A little more than a year ago, my colleagues and I started tracking media mentions of drug pricing issues. We assumed it would be an important but relatively laid-back endeavor. Discussions about the cost-savings of hepatitis C drugs was still ongoing, and cancer conferences were putting a spotlight on oncology prices.

Then, on Sept. 18, an intern at USA Today wrote a story about a company no one had heard of raising the price of a drug no one had heard of by an amount no one could have imagined. The headline read: “Company hikes price 5,000% for drug that fights complication of AIDS, cancer,” and it was the first introduction most of America had to Turing.

The name “Martin Shkreli,” Turing’s founder and CEO, never appeared in the USA Today piece, but his obscurity would vanish in a matter of days. Within two weeks, he was being mocked by Miley Cyrus on Saturday Night Live.

The Turing controversy turned our sleepy media-tracking operation into a borderline sweatshop. In the year that followed, we logged 1,300 stories of significance. But not all stories are created equal, so last month, we used our MDigitalLife offering to look at those stories in more detail, trying to answer the question of what pricing/value stories resonated with different audiences: the general public, media and physicians. You can see the results in the infographic below.


That analysis generated seven learnings:

  1. There is no common source for news. The idea that different groups have different information sources isn’t novel, but the top 10 lists show clearly that while the Wall Street Journal is driving the media agenda (five of the top 10 stories, plus a near-miss for a Pulitzer), it’s not having a similar effect on consumers (zero of the top 10 stories). In contrast, two Huffington Post stories had an outsized impact with consumers, but media and physicians barely noticed.
  2. Physicians are aligned with patients. The stories most-shared by physicians were, in general, stories about the impact of various decisions and policies on patients, not broader topics.
  3. The media are fascinated by the system itself. Most of the top stories among reporters had to do not with a single example but with how the pricing system works from generally, underscoring the complexity, and–perhaps–an instinct toward hating the game, not the players.
  4. Bernie Sanders is a Facebook superstar: Sanders unlikely success in the race for the Democratic presidential nomination showcased the social chops of his supporters. While Sanders wasn’t ahead of the game on any pricing issue, he managed to go viral whenever he inserted himself.
  5. Orphan drugs got a pass. EpiPen, Daraprim and Valeant’s business practices got a once-over in the past year, but drugs for small populations (which tend to come with large price tags) garnered little attention, even as they made up an increasingly percentage of new drugs. That’s a phenomenon that also played out, albeit with some exceptions, with oncology drugs.
  6. Shkreli was catnip. Nearly 2 million people have watched this Vice interview with Shkreli on YouTube, making it one of the most-shared links in our database. First, Shkreli was a news driver. Then he was an icon of generally startling behavior, generating more ink. Finally, he was an all-purpose metaphor for think pieces. Regardless of the role he played, Shkreli was the thread that tied the year’s coverage together.
  7. We’re now in the year of EpiPen. As wild and multi-faceted as the Daraprim story was a year ago, link-sharing behavior around EpiPen stories is unprecedented. How that story evolves will determine in large part where the pricing narrative flows in the year to come.

Our database will continue to grow and inform what we do at W2O Group, internally and for our clients.  We look forward to continuing to listen to conversations in the health care space. Here’s to the next 1,300.


We’ve recently had the pleasure of collaborating with on a segmentation project geared toward better understanding men. Joe launched in 2015 and already is one of the biggest premium male-focused publishers in the UK. With an array of mottos like “for men, not lads” and “the voice of British men,” the website focuses on giving men something ‘better than what they’ve been used to over the past few years of online publishing.’ We wanted to use research to demonstrate that men are more complex than we give them credit for, and that current advertising aimed at men is missing the point.

The project entailed two different parts: a deep qualitative survey and a detailed audience segmentation.

While the survey is not the focus of this blog post, its results were startling and confirmed our initial hypothesis:

  • Only 17% of men felt that UK media represented them
  • Over 73% of men did not identify with the word ‘lad’

Perhaps more interestingly, modern men revealed that they felt significant pressure in their everyday life:

  • 40% admitted to have suffered from depression
  • 71% felt too much was expected of them
  • 68% wished they had better ways to deal with stress

The survey results provided a lot of detail about how men felt, but what were they interested in? What is segmentation and how can it be leveraged to reach an unparalleled depth of audience understanding?


Click on the gif for a slower animation.

Segmentation is divided into three phases. First, an audience and a normative are selected. For this project the analysis was made up of approximately 38,000 Twitter accounts of followers, and a UK normative of 41,000 UK Twitter accounts selected at random.

The second step is to collect and categorize every handle the audience and normative follows. We then group these handles by “interest,” using a technique called clustering, which clusters handles that are commonly associated. An interest is typically comprised of 50-100 clustered handles, and—while this is fundamentally a mathematical process—results are later interpreted by human analysts who identify the underlying themes and patterns.

For example, we may see that “Interest 1” is comprised of a strong followership correlation amongst David Beckham, Gary Lineker and Rio Ferdinand. We may wish to call that interest “ex-England Footballers,” but if the next most active handle belongs to Cristiano Ronaldo, we have to revise our title to “famous football players.” This scenario is likely to occur more than once and is why analysts are brought in to avoid inaccurate categorization.

Finally, groups of handles are bundled into segments. This is done through a technique called agglomerative clustering that allows us to see the proximity between interests. Like clustering by interest, this process is a combination of technology and mathematics, layered with human insight. As interests are refined and renamed, segments begin to take shape.


Segmentation of the UK male audience, based on research by W2O Group and JOE Media
Segmentation of the UK male audience, based on research by W2O Group and JOE Media

We identified eight major segments within the Joe set, each with their own idiosyncrasies.

While we don’t surface the majority of the data on this blog post, there are three elements to think about when looking at the data below.

  1. The segments themselves, which represent a unique audience through a combination of interests
  2. The Interest Reach, which represents the percentage of people in the Joe audience who followed a particular interest
  3. The Interest Index, which conveys the rarity of an interest by showing (in the form of a multiplier) how many times more likely the Joe audience was to follow this interest than was the UK normative

Let’s look at a few of the segments from our case study in more depth:

Example of Starstuck segment from W2O Group & JOE Media analysis.
Example of Starstuck segment from W2O Group & JOE Media analysis. Click for a slower/larger animation.

Starstruck is one of our two most mainstream segments. The leading interest here is Hollywood Actors, which is comprised of predominantly male celebrities from Hollywood. Household names like Simon Pegg, Leonardo DiCaprio, Samuel L. Jackson and Tom Hanks are accompanied by more recent additions, like Aaron Paul and Seth MacFarlane. This audience is also very interested in UK Comedians and Shopping, both of which are also popular with the UK normative. An interesting aspect of the latter is the lack of differentiation between online brands (e.g. ASOS and Amazon) and traditional street brands (e.g. M&S and Ted Baker).

Londonites are all about metropolis lifestyle.

Londonites epitomize big metropolis living. They represent both young men living in London (and other large urban areas) and those striving to do so. Londonites are well versed in current affairs and follow a bipartisan list of news outlets and politicians. They have no discernible political stance at this level, with near-equal levels of followership for Jeremy Corbyn, Nigel Farage or even figures like Edward Snowden. They also enjoy the lighter side of current affairs, with Armando Iannucci and Charlie Brooker reoccurring in their News & Politics interest.

Londonites also follow interests related to both London venues and restaurants. Their interest include the Glastonbury music festival, Time-Out recommendations, restaurant reviews from Nigella Lawson, and weekend public transit updates. Londonites are all about making the most of what the city has to offer.

I Bet You is all about competition and gamification

I Bet You is a segment about competition in all its forms. While a large part of the Joe audience is focused on football, this segment is deeply interested a wide range of organized competition. They’re very interested in Rugby, Horse Racing, Darts, Snooker, Cycling and the Olympics. Beyond watching competition, I Bet You enjoys betting on sports outcomes; a rare interest when compared to the UK normative data.

Gamers have multiple interests outside gaming including UFC and television shows such as Walking Dead and Game of Thrones

The Gamers segment is primarily driven by Video Game Companies and unique UK Streamers and YouTube celebrities. While these interests are common across all gamers, the specific streamers found in the interest were unique, featuring the likes of the KSI and MrSyndicate. Additionally, this segment boasts the second rarest interest within the whole research, UFC fighting, which it is 5.4 times more likely to follow than the UK average. The whole segment is rounded off with an affinity for US scifi and fantasy shows, led by interests comprising the casts of The Walking Dead and Game of Thrones.

As you can see, each segment provides a unique glimpse into the mind of an audience. Some represent mainstream appetites, which often align with those in the UK normative data. Other segments embody niche interests, which diverge from the mainstream and represent unique behaviors. Together, the segments provide a more nuanced understanding of Joe’s diverse audience and an unique window into the various affinities vying for their attention.


Whatever you are writing – be it a novel, a tweet, an article –  you can’t land your story without knowing your audience. But as marketers and communicators, how do we best align the right message to the right customer?

According to W2O Group’s Bob Pearson, that’s where the practice of Storytizing comes in. Storytizing, or audience architecture & segmentation, is rooted in using data and analytics to uncover not only what audience segments are saying about themselves, but also what their online behavior is saying about the types of stories that will have the most impact on them.

This was the topic of conversation last week as we hosted our first ever W2O Group GeekATea during Social Media Week London.

Fortnum & Mason’s afternoon tea

But first: what is a GeekATea, you may ask? As Annalise Coady, President of our agency Twist Mktg and head of our London office, noted as we kicked off the event, the name is a British nod to our annual GeekACue event we hold during SXSW in Austin, Texas where we celebrate our appreciation of both barbecue and digital communications. It was only fitting in London that we celebrate the tradition of Afternoon Tea as a backdrop to a lively discussion on this ever-important practice of audience architecture.

At this intimate afternoon tea at the renowned Fortnum and Mason, we were honored to have with us:

  • Will Hayward, CEO of JOE Media, biggest premium male-focused publisher in the UK, Board Member of Social Media Week London.
  • Bob Pearson, President of W2O Group and author of Storytizing
  • Lucas Galan, Head of Analytics Productization at W2O Group

The session kicked off with an overview of trends in storytizing from Bob, and then segued into an introduction to JOE Media’s audience strategy from Will. Following, Lucas provided a deep dive into a case study of audience segmentation research conducted by W2O Group and Joe Media about a very specific and, as we learned, underestimated audience segment – the British male.

Will Hayward discussing JOE Media's audience strategy
Will Hayward discussing JOE Media’s audience strategy

Here’s a brief recap of the major takeaways:

Chapter 1: The new media outlet. New media outlets are built based on audience need, not advertisers’ need for reach. The editorial plan is based on the actual needs of the audience. In this sense, a “media outlet” is defined by customers – we decide what matters and where we spend our time. The good news: a brand site could become a media outlet in its own right… if it is authentic, dependable and it truly understands the audience’s needs.

Bob Pearson of W2O Group, discussing Storytizing
Bob Pearson of W2O Group, discussing Storytizing

Chapter 2Habits win. Media sites will succeed and fail as fast as bars and restaurants. Those who succeed will have built audiences that depend on the outlet as part of their daily ritual.

Chapter 3: Media planning will become AUDIENCE planning. Media planning of the past told us where our audience is, but not what content is going to have an impact on that audience. Micro-audience segmentation will lead to hundreds/thousands of segments to address.

Segmentation of the UK male audience, based on research by W2O Group and JOE Media
Segmentation of the UK male audience, based on research by W2O Group and JOE Media

Chapter 4: Segmenting your audience by analytics, not stereotypes, is the future. Audiences are complex and they do not want to be lumped into broad stereotypes. Our analysis of JOE Media’s audience showed that only 17% of men feel that UK media accurately represent the type of man they are.

Example of Starstuck segment from W2O Group & JOE Media analysis.
Example of Starstuck segment from W2O Group & JOE Media analysis.
Example of Londonites segment from W2O Group & JOE Media analysis.
Example of Londonites segment from W2O Group & JOE Media analysis.

Chapter 5: Surveys vs. analytics. It’s not enough to ask people what they want. People tell you what you want to hear if you ask them, observe them and you will get to know what they really think. The power of these combined allows us to figure out how to most effectively activate specific audience segments.

Chapter 6: Be new school. The value of traditional journalism is huge but the economics is no longer there. Operate under the mentality that everything you are doing is “old school” and always be seeking the “new school.” Will Hayward says at JOE Media, they take the best of old school journalism and bring it into the future (stay tuned for a major announcement from JOE Media that illustrates this concept in the upcoming days!).



Social and digital data predict trends the surveys can miss

If you’re like most companies, you’re using some form of traditional, survey-based market research or brand tracking in order to inform both your marketing and broader business strategies. But what if we told you that surveys alone aren’t all that accurate for some consumer demographics? And that those are the demographics that tend to be most sought-after?

Over the last few years, several members of our analytics team have been focused on predictive analytics. How do we help our clients better predict business results and market trends, so they have the information they need to make sound business decisions?

What we’ve found is that social and digital data provide meaningful signals into market trends that surveys sometimes miss.  When we add digital and social data to more traditional models, our forecasts are generally 25-40% better, and we’re able to predict business outcomes for our customers with at or above 90% accuracy. Even more interesting, digital and social data is often more predictive of behavior than self-reported interest and intent for younger audiences, including Millennials.


Why are social and digital metrics so powerful?

There are a couple reasons this could be the case. First, there could be a spontaneity effect at play. We were dealing with a relatively low-consideration purchase, and younger people tend to have more flexibility with both their time and their finances when compared to older adults, who tend to have more commitments like young children.

The second effect we could be seeing is a social desirability one. We know that younger people are more susceptible to saying they’ll do things they don’t actually plan to do in order to be seen as ‘cool’ or otherwise on-trend. For instance, research has shown that millennials say they prefer chocolate brands with ethical sourcing in focus groups, but actually chose brands based on high-fat content and a relatively small number of pronounceable ingredients*.

To determine which of these was responsible for our results, we decided to take spontaneity out of the equation by looking at a high-consideration purchase: automobiles.

What we found is that for high-priced vehicles, we saw the same general trend as we did for our low-consideration purchase: self-reported intent was a better predictor than digital and social metrics for older audiences, whereas digital and social data better predicted purchases for younger groups. However for low-priced vehicles, we saw the opposite trend. Younger audiences’ tendency to report they were going to purchase a low-priced vehicles was a better predictor of their behavior, but for older audiences, digital and social metrics were a much better indicator.

What does this mean? The >35 crowd is find with saying they’re going to buy a BMW, but if they’re planning to buy a Kia, they’re less likely to tell you about it. This suggests not only that social desirability affects the predictive value of a survey, but also that any demographic group can fall victim to the bias.

What kinds of social and digital data matter?

While most marketers look at conversational metrics to calculate share of voice or get at voice of the consumer, it’s actually behavioral metrics that add the most predictive value. It’s the relatively anonymous nature of things like YouTube video views, website visits, and search that provides us a glimpse into what people are really interested in, especially when that thing is not particularly “cool.”

Social desirability still plays a role when you’re posting on Facebook to a group of 1,700 of your closest friends, but not when you’re one of the nearly 15 million people watching “A Goldmine of Blackhead & Whitehead Extractions” on YouTube at 1 am.

Insights in an echo chamber

Companies who rely solely on survey data run the risk of making critical decisions in an echo chamber. Consider a brand developing a new product.  They put a survey in the field to understand which consumer segments may be interested in a premium product like theirs, and find that it’s more established consumers—those over 45—who show the highest definite interest and intent to purchase.  They rely on this insight as they develop their creative and select their marketing placement, heavily investing in primetime television and less on digital channels.

The problem with this scenario is that the survey didn’t capture the whole picture. There is a group of millennials—affluent, urban, 25-34—who are interested in the product, but don’t say so because it’s not particularly environmentally-friendly. The result is a self-fulfilling prophecy: the younger audience didn’t say they were interested, so the brand didn’t effectively market to them, and, as a result, that high-value audience didn’t buy the product.

The brand looks at their purchase data, relieved that their insight was ‘proven,’ but they actually missed out on a large pool of potential revenue as well as an audience that could carry their business into the future.


Market research moving forward

We no longer live in a world where survey research is the best predictor of consumer behavior. Given our ability to overcome the limitations of traditional market research by bringing in new metrics, it’s irresponsible for brands not to leverage social and digital data in brand tracking and forecasting.

We’re so excited that our work was presented at Esomar’s 2016 Congress, because it shows that the broader Market Research community is coming around to the idea of fusing traditional metrics with digital and social signals, as well.


Seth Duncan and Kelley Sternhagen

*Young & McCoy (2015).


Kelley Sternhagen has delivered consumer insights for clients across verticals, including healthcare, tech, retail and consumer packaged goods. A range of skills from predictive analytics to consumer segmentation and planning allow her to deliver critical insights to brands to optimize their content, targeting and media mix.

Kelley began her research career as an undergraduate at Princeton University. Prior to joining W2O Group, she worked as a product marketer at Google and was an on-site analyst for Nike.

She holds a Bachelor of Arts in psychology and gender studies from Princeton University

Connect with her on LinkedIn or say hi to her on Twitter at @ksternhagen.


All social media platforms serve as a tool for storytelling.

Whether users are telling the stories of various humans of New York, sharing the story of their weekend through a series of 10-second photos, or capturing the perfect meal through the proper filter, all social networks empower users to publish their lives or the lives of others.

Twitter is no exception to this rule. However, the platform has always been unique in comparison to its competitors:  Facebook, Instagram, and Snapchat. Perhaps Twitter’s most defining differentiator is its less visual platform, specifically compared to Snapchat and Instagram. Even so, Twitter is still competing (and generally winning) in the social media space. The question is… how?

What Makes Twitter Work?

Twitter has an ace in the hole:  it is the undisputed platform of choice for live events and iconic pop culture moments. From the 2016 presidential election to social justice movements to sporting events, people turn to Twitter for real time updates and to express their reactions and observe others’.

What is going in world news? Follow @CNN.

What is the score of the latest game? Follow @espn.

Who is the next big musician? Follow @allsongs.

How was your friend’s last day of work? Follow her account. (Don’t, though, because she is really MY friend.)

From the personal news of your friendship to global news, Twitter perfectly curates and delivers your content. Twitter has transformed into the modern day newspaper and that is why users love its platform.

Moon Landing





Kim K_TSwift Reaction

Additionally, the platform empowers users to create their own iconic moments for the world to celebrate, criticize, and join (e.x. #GIFOscars#FamousMelaniaTrumpQuotes#CareFreeBlackKids2k16).

Users are notified when people within their network are tweeting, retweeting or liking the same topic or trend– so even if a user is not privy to the conversation, they can join in without missing a beat.

Melania Trump Quotes

How Will Twitter Grow in the Future?

Twitter’s latest move toward growing their cash flow and number of users is broadcasting sports within the platform. Twitter recently signed its first broadcasting deal with the NFL, which will allow the company to stream 10 Thursday night football games this upcoming season.

And football is not the only sport Twitter is interested in:  last month the platform hosted live coverage of Wimbledon. Plus the social media company has signed deals with the MLB, NHL, NBA and Pac-12 Networks.

Livestreaming sports may address the two main barriers of the platform. First, creating a steady flow of revenue. Potential advertising for brands most likely will pique the interest of marketers, providing a constant, dependable flow of cash. Second, broadcasting sports may help grow the amount of people who use the platform. The games will be available for people to stream regardless of if they have a Twitter account. This provides access to those who may have never interacted with the platform the opportunity to become familiar with the social media network. If people would like to join the conversation where they are tuning in, Twitter is the most natural fit, which may lead to more people signing up.

Case Study: Olympics 2016

Over the past couple weeks, our analytics team took a look at Twitter presence of the five members of this year’s USA women’s gymnastics team. Below you will see the number of mentions each member accrued over the entirety of the artistic gymnastics competition (August 06, 2016 – August 11, 2016; August 14, 2016-August 16, 2016).



If this volume of mentions can be amassed with merely five athletes in the span of two weeks, Twitter may be on the verge of becoming an untouchable competitor in the social media space.

Even outside of the five gymnastics’ superstars, Twitter has been bustling during the Olympics.

  • Tweeters have empathized with parents of competitors.
  • They have created overnight memes such as #PhelpsFace.
  • They have shamelessly praised their favorite athletes.

Katie Ledecky

It is clear the platform succeeds during live events, regardless of whether the event is streaming within the platform or not. However, one can only imagine how much success the network could have if the events that people were speaking about actually live-streamed in their feeds.

The combination of creating a platform where users can share their thoughts in real time while never having to leave the app could be the reason that Twitter does not just survive, but thrives. Plus, it eliminates the pain points of cross screen marketing for marketers, which may also be a reason they pay big bucks in ad dollars.

If streaming sporting events serves as a lucrative vehicle to gain revenue and users, Twitter may continue to strike up additional partnerships with other live events outside of the sports market. For example, Twitter broke social media ground by live-streaming the 2016 Democratic and Republican National Conventions.

Twitter is thinking outside the box and launching into new spaces, and we are excited to continue to grow our clients through this platform.

Jon Carillo Headshot

Creative for this project was provided by Jon Carrillo, a graphic designer at W2O Group. Connect with him on LinkedIn and if you don’t mind the occasional rant about the San Antonio Spurs, follow him on Twitter at @JonnyCTweets!


Brad Snyder Headshot

Research assistance for this project was provided by MDigitalLife intern, Brad Snyder. Brad attends Tiffin University where he is a business major with a focus in marketing and a minor in international business. Connect with him on Linkedin and if you can tolerate Cleveland sports follow him on Twitter at @b_snyds!


The recent Zika infection of 14 people in a Florida community near Miami – the first case of U.S. transmission from local mosquitoes – has further raised fears that the U.S. will face a large domestic Zika outbreak. With increased news coverage on the domestically transmitted cases as well as the upcoming Olympic opening ceremonies on Friday, August 5th in Brazil, a country hit hard by Zika, there is likely to be a large amount of public discussion about the looming domestic Zika threat.

Studies have shown that physicians are one of the most trusted sources of online health information, so it is likely that patients will be turning to their physician to understand how concerned they should be about Zika and the steps they should take to avoid contracting the virus. But what are physicians saying about the Zika virus?

At W2O Group, we set out to answer that question by querying the MDigitalLife Health Ecosystem Database, the world’s first database to link physicians’ online content to their national physician identifier records, to gain insights into the networks, relationships, and social activities of online physicians.

A Brief Review of Physician Zika Conversation Timeline

Specifically addressing Zika social conversation, there have been over 51,000 posts from more than 4,700 unique physicians since January 11, 2014. U.S. physicians contributed over 29,000 posts from 2,700+ authors and Non-U.S. physicians contributed more than 22,000 posts from 2,000+ authors.

Zika Timeline With Callouts Dark

Physician Zika conversation grew rapidly on January 11th of this year when the first Zika case in the U.S. was reported in Houston, TX. A doctor led the charge with this news as Dr. Umair Shah, Executive Director of Harris County Public Health & Environmental Services, shared the news via his personal twitter account as the news was released to the public. The Houston Chronicle quickly picked up the story (and included quotes from Dr. Hotez), followed by Sarah Begley’s piece in TIME and then similar pieces in nearly every other outlet. Conversation was driven higher on January 16th after Hawaiian officials announced the first baby born in the US with microcephaly linked to Zika and the CDC officially advised pregnant women to avoid traveling to areas where
Zika is spreading. Physician conversation continued to rise on
January 28, when Margaret Chan, WHO Director-General, announced that “the virus was detected in the Americas, where it is now spreading explosively”. Conversation reached its current peak between February 1-3, when physicians posted over 3,300 times after the WHO declared the Zika virus and related links to possible birth defects an international public health emergency.

Watch Monthly Physician Conversation Evolve From 2014-2016 In The Interactive Map Below

ZikaGifAfter global physician conversation peaked with over 15,000 posts in February, conversation began to decline; US physicians maintained an average of 3,400 posts per month while ExUS Physicians contributed an average of 1,700 per month between March and July. U.S. Physicians did see a small uptick in conversation in April when the CDC confirmed the link between microcephaly and Zika. Additionally, there was a rise in conversation in July as the first case of US transmission was announced in Miami. We expect to see a significant spike in global conversation as we approach hurricane season in the U.S. and the Games kick off.

Conversation Over Time

Zika NGram Graph2

We tracked five different keywords throughout the course of the conversation to examine the evolution of the language used by physicians discussing Zika. During late 2015 and early 2016, a significant portion of the tweets contained some mention of “Brazil,” the origin of the outbreak. The volume of tweets mentioning the word “women” was proportional to the overall volume of the U.S. physician Zika conversation, underscoring the relevance of this conversation to women’s health. In April of 2016, tweets started emerging that contained the word “funding,” with many questioning the level of government funding to combat the growing outbreak. July saw the introduction of both “transmission” and “Florida” at a high rate, correlating with many reports examining the transmission of Zika and the rise of domestic cases in Florida.

Going Back to the Start

So who actually started the online physician conversation about Zika? The first post we recorded from a U.S. physician was by Dr. Peter J. Hotez, Founding Dean of the National School of Tropical Medicine, Texas Children’s Hospital Chair in Tropical Pediatrics & President of the Sabin Vaccine Institute, on August 3, 2013. Dr. Hotez’ initial post shared a study released in the Journal, PLOS Neglected Tropical Diseases, which discussed the emergence and spread of Zika in Africa and Asia. The initial post picked up a small amount of traction with four retweets. This did not cause a blip on Twitter’s radar. Though perhaps more people should have been paying attention, since Dr. Hotez proved to be rather prophetic with his post in March 2014 on Zika.

Conclusion and Acknowledgements

Our analysis demonstrates how the online physician conversation about Zika has grown over time, and evolved both geographically and linguistically. It is our hope that monitoring this conversation can be helpful to both physicians and the public at large in keeping abreast of the current issues being discussed in the ongoing Zika outbreak.

This post was co-authored by Dr. Yash Gad, Chief Data Scientist for MDigitalLife.

As a part of a larger research initiative and partnership, we thank our colleagues from the lab of Dr. Wenhong Chen at the University of Texas Austin who provided insight and expertise that contributed to this article.

An earlier version of this post incorrectly stated Dr. Hotez first tweet on Zika was in January 2014.

Earlier this month I began to explore the correlation between Twitter analytics and Britain’s possible departure from the EU. Things are heating up in the Twitterendum, but not necessarily getting any clearer. The volume of tweets in the five weeks following our last post was effectively the same as the previous twenty combined. This represents an huge increase in the raw number of tweets, but also in unique contributors, meaning a lot more people are being drawn into the debate.


In spite of the increase in activity, the Twitterendum results remained remarkably static. The number of ‘stay’ unique accounts remained around 10% higher than their opposition whilst Brexiters continued to generate more per capita tweets than Bremainers, around 20% more content.


In the ‘leave’ camp, Burnley, Eastbourne and South Northamptonshire all doubled down on their positions. The number of ‘stay’ supporters increased dramatically in Woking and Manchester, however, though they are now more contested, both still generated a far greater number of ‘leave’ tweets.

Only one new joiner stood out – Oldham. Instead of making gains in key decisive swings of opinion, the Brexit camp found success in edging several smaller Local Authorities. Doncaster, Sunderland and Kingston upon Hull all moved from ‘undecided’ to ‘leave’.

Similarly, the ‘stay’ camp continued to reinforce its position in university towns and Labour strongholds. All ‘stay’ frontrunners from our last update increased their #StrongerIn scores.

London continued to be the centre of extremely heated debate with the total number of tweets only just favoring the stay camp. However, in terms of unique users it enjoyed one of the highest ratios of Bremainers: Brexiters in the entire country…perhaps unsurprisingly. The animation we’ve included shows London’s gradual movement from undecided firmly into the ‘stay’ camp, mobilizing more and more ‘stay’ supporters with each passing week.

The remaining most populous centers experienced very similar phenomena. Sheffield, Birmingham and Leeds all recorded surges in the number of unique ‘stay’ supporters, while simultaneously recording disproportionate levels of ‘leave’ tweets.


From this, it seems like the race is rather contested… though the higher proportion of unique accounts means that the ‘remain’ camp is pulling slightly ahead. What will the results show on the eve of the vote? Will there be any correlation between Twitter trends and the final referendum results? Join us next week!

Lucas-Galan-headshotLucas Galan currently serves as the Head of Analytics Productization at W2O Group’s London office. Connect with him on LinkedIn!


Similar to the rest of the country, we at W2O Group have been glued to our TVs, phones, tablets and laptops watching the NBA postseason.

Who could blame us? The theater and caliber of play throughout the past several weeks has been some of the best basketball that the league has seen in years. The series of games that preceded the finals captivated homes, cities and fans across the country, whether you were watching the underdog Toronto Raptors battle it out against King James and the Cleveland Cavaliers – or if you were engrossed watching the Golden State Warriors and the Oklahoma City Thunder each give it 100% in a seven-game series. The NBA postseason has been nothing short of entertaining.

The Analytics Formula

As irresistible as we find the competition, we find the analytics surrounding the sport just as alluring.

Similar to our clients, we had preconceived notions of which fans would be the loudest  and in some respects were shocked by the story the analytics told us. The exclusivity our analytics provided us for this project exemplifies how we position our clients to dominate against the competition… just like the pros.

The focal point of our curiosity stemmed from which teams had the loudest fans during the regular season. Thanks to our amazing analytics team, we were able to take a deep dive into the data and discover which fans were making the most noise.

For this analysis, we defined the “loudest” fans based on how actively they engage with their teams on Twitter. Engagement was calculated as total retweets plus favorites of content posted by each NBA team. Additionally we accounted for the following metrics:

  • City population (market size)
  • TV households in market area
  • Number of tweets posted from each team’s offical Twitter handle during the regular season

The Results: The Loudest

This approach identified the following teams as having the loudest fans on Twitter: the Warriors, the Spurs and the Thunder.

(We cannot say we are surprised that the Warriors were included in the top three. There is no doubt that our CEO Jim Weiss and our San Francisco headquarters’ team members have contributed heavily to the conversation.)

Based off of our analytical approach, the Spurs had the highest engagement rate per potential fan, making them the loudest Twitter fan base in the NBA with a “noise” index of 100. The Thunder rank second with a noise index of 70 with approximately 1.44% of total fan engagements in the regular season vs. Spurs. Finally, the Warriors rank third with a noise index of 55.2 with approximately 1.28% of total fan engagement in the regular season.

W2O_LoudestNBATeams_1024x512 (1)

Below are the top 10 loudest fan bases on Twitter in the NBA.


It is key to take into account outside factors such as number of cumulative tweets posted and city population. Although raw numbers are helpful, often times, they do not tell the complete story. If we only based engagement levels off of the raw data, retweets and favorites, the Warriors would have the loudest fans, followed by the Spurs, and the Thunder would round out the list.


However, our team understands the need to level the playing field when discussing pro teams. Some market sizes are larger than others (i.e. San Francisco) and some teams are not as active as others on Twitter. Note the Chicago Bulls sent 2,767 tweets during the regular season vs. the Warriors who sent 8,031. All of this context affects the bigger picture. Numbers always tell a story; we are lucky to have an amazing group of authors in our analytics team.

The Results:  The Most Appreciative Fans

We also wanted to know which teams’ fans engaged the most with tweets from the team’s official Twitter account following a win. Based on engagement rate per win, the following teams have the “most appreciative fans” in the NBA:

1. San Antonio Spurs

Avg. Engagement Rate Per Win: 84

2. Oklahoma City Thunder

Avg. Engagement Rate Per Win: 71.2%

3. Philadelphia 76ers

Avg. Engagement Rate Per Win: 69%

4. Sacramento Kings

Avg. Engagement Rate Per Win: 66.7%

5. Milwaukee Bucks

Avg. Engagement Rate Per Win: 49.1%

It is easy surmise why the Spurs, Thunder, Kings and Bucks had appreciative fans throughout the regular season. The Spurs and Thunder were both top contenders throughout the playoffs. The Thunder going seven games against the reigning league champs, the Warriors. The Kings had a great season with new talent on its roster, and the Bucks handed the Warriors its first loss of the season. However, the Philadelphia 76ers had an embarrassing season, winning only 10 games out of 82. They nearly set a record for least amount of games won in a season, only narrowly escaping that fate by winning more games than the 1973 76ers who went 9-73.

Whether it was relief, sarcasm or praise, Philly’s fans made some serious noise on Twitter on the rare occasion the 76ers were able to deliver a win.

76ers Fan

The Results:  The Least Appreciative Fans

Reversely, the following teams’ fans had the lowest engagement rate per tweet following a win, deeming them the “least appreciative fans” of the NBA:

1. Brooklyn Nets

Avg. Engagement Rate Per Win: 3.9%

2. New York Knicks

Avg. Engagement Rate Per Win: 6.3%

3. Orlando Magic

Avg. Engagement Rate Per Win: 7.6%

4. Los Angeles Clippers

Avg. Engagement Rate Per Win: 9.1%

5. Boston Celtics

Avg. Engagement Rate Per Win: 12.6%

Perhaps it is easy to recognize why the Nets’ fans are least appreciative: the franchise only won a pitiful 21 games out of 82. However, the Clippers had a fantastic season, winning well over 50 percent of its games, yet its fans did not seem too impressed on Twitter. Even when the Clippers play excellent basketball, its fans do not have much to say about it via Twitter.

Looking Forward

As the NBA prepares to wrap up its 70th season, we enjoyed looking back on the highs and lows of the season through the lens of analytics.

W2O always appreciates the incredible power our analytics can provide and the stories they are able to tell us, whether it is for our clients or our own passion projects. Our suspicions might be confirmed or we might be surprised, but we are always informed and empowered to create a strategy that places our clients in a position to succeed. Have any questions about our analytics? Contact Us

Jon Carillo HeadshotCreative for this project was provided by Jon Carrillo, a graphic designer at W2O Group. Connect with him on LinkedIn and if you don’t mind the occasional rant about the San Antonio Spurs follow him on Twitter at @JonnyCTweets!

If tweets were referendum votes then, come June 23rd, the UK would be departing the EU. Looking at the period beginning early 2016, spanning the official launch of both the ‘Brexit’ and ‘Bremain’ campaigns, and culminating in the London mayoral election, it is clear that the latter, #StrongerIn Bremain campaign enjoys a small but significant lead in the number of unique accounts pledging support. However, it is the Brexiters that appear more ardent in their beliefs and more vocal in their call to arms, tweeting roughly twice as much as their Bremain counterparts.

To try to gain a greater understanding of the distribution of Brexiters and Bremainers throughout the UK, the socioeconomic context in which they exist and the correlation (if any) between Twitter activity and actual referendum outcomes, we have created a model that combines the absolute number of tweets, the number of unique accounts and the total populations for each local authority district (LAD).

Screen Shot 2016-05-27 at 16.48.28

LADs known to have concerns regarding high levels of immigration and demonstrating above average support for UKIP, the country’s Eurosceptic and right-wing populist party, were (somewhat predictably) Brexit strongholds. Burnley, Eastbourne, Hartlepool and Bournemouth all ranked highly in Brexit ratios and all fit this profile. However, UKIP voting alone was not a definitive predictor of Brexit performance.

Clacton, Boston, Thurrock and Rotherham all recorded high levels of UKIP support in the 2015 election but do not appear at the forefront of the Brexit campaign as fewer residents utilise Twitter and those that do tweet with low frequency.

Brexiters also tended to lean Conservative and showed strength in traditionally Conservative strongholds such as Woking (one of the safest Tory seats in the country), Northamptonshire South and West Dorset. The results for Manchester, however, ran counter to this idea of traditionally more conservative populations favouring a departure from the EU. An ultra-safe Labour seat and beacon of multiculturalism, its inhabitants had relatively more unique Twitter accounts and total numbers of tweets in favour of the Brexit campaign – though, admittedly, the number of the Bremain accounts was also high.

The #StrongerIn camp was, in general, more predictable, led by university towns with a strong base of Labour support (Oxford, Cambridge, Cardiff and Exeter were frontrunners). Scottish urban centres also leaned in this direction. Edinburgh and Glasgow, Scotland’s most populous cities, showed a predominance of support for the remain campaign by unique Twitter account numbers but almost even number of tweets for both camps.


These results come from a preliminary analysis of the data and it may be that as the volume of referendum-related Twitter content increases in the approach to June 23rd, so the results shift. Of the top seven most populated places in the country, only one – Manchester – sits definitively in either camp. The other six remain undecided but with strong showings from each camp. It is in these large urban centers where the Twitterendum will be most intense, and we expect this to be highly contested in the coming weeks.

Gif 6



Of course, tweets are not votes. Twitter users do not reflect the UK population as a whole. Twitter users account for roughly a quarter of the population (23%) and tend to skew young and urban. Perhaps most pertinently, Twitter may be less representative of the British because of something inherent to Britishness. Twitter is very much a public soap box and, for a nation so often aligned with a strong sense of propriety, it is not everyone’s cup of tea.

So, while the Twitterendum results should not be used as an analogue to real voting attitudes, these observations can still provide an interesting barometer through which to measure the winds of sentiment sweeping the nation. As the debate heats up in the coming weeks and traditional media becomes saturated, we’ll continue to turn to this model to see what further insights we can glean, so please join us for regular updates of our Twittterendum coverage!

Lucas Galan headshot

Lucas Galan currently serves as the Head of Analytics Productization at W2O Group’s London office. Connect with him on LinkedIn!

Learn more about W2O Group:  About  Work  Contact