety of emotions and themes expressed about the show
  • – Appraisals of individual actors or performers


To wrap up, we think official TV ratings measure one aspect of a TV show’s commercial value – it’s ability to generate advertising revenue.

On the other hand, Twitter message levels might point to a broader set of TV show attributes that signal longer term content-franchise value.

To learn more about how Colourtext analyses Twitter data, and what we can reveal about your brand, please email

Why Dove Real Beauty sketches video is the most watched of all time

Leave a Comment

Dove sketches 3

Colourtext Analyses Social Chatter Surrounding The “Most Watched ad Ever”.

The viral sensation sweeping screens all over the world in May this year has now been confirmed as the most watched video ad of all time. Dove’s “Real Beauty Sketches” ad released in May 2013 has knocked Evian’s “roller babies” from the top spot of most viewed video ad ever.

The Dove Real Beauty Sketches ad has been described as an inspiration to millions of women all over the world. Colourtext was commissioned by social network giant Twitter to find out why.

Twitter’s Head of Planning in the UK, Oliver Snoddy, commissioned this project and commented:

“Colourtext provided Twitter with a scalable data analysis solution that enabled us to quickly understand why so many of our users loved and shared Dove’s ‘Real Beauty’ campaign.

Colourtext’s system yielded deep insights into the contribution made by Twitter to the propagation of Dove’s commercial across the web. When a deep analysis of social conversations is needed, Colourtext has a compelling solution.”

Colourtext has developed a world-leading text mining system that automatically ‘reads’ social media, market research and customer satisfaction messages. It reveals what customers think or feel about brands and how perceptions translate into behaviour.

Colourtext’s technology was able to define audience reaction to Dove Real Beauty Sketches by mapping comment and reaction on Twitter  to reveal key audience themes and how they relate to each other.

Colourtext revealed that female Twitter users “hyper-recommended” the video and embraced its clever use of CSI-style forensics. Dove’s target female audience strongly urged, and in some cases appealed, to other women to view the online ad. Reactions to the ad on Twitter went far beyond conventional ReTweeting, suggesting viewers embraced and internalised the ad’s message in a deeper way.

To find out more about Colourtext’s analysis of Dove’s ad campaign you can download our Dove Real Beauty Sketches case study here.

If you would like to get in touch with Colourtext please email

Social analytics are rubbish – discuss

1 Comment


To date, the strongest driver for social analytics investment has been to engage in rear-guard fire fights when bad karma about a brand gets propagated through social networks.

A whole industry of dashboards, social metrics and stat analytics has sprung up to service this need. Indeed, an interesting new form of Social-CRM is beginning to emerge.

That’s fine so far as it goes, but there’s a general suspicion it severely underestimates the full contribution that social listening data could, and should, make to business success.

We often hear marketing, research and communication professionals dismiss the producers and users of these tools as mere “number-wangers”.

This indicates that Social Listening, and the social analytics industry that promotes it, is getting something badly wrong. There are lots of viewpoints on this issue, but here’s ours.

People who use buzz monitoring systems continue to stare at reams of natural language messages and text, shovelled faithfully onto their PC screens by social analytics platforms, and still wonder, “what does it all mean, what are the narrative around my brand?”

We’re not saying the social analytics emperor is wearing no clothes, but it feels darn close sometimes. The intuition, the overriding faith, expressed by most social analysts I speak to is that something useful must be buried away under all that text data. But no one is really feeling the love yet.

People want ‘insights’, preferably ‘actionable’ insights if they can get them. So, what does that mean?

To find out more about Colourtext, you can download our presentation “Introduction to Colourtext” here.

If you would like to get in touch with Colourtext please email