The Data Evangelist

Using data whenever possible because humans suck

[emilychambliss]

Topics Trending in Tandem: A Social Timeline of Ferguson Topics

Came across this post I wrote on the Carrot blog based on my social analysis after Ferguson.  In some ways, it’s outdated, but also still relevant.  Originally posted here: https://carrot.is/blogging/industry/ferguson_analytics -----

Social media is now the fastest way to share information on a local and global level. From a quick access source for the latest news, to a public forum for large scale discussions around human rights and the role of government, social media provides a space for both. It is the medium that transmits data in multiple forms—text, photo, video—with content ranging from cute cat pics to more serious, philosophical subjects. Topics, when tracked in social media, show distinctly different trends depending on the nature of the discussion taking place.

The chart below depicts topics mentioned across social media to show trends in the online discussion surrounding the death of Michael Brown in Ferguson, Missouri. Using Netbase, a social listening tool, we collected mentions of Ferguson and relevant topics, isolating the mentions for each topic to see how they trended.

Timeline

Saturday, 8/9/14 Ferguson police shoot and kill Michael Brown.

Sunday, 8/10/14 Ferguson top topics in social are: police, Michael Brown and murder (132k total mentions).

Monday, 8/11/14 Looting mentions trend (46k). Reports of rubber bullets used to disperse crowd, but not yet trending in social (22 mentions).

Tuesday, 8/12/14 Tear gas starts trending (60k), as well as rubber bullets (16k).

Wednesday, 8/13/14 Police arrest two reporters covering the protests. Ferguson mentions dip 23% from the day prior.

Thursday, 8/14/14 The day President Obama comments on Ferguson, police, murder, tear gas and protests, topics trend together, while civil rights (144k), militarization (102k) and rubber bullets (57k) mentions peak. Michael Brown mentions also peaked on 8/14 (483k) and then again on 8/18 (333k) trending in tandem with mentions of police, protests and civil rights, suggesting the use of Michael Brown's death to facilitate social change.

Friday, 8/15/14 Police release name of the officer and surveillance tape. Darren Wilson mentions peak (62k). Highway Patrol mentions trend (50k) while protest mentions decline (by 56% from the prior day) with most mentions acknowledging Highway Patrol’s success at keeping the peace.

Saturday, 8/16/14 Looting mentions peak again (63k, +505% from the day prior).

Sunday, 8/17/14 Mentions of civil rights start trending again (+55% from the day prior).

Monday, 8/18/14 MO Governor brings in the National Guard (106k mentions).

Tuesday, 8/19/14 Protests becomes the third most mentioned topic for the first time (194k mentions).

Wednesday, 8/20/14 Ferguson mentions decline 42% from the day prior.

Thursday, 8/21/14 Ferguson mentions continue to decline another 36%.

Friday, 8/22/14 Although Michael Brown mentions trend up slightly (+14%), overall Ferguson mentions continue to decline (-34%). On 8/22, Huffington Post releases an article entitled "21 Numbers That Will Help You Understand WHy Ferguson Is About More Than Michael Brown," that got roughly 11k FB likes, 25k FB shares and nearly 600 tweets to date.

Saturday, 8/23/14 Ferguson mentions continue declining (by 35% from the day prior).

Sunday, 8/24/14 Highway Patrol and Rick Perry mentions start trending (increasing 26% and 67% respectively from the day prior), although overall Ferguson mentions continue to decline (-23% from the day prior).

Another Look

This chart shows the same topics as a percentage of Ferguson mentions to show how much each topic was included in the daily conversation. Looking at this chart, we can see that initially, the story of what happened in Ferguson was relatively simple. On Saturday 8/9 and Sunday 8/10, the conversation focused primarily on the police murdering Michael Brown.

On Monday 8/11 and Saturday 8/16, we start to see the emergence of looting in the overall conversation, as well as other topics that flow out of the initial event. Tear gas becomes one of the top topics mentioned from Tuesday 8/12, and in the following week. We see civil rights mentions gaining traction in the conversation building from 1% on 8/10 to 4% on 8/14, and back down to 1-3% in the days to follow. Similarly, militarization discussion emerges after 8/11 and remains steady in the conversation, included in 1-3% Ferguson mentions.

So What?

There are topics that trend in social that are a direct response to the occurrence of an event, such as looting (8/11, 8/16), the involvement of the Highway Patrol (8/15) and National Guard (8/18), and the release of Officer Darren Wilson’s name (8/15). These events caused an appearance of a new topic in social media as people responded to new news. In the chart above, we see these topics emerge suddenly as a mass of color that did not exist prior; they tend to fizzle out as the news becomes older.

However, there are also topics that emerge as part of the deeper conversation around the news events. These topics include civil rights and militarization, both emerging around 8/11 and 8/12 and steadily making up 1% to 4% of mentions in the days to follow. These topics behave differently in social media in that they are a consistent, ongoing theme of the discussions. They actually look different in the chart above; they are the continuous threads of color, unlike the news event reactions which bubble up and quickly dissipate.

These thread-like topics that remain consistent throughout the evolving discussion reflect more than just the public’s reaction to the news event. These topics involve a deeper look at the state of our government. They are political and impassioned. They go beyond the initial news event and venture into larger conversations about right, wrong and the grey area between them. They have a deeper significance, and they live among the quick and impulsive reactions in the social web.

While some brush off social chatter as flippant and shallow (i.e. slacktivism criticisms), others are having deep, meaningful discussions that have great potential to influence others and drive change.

References: “A new Pew Research Center analysis of media coverage of the event and subsequent protests finds that the story emerged on Twitter before cable, but the trajectory of attention quickly rose in tandem, peaking on both mediums the day after two journalists were arrested and protests turned more violent.”

Top Mentions By Klout

8/10 @FoxNewsRadio (Klout: 87, 113K followers) “Tensions high in Ferguson, MO, a suburb of St. Louis over the fatal shooting by police of an unarmed black teenager.”

8/12 @AP (Klout: 99, 3.76M followers) “Father of teen fatally shot by police asks crowd in Ferguson, "Come together and do this right... No violence": http://t.co/kMNkx7XXIL.”

8/15 @WashingtonPost (Klout: 99, 3.16M followers) “This is what Ferguson police say occurred in the minutes leading up to Michael Brown's death http://t.co/TV7zCf5TtT.”

8/17 @HuffingtonPost (Klout: 99, 4.5M followers) “RT @BuzzFeedNews: Capt. Ron Johnson: "I am you. I will protect you, I will protect your right to protest." #Ferguson http://t.co/3UNk7arw5h.”

8/18 @TIME (Klout: 99, 6.11M followers) “Obama voices concerns on police militarization in wake of Ferguson http://t.co/pa1oFpd2yZ.”

8/19 @TIME (Klout: 99, 6.11M followers) “Paul Ryan: Don't rush to judgment on Ferguson http://t.co/ETT2Sj3S8f.”

8/22 @HuffingtonPost (Klout: 99, 4.5M followers) “21 numbers that help you understand why #Ferguson is about more than Michael Brown http://t.co/g5lhhpDUrU.”

[emilychambliss]

Charting Discipline & Being Kind with Colors

I’ve resumed exercising and documenting my work outs.  The manual process of inputting data provides a satisfaction akin to a pat on the back or crossing an item off a to-do list (Type A people will understand this). Looking at the data for the first time since collection began (on 3/1/15), my discipline--as determined by # of work out days--has tapered over time with April being the most disciplined month.  Beginning an exercise regimen takes a few weeks to develop into a routine and to see results.  After the first month of working out at the gym, I began to notice my body changing, which provided incentive to work harder.  After reaching my “happy weight” at the end of April, my perceived need to exercise at the gym declined, as did my work out frequency.  

Color coding in charts provides a way to digest information quickly.  Avoiding the color red to indicate non-work out days because it is “too mean,” I opted for purple, a color that recalls tubby dinosaurs and Prilosec

[datavizminusdata]

h/t @x0xMaximus

[obeserhino]

//sketch 12.8.14

1000 digits of e A Circle represents the repetition of the same number in a row A Circle above that indicates yet another repetition A Circle above that is yet another (the same number 4 times in a row) 

[emilychambliss]

Inspired by Martin Macko. http://people.ksp.sk/~martin/fractal/

Original image: http://spikedmath.com/570.html

[sunfoundation]

LEGO New York, 3D Design Based on Maps & Satellite Imagery

New York 3D artist and motion designer J.R. Schmidt created a beautiful 3D LEGO New York topographical design by using various elevation maps and satellite images of New York (as seen below) to properly set the elevation and colors of his rendered LEGO blocks.

[carrotcreative]

Meet Emily, an analyst at Carrot.

Would you rather lose your arms or legs? Why? Legs - because I’d install monkey bars and handles all over my apartment and get super buff and be awesome.

Dogs or cats? Why? Cats because they are more like humans. They have complex personalities that...

[emilychambliss]

Resuming Quantified Self Practices

A few years ago I tracked my eating and exercise habits in order to gain insight into the trends and limitations around my energy consumption and expenditure.  I used Excel over one of the many apps available (primarily because I didn't have a smart phone).  By color coding my entries, I developed a system of discipline and organization.  I was able to glance at my exercise worksheet and see what muscle groups I hadn't exercised recently, how often I had skipped work outs, and whether I was meeting goals.  In my consumption worksheet, the color coded system would tell me how closely I adhered to my dietary restrictions.  

Not only did I learn about how I behave on different days of the week and times of the day, tracking these behaviors did help regulate them.  At the time I was far fitter than I am today.  

I presented the data about a year and a half ago at a Quantified Self Meetup.  My slides are now available on SlideShare for whomever is curious.  I also recently found some notes by Measured Me on the event.  

So why am I concerned with this now?

After binging on meats at a BBQ this Easter, I decided it was time to do something about my consumption, which has become excessive, particularly on the weekends.  And so, I have resumed my consumption tracking as of yesterday, and again, I am doing it in Excel as opposed to an app.  I choose Excel because I feel the process of manually entering items and checking nutritional info online to help internalize the reality of  consumption behavior.  Of course, I can also do some quick and easy manipulations to analyze the data (and I stare at my phone too much as it is).

So the QS practice resumes.  I plan for tracking exercise with a Speed and Cadence Sensor by Wahoo that attaches to my bicycle to give me data around my cycling since I cycle everyday.  

[emilychambliss]

Neat!  But does this make sense for visualizing social networks and information sharing?  How can this visual representation be improved to be more comprehensive or useful?   An interactive visualization of three online communities: StackOverflow, GitHub, FriendFeed

"Ekisto is an interactive visualization of three online communities: StackOverflow, Github and Friendfeed. Ekisto tries to imagine and map our online habitats using graph algorithms and the city as a metaphor.

A graph layout algorithm arranges users in 2D space based on their similarity. Cosine similarity is computed based on the users' network (Friendfeed), collaborate, watch, fork and follow relationships (Github), or based on the tags of posts contributed by users (StackOverflow). The height of each user represents the normalized value of the user's Pagerank (Github, Friendfeed) or their reputation points (StackOverflow)."

[emilychambliss]

Big Data restrictions pertaining to marketing efforts.

This article brings up a good point to keep in mind about the limitations of big data.  "Data only tells you about the population and actions of individuals it represents, which inevitably excludes part of the population."  Though the examples of two individuals represent extreme opposites (the first being highly valuable to marketers), the point is about being aware of restricted data flows, which is pertinent to what I do everyday.   Let's say a client wants to expand their social reach to a new market demographic.  We have data on their current audience, but lack data on the target audience--the untracked populations.  We may consider ecosystem or competitive tracking around a topic or similar brands, or consult outside research.  

We have even less of an understanding of the populations whose behaviors occur offline.  For clients who aim to reach audiences that are less active online, the restriction of data flows from this population presents a major difficulty.  Mindfulness and careful planning are worthwhile investments to accomplish the brand goal and related business objectives.  

Not only do limitations include the lack of data on all populations, they also include lack of data for certain metrics.  For instance, age data collected through social listening tools often come from two sources: Facebook and YouTube.  We can make inferences about age segments of the population on these networks, but reaching them on other networks where age data is unattainable requires extrapolation.  

Just some things I keep in mind...

[emilychambliss]

Cats vs. Dogs: Who Really Gets More Social Media Love?

Early in the year, I conducted silly research on cats vs. dogs to determine which was actually more popular on social media.  Here it is for your procrastination pleasures. 

---

The question of which is the better—cats or dogs—is a popular, trivial debate.   Some have noted that cat content is widely prevalent in social media and speculated on the reasons behind why social media users focus more on cats than dogs.  But do people really favor cats over dogs online? 

I ran a quick query search using Sysomos, a social media measurement tool, to gauge the relative popularity of cats versus dogs, comparing results for the following terms for the time period 5/25/12 to 1/23/13.  The results were counterintuitive.

Search Terms for Cats                   Search Terms for Dogs

Cat                                                  Dog

Cats                                                Dogs

Kitty                                                 Puppy

Kitties                                              Puppies

  In terms of overall mentions of the above search terms, cats had 78.0 million mentions compared to 89.5 million mentions for dogs, positioning dogs as 14.7% more popular in terms of overall social media mentions.  But on which networks are dogs more popular than cats?

  Both animals are most mentioned on Twitter, followed by forums, blogs, and news sites.  A notable difference in dog to cat mentions by networks showed dogs as having 86% more mentions on news sites compared to cats.  I browsed through popular news articles and concluded that dogs are more frequently discussed on news channels because they have more utilitarian function for humans and are more integrated into the public sphere.  News articles included stories about rescue dogs, hunting dogs, dogs in public spaces and the subsequent problem with dog defecation, and so on.  

 Comparing age data collected from blogs showed the majority of people talking about either animal were between 21 and 35 years old, followed by people from 36 to 50 years old.  The most striking difference was with the 51+ age group—dogs had 23.6% of people over 51 years old compared to cats with only 4.3%.  What could account for this difference?  Have the cat people died off, or are older dog people more active on social media?  I can only speculate.

         And since I’m looking at the classic debate of dogs versus cats, I couldn’t resist the male/female comparison for people talking about each animal. 

        Although females had the majority for each animal, males mentioning dogs were nine percentage points greater than males mentioning cats.  This evidence could be used to support an array of arguments about the nature of gender differences.  One could also argue that the incompleteness of the dataset invalidates the assumption that males are more invested in dogs than cats, especially since the data was pulled from only one social network (due to the limitations of personal information collection, Sysomos draws gender data from Twitter). 

In conclusion, although cats appear to be more popular than dogs online, these results suggests otherwise.  Possible explanations include a variety of factors.  The search terms may have multiple meanings with multiple uses.  For instance, people may refer to a person as a “dog” or talk about “dog-eared” pages or “cat scans.”  Although the overwhelming use of the search terms appeared to be relevant, the variations in the usage and meaning of the search terms certainly contributed some noise.  Additionally, the Sysomos does not collect data from Facebook, Instagram, Pinterest or other social networks, nor does it yield as accurate results as other tools.  All in all, the data comparison may be misleading, but it could also be revealing.  More research is needed to answer the ongoing debate of dogs versus cats, but until then, I’ll be posting cat pictures on Instagram and Facebook in an attempt to sway the outcome.  

9.5 million mentions for dogs positioning dogs as 14.7% more popular in the overall social media environment.

   Sources

http://www.sysomos.com/

http://www.onlineschools.org/visual-academy/cats-vs-dogs/

http://mashable.com/2010/10/21/why-does-the-web-love-cats/

[emilychambliss]

Avoid Gratuitous Data Puking.  "All data in aggregate is crap." (from a deck I don't recall)

[emilychambliss]

Mashable used my data analysis to conclude that Mitt Romney most likely bought a bunch of Twitter followers.   

[emilychambliss]

Social Media Allows Regular People To Have Political Influence

In my latest social media exploration, I look at influential posts that drove peaks in social media activity related to political topics and the elections. Unsurprisingly, I find that authors of influential posts were typically news sites (The Huffington Post, CNN, The Wall Street Journal) and politicians (Mitt Romney and Newt Gingrich) on Facebook. However, one influential post was authored by a random, politically-progressive Tumblr user who evoked 2,682 people to interact with the post. I conclude that through social media, individuals can wield political influence that rivals that of political authorities. I also measured conversation densities of influential posts to see the average amount of interactions per participant. I believe this is a suitable metric to gauge the intensity of conversations, i.e. high density means people interacted multiple times, responding to each others' comments vs. low density indicating that participants contributed one interaction, but did not participate further. It's interesting to see what riles people up and to make inferences about their attitudes and communities.

[emilychambliss]

Political topics related to the presidential elections trending on social media.

[emilychambliss]

I analyzed social media activity surrounding political topics related to the presidential elections to gauge what people are most interested in.

[emilychambliss]

Super Tuesday:  Social Media Participants vs. Voter Turnout