Here is a link to the Klurig Twitter analytics covering the Boston Marathon, 2014, for additional maps, graphs and analytics, including sentiment analysis.

Below is Dom Nicastro’s article.

B2B Marketing and Twitter: Strong Marriage, Legitimate ROI?

is an annual long distance (90 km) cross-country ski race (ski marathon) held on the first Sunday of March in northwesternDalarna, Sweden between the village of Sälen and town of Mora. It is the oldest, the longest, and the biggest (in terms of participants) cross-country ski race in the world.[1] The race was first run in 1922, inspired by a run by King Gustav Vasa in 1520.”

For runners, in road races like marathons, all finishers receive a medal. In Vasaloppet, you only get a medal if you ski faster than 150% of the winning time.

To give you an idea what that means: take Chicago Marathon, 2013.  The winning time for men was 2:03:45. For a medal, you would have needed to run at 3:05:37 (150% on winning time) and for that race, this would have put you in the top 6% (place 1237 of 21488) in the men’s division. Running a marathon close to 3 hours is really hard and involves serious training and time commitment. Men and women use the 150% medal time based on winning male and winning female respective.

So, what this means is that getting a medal in V asaloppet is a big deal.

A friend of mine, Kalle, participated in Vasaloppet this year. His time 6:47:03 did not break the medal time at 6:21:50 but he wasn’t that far off either. So I figured I do some analysis to see what is going on. Vasaloppet is 90 km or 56 miles.  There are 8 legs of about 9 – 15 km.  Each leg goes to a different town, akin to Boston Marathon, which goes from Hopkinton to Ashland to a few other towns and finishes in Boston.

I downloaded the results for Vasaloppet, which includes splits for each leg for Kalle’s class (men, 55-59).  There are 832 skiers in that class. During the pre-processing step, for each skier, for each leg, the pace (min/km) was calculated and used in all subsequent results. I could simply have used the split but I wanted to see the actual pace to get a better feel for how fast they ski.

A few goals with the analysis:

1. For each leg, I wanted to create a box plot to see which leg seemed slower compared to faster.
2. I wanted to overlay Kalle’s time to see how he did on a per leg basis compared to his competition.
3. To see how Kalle fared compared to medal time, I also overlaid the medal time.
4. This was a good opportunity to look at violin plots, which I haven’t used in the past.
5. In addition, as a secondary plot, I wanted to use a heat map to use colors to see the differences between each leg as well as between each skier.

Using R and the ggplot2 graphing library, it was very easy to create the graph (let me know if anyone is interested in the raw data and the code). A few observations

1. There are actually two plots as I overlaid the box plot on top of the violin plot.
2. The first leg, from the start to Smagen, there is a 2 km uphill which causes a bit of a zoo-like conditions. That is why the spread from fastest (~3.5 min/km) to the slowest (~13.8 min/km) is so large. The remaining 7 legs are more normal.
3. Legs 2, 3 and 5 seem faster than the other legs, which makes sense as the terrain is more conducive to faster skiing (read: more downhill)
4. Kalle’s splits are in red and the medal splits are in green
5. Kalle is doing well: in all legs, except the first leg where he got caught in the zoo, Kalle is either faster or almost faster than 75% of the skiers in his class.  On leg 3, he beats the medal pace.
6. Based on the plot, it is easy to see that in order to win a medal, your time needs to be clearly better than 75% of all skiers. Note though that the winning time is based on first overall (not first in class) and in this case, we are only looking at one of the classes.
7. Based on the violin plots, we can see that early in race, the violin plots are more pear shaped.  As the race goes on and people start to become more tired, the max thickness of the plots is moving upwards (the majority of the skiers are getting slower) causing the last plot to have a body like a bodybuilder.

The next plot is a standard heat map, also based on R and the ggplot2 library. The plot is small but high resolution because there are 832 skiers vertically. For better details, download the png and zoom in.

The plot shows the faster skier on top and slower skiers (based on finishing time) at the bottom.  Thus blue color means faster pace and red color means slower pace.

Two observations stand out: First, yes, leg 1 was a zoo and you can see the 10 – 14 min/km pace in red colors. Secondly, as mentioned above, legs 2, 3 and 5 are clearly the faster legs, perhaps leg 5 is the fastest leg.

Let me know if you want the code or the data.

In this post, we will use Social Network Analysis to dive deep into the Twitter followers and friends of 17 of the speakers and panelist of the week-long event.  The reason for 17 is simply that out of 30 people, only 17 supplied a Twitter handle. We also added the handles for some of the organizers including myself (@dagh) for a total of 22 Twitter handles. Below is the graph, and below the graph are the details.

Some quick observations:

1. KDnuggets with 18,400 followers rocks!
2. The main clusters are based on the 22 initial seed users. That makes sense because that is where we started.
3. VishalTx has a large following, specially if you consider that Vishal Kumar is the CEO of Cognizeus and therefore is running three Twitter handles in @VishalTx, @Cognizeus and @AnalyticsWeek
4. Other prominent users are @gretaroberts, @paulsonderegger and @DeborahMCooper.

The included Twitter handles and their respective names are:

1. PaulSonderegger – Paul Sonderegger – Oracle
2. kdnuggets – Gregory Piatetsky – KDnuggets
3. chazard – Chip Hazard – Flybridge Capital
4. molecularist – Charlie Schick – IBM
5. greels1 – Michael Greely – Foundation Medical Partners
6. Lynch_BigData – Christopher Lynch – Atlas Ventures
7. DeborahMCooper – Deborah Cooper – DeborahmCooper.com
8. eureqa – Michael Schmidt – Nutonian
9. judah – Judah Phillips – Smart Current
10. clarkjacker – Ben Clark – Wayfair
11. analyticsraj – Raj Aggarwal – Localytics
12. rama100 – Rama Ramakrishnan – CQuotient
13. cesarbrea – Cesar Brea – Force Five Partners
14. dxsimmons – Bill Simmons – DataXu
15. suthoff – Brian Suthoff – Localytics
16. gretaroberts – Greta Roberts – Talent Analytics
17. imdaviddietrich – David Dietrich – EMC
18. AnalyticsWeek – run by Cognizeus – tweets by @dagh during the Analytics Week
19. VishalTx – Vishal Kumar – Cognizeus
20. skbhate – Sachin Kumar Bhate – Cognizeus
21. cognizeus – twitter handle for Cognizeus
22. dagh – Dag Holmboe (myself) – Klurig Analytics

Through this analysis, our goal was to combine the 22 twitter handles into a social network analytics chart to see if there are individual users or clusters of users, who seem to be more prominent, perhaps have a bigger influence on the Boston and national and international analytics community.

There are different ways to measure influence using Twitter as well as using standard social network analytics.  Within Twitter, we often look at five measurements:

1. Followers – sometimes is mostly a popularity contest but it still gives an indication of influence. A person with a million followers wields probably more influence than a person with 10 followers
2. Mentions – the more you are mentioned, the more influence you have? Not necessarily but it is an indication.
3. Retweets – retweets is probably the most reliable measure of influence. Because a retweet is a reflection of ourselves, we often only retweets things that we think are good.  So, someone who is often being retweeted, often tweets good tweets and can therefore be considered an influencer
4. Recency – is the person recent? Is the person still tweeting periodically? If a person is not recent, the person lose influence quickly
5. Topic – a person might be an influencer in one topic but perhaps not the topic that I am interested, thus for me, the person is not an influencer

We can also look at measuring influence using social network analytics centrality measures such as closeness, betweeness and eigenvectors.

1. Closeness measures how close you are to all other nodes in a network. The person who is closest to everyone else could be called an influencer because if I want to distribute a message to the whole network, I can go this person because this person can, in theory, do it most quickly and cheapest.
2. Betweeness measures on how many shortest paths you are on.  The person who is on most shortest paths could be considered an influencer because for any person to reach any other person in the network, they have to go through this person more than any other person.
3. Eigenvector is similar to the Google Pagerank in that your importance is partially measured based on the importance of your network neighbors.

Even though we have used the above techniques many times in the past, in this case, we did not use them.  Instead, we were simply interested in creating an exploratory social network graph.  To do that, using R and the twitteR library, we downloaded the profile for each of our 22 users, as well as the profile of each of their followers and friends.  Using profiles, we ranked them based on topic (analytics), removed profiles who did not seem to be analytics inclined. We started out with 49,136 users and wilted them down to 3,639 users connected via 5,468 edges.

Using the igraph library, we exported the network data to Gephi where we did some additional pruning and created the AnalyticsWeek network graph.

AnalyticsWeek was organized by Cognizeus.

In this post, we will walk through the AnalyticsWeek tweets based on the social media analytics. Klurig Analytics’ specialties are social network analytics and text mining.  Combing those two disciplines with data Twitter data and we get a powerful set of knowledge about a community and an event.

Each evening, we tweeted a very comprehensive tweet stream (at times, many tweets per minute) with the hashtag #bosaw2014. In addition, while the event was on, periodically and automated, Klurig use the Twitter API to download the latest tweets for the hashtag to automatically generate graphs and tables for up to the minute understanding of users and keywords. Through each presentation, we shared the names of top user (most frequent Tweeters) information on Twitter together with a message to each of the top users.  This personalized message help generate additional buzz about the event.  For the whole week, we had 125 unique users tweeting about the event. We also shared the information at the end of each evening with the organizers so that they could give out a prize to the user who tweeted the most.

The above graph shows the total number of tweets for each evening.  It should be pointed out, as we will see further down, that @AnalyticsWeek was the top tweeter by far and these numbers includes @AnalyticsWeek. The first night, Monday night was a bit slow but mostly because this was the first night. The remaining nights, we jacked it up to about 300 tweets per night – a tweet a minute –  with a top on Thursday night at the Microsoft NERD center discussing Marketing and big data analytics.

These are the unique tweeters for each evening and it is on average 35 unique tweeters per night.  As we can see, Healthcare Informatics at Microsoft NERD got most unique users tweeting. It is interesting to see that both top nights were at Microsoft.  Were there more people at Microsoft because it is more centrally located, closer to Kendall Square, more known than District Hall or Hack/Reduce? Or are there other reasons?

It makes sense that Friday night is a slower night as some people check out immediately after work. It also seemed that there were less people attending on Friday.

Sentiment analysis is not an exact science but it gives an indication of the mood of a group. In our case, the mean sentiment per evening is interesting. Clearly, Friday is the happiest evening – perhaps because it is Friday.  On the other hand, my guess is that people displayed a good mood because Greta Roberts gave such a great keynote address with many surprising and positive comments.

Wednesday at Hack/Reduce talking about Finance and Insurance also got people in a better mood.  Christopher Lynch gave a fantastic keynote address with a good amount of humor included, which possibly helped raising the mood. Cause and effect is always tough.

Either way, based on the sentiment, overall for the AnalyticsWeek, it was a very successful week with people enjoying the keynote speakers, moderators, panelists and the great opportunity to network with great people. Kudos go to Cognizeus for the organization of the event.

.

This chart shows the number of tweets as measured in 10 minute intervals during each night from 5pm to 10pm.

Two take aways; first, each evening started on time except for Friday which started about 30 minutes late as we were waiting for additional people. Secondly, it seems that there were more tweets during the keynote address as compared to the panel discussion for Wednesday, Thursday and Friday (close), while the opposite on Monday and Tuesday. Not sure why. It might be easier to tweet during keynote address because it is only one person talking while more jiggered during the panel discussion because more people talking.  Unclear on that one (sometimes, there is a tendency to pattern-match random information).

Showing the sentiment in 10 minute intervals.  Not much to see except for the fact that on Friday, people seemed to get happier as time when on.  The other days, things seemed to stay the same over time.

In this chart, we are using cluster analysis to try to understand if we can see a difference between each evening event.  Each event has a different topic so a cluster analysis could potentially detect that. Based on the tweets, it is tough to see a difference, except for potentially Tuesday, Healthcare Informatics, which seems different from the other evenings.

Top keywords over the whole week, which words were used the most in all tweets (after stopwords have been removed). Boston is used the most, perhaps not surprisingly.

Every time we tweeted (as AnalyticsWeek), we added the name of the speaker to the tweet so the number of times a person’s name shows up is a function of the how long they spoke and how many times they were tweeted during their speech. In this case, it seems that Judah Phillips and Greta Roberts got the most tweets.

There seems to be a lot of tweets on the health track with marketing and workforce analytics close after.

Note that this is one of these charts that can be done a few different ways.  For instance, should you include the twitter handles of those people speaking?  Should you include potential words such as day and top? It all depends on what you are looking to display. In this case, apart from stopwords, we display it all.

The above graphs shows the top keywords on a nightly basis, primarily used just to verify that there are differences, which we also noticed for the cluster analysis above.

Yes, the top tweeter was yours truly, @AnalyticsWeek, as stated above. The remaining tweeters are as listed above and we say thank you to all of you who helped us socialize this event.  Interesting fact is also that there were a total of 125 unique people tweeting during this event.

All work is done in R, using the twitteR, ggplot2 and ggthemes packages – thanks to the authors for making these packages available to the open source community.

In a later post, we walk through the social media analytics from the event

The keynote speakers for each day started with Paul Sonderegger from Oracle, Gil Alterovitz at Harvard med school, Christopher Lynch at Atlas Ventures, Judah Phillips at Smart Current, and finally Greta Roberts from Talent Analytics.

Klurig Analytics was a Silver sponsor of the event with our team responsible for social media and social media analytics. Using the handle @AnalyticsWeek and tweeting both text and images to the community, enabling non-attendees to follow along as well as leaving a historical trail of the event. For instance KDnugget used our tweets for a great post about Monday night where Gregory Piatetsky moderated a panel.

MONDAY

The first night, at District Hall, was all about Big Data Analytics.  Paul Sonderegger from Oracle gave a really good keynote address with key points such as

1. there is more data being born outside of a company than inside the company, about the company
2. there is a difference between data-first and model-first but you need both. The whole is greater than the sum
3. predictive analytics do not predict big things; predictive analytics predicts small things
4. build a data strategy
   1. think in terms of data market share
   2. create proprietary data assets. Take public data and combine it with own data and get better data
   3. use data to make data, like how Google is using data. Create a big data flywheel

The panel discussion was moderated by Gregory Piatetsky. A selection of comments:

Gregory Piatetsky: Data scientist will be replaced by automated algorithms? Data science in a box? Can data science be automated?

David Jegen: Humans do what humans do really well. Machines do what they do really well. There will be a coexistence. PayPal had problems and put machines on it, realized that you need humans and now they coexists. Machines to curate data. When do you bring in a human to curate, to train the machine. Humans are artists.

Anand Rao. In AI, we have proven that it is really hard to replace a human brain.

Chip Hazard The data preprocessing part will be handled by machine. The key is to ask the right question,  the right question is really hard to find and that is where we need a human.

Paul Markovitz: data scientist will be doing less grudge work (machines handles that), instead they will be doing the fun work.

TUESDAY

Tuesday night, at the Microsoft NERD center, the discussion centered on Health Informatics and Analytics. Gil Alterovitz gave another great keynote address including an overview and a presentation of SMART. Key pointers are:

1. re: cloud computing – the analyst just need the data, doesn’t care of where the data is
2. today, if you have a disease, your healthcare provider gives you advice. Using sensors, ques can be given to you automatically. For instance, when the sensor senses that you are low on blood sugar, ques can be given to you to certain food at certain times
3. Genomics will be a game changer. For about $1,000, you can get your whole genome sequenced
4. There are many genomic standards, we need an adapter that can talk to all standards and we have one named SMART
5. create an eco system of SMART apps, compete on SMART apps
6. SMART Genomics Advisory Platform – different modules can be re-used in other apps
7. China is very interested in SMART because their data is more centralized
8. Key is to abstract away the tough parts of building apps, it should be easy to build apps

The panel discussion was moderated by Charlie Schick and consisted of Michael Greeley, Allen Kamer, Kris Joshi and Timothy Andrews.

Charlie Schick: Even though there are hackfests in healthcare programming, but it is too slow. How do we overcome the slow speed? How do we get all together so providers have access to data and systems.

Michael Greely: Many great initiatives and accelerators but they are not adequate yet. Optum Labs is better to teach ecosystem. Optum Labs is great place to learn the ecosystem of healthcare as for analytics and bigdata.

Allen Kamer: We struck partnerships with different groups to partner to help us build products, early adopters such that we can prove. Partnerships are a great way of building products. Helps patients getting help faster and more cost effective. Optum Labs is helping smaller companies to get started. Companies can reach out to Optum Labs with different initiatives to partner around things.

Kris Joshi: The innovation is very local. The flipside is different from other places in the country. It is important to partner. Need to get the right asset built and then a different way of scaling the asset.

Tim Andrews: Healthcare is very regulated. There is no one institution in charge of healthcare. It is very hard to get them to share. Need to get hold of multiple targets to work together, but it is really tough in healthcare.

WEDNESDAY

On Wednesday, at Hack/Reduce in Cambridge, Christopher Lynch gave a keynote address about Finance and Insurance.  Key points:

1. Big data becomes transformationable when power of analytics is given to the common man
2. The cognitive impact of data science – infrastructure has little intelligence
3. Deploy technology for actionable outcome – deliver the right data at the right time
4. Realtime is old. Intime is more important, as ‘in time’ to make a decision. Time to value!
5. How do you impact time to value. That is what is important today
6. Big data is to look at all data inside a company, including social media data and financial data and all data in between
7. Simplicity is transformationable. Need to deliver quick data in easy to understandable way
8. Opportunity for New England to be a center of competence in big data and analytics
9. Make sure your technology is giving back to the community
10. A great company is the company that cracks the code of delivering big data analytics to the rest of us

The panel was moderated by Bill Fearnley.  Sample question:

Bill Fearnley: What is the key challenge that lead to big data solutions?

John Muller: Analytics is the key thing. We create big data and that is a challenge. But the main idea is the analytics and the data visualization

John Raguin: Marketing data analytics  have been around for decades. But tools doesn’t mean anything. The real question is: will this customer be a profitable customer for me? Technology is immaterial as long as it gets the job done.

Deborah Cooper: Without big data, my group would not be as responsible as they should be.  Big data is key. The challenge is too many data non-integrated data sources, and too expensive.

Michael Schmidt: Important to be able to explain what goes into a prediction. His team created a bracket for March Madness. 75% accuracy with minimal effort. Easy to explain, easy to use.

THURSDAY

Back to Microsoft NERD center for an evening discussion analytics in Marketing.  Judah Phillips gave a fantastic overview of various marketing concepts related to big data analytics. Sample comments:

1. When is the right time to quantify? When you have the data, need information about the data, consistency.
2. Right time when the company is cross-functional, when you can improve a situation and when revenue is at risk.
3. All data is not actionable.
4. Analytics value chain: 1. understanding, 2. collecting data, 3. reporting, 4. analyze and socialize, 5. optimize – 6. demonstrating that the analytics team provide economic value, that the analytics have an economic impact on the bottom line.
5. A/B testing yields significant gains if executed correctly
6. What does a complete analytics team include? you need the architecture team, the reporting team and the analysis team.
7. Analytics can serve marketing, product, sales, executive and IT

The panel was moderated by Cesar Brea. Sample question and comments:

Cesar Brea: what is the undefined problem that you are trying to crack?

Raj Aggarwal: Instead of building apps, Localytics built tools for manage apps.  Sell shovels instead of digging for gold.

Rama Ramakrishnan: The average customer only shops twice a year, therefore it is really hard to build a recommendation engine because person who bought product A might not want product B

Bill Simmons: if you build stuff for NASA, ensure it needs to be built in a large number of congressional districts

Ben Clark: Radical about keeping it simple stupid. Prefers A/B testing to MVT (multivariate testing). Business value can be evaluated with simple metrics; traffic, conversion rate and average value. Almost all marketing analytics use the same three data points.

FRIDAY

At District Hall, where Greta Roberts, Talent Analytics, CEO, give a really fun keynote address about workforce automation.  Key points:

1. Insight comes from many data sources. Need to additional value.  Torture data until it confesses.
2. The important thing is how we mix the data, how it is pulled together.
3. Tracking to outcomes matters immensely in data science.
4. Workforce analytics is about optimizing organizational layout
5. Solve the business challenge (not the HR challenge)
6. Using workforce analytics, we can do things like: we can predict that this person will be a top data scientist
7. New Deloitte report – HR analytcs is in the top 5 global trends – HR Analytics is huge and it will grow!
8. It is a huge trend because employee expense is the top expense in any company
9. Start with HR analytics even before you get complete data – worry less about completeness – get started
10. If employees are leaving prior to reaching breakeven, they are a loss for the company.
11. The key is to use workforce analytics to help predict which employees will stay past the break even point.
12. By using prediction, we can also see what makes a good employee good, compared to a not so good employee
13. When you make decisions, you need to use both the analytics and the human intuition.
14. The most exciting is reducing attrition and increasing performance in HR
15. HR analytics helps to put people in the roles where they want to be.

Abby Cashman led the panel as a moderator and among other great question, Abby asked the panel the following question: how do you define ROI, how do you measure this, what are the metrics that you use?

David Parker: We need to keep clean: red, yellow, green – we know what color the banker. Started at 6% green, now 24% green – good trend. Coach to behavior based on metrics.

David Dietrich: Measure to ROI – innovation work – people were volunteering – free and open source tools. Who is creating new IP and who can replicate that – this is really important for ROI

Lisa Croteau: measure metrics by having people using the system that they created. Having people in the company to look at standardized data. Good record keeping system. Good expe
rience working with techies. Measure success by counting phone calls when people asking questions about the system

Melissa Arronte: Measure by customer satisfaction surveys – do you use the data that we provide? A few successes in analytics will pay for the team.

Digital Marketing Files:  Are Marketers Forgetting to Talk Today?

When I was the editor of a small daily newspaper in Massachusetts, we had monthly reader advisory board meetings.

A dozen or so readers — picked carefully by our team based on demographics and engagement with the newspaper — literally came to our office, and I greeted each one at the door. We had a targeted agenda, maintained a lively discussion and produced action items.

The result? Story ideas for my staff, rotating columns in our editorial section by advisory board members and some brand recognition in our four small communities.

Continued

Phase 1 – Following.  In the early days of social media analytics, the big thing was the number of followers or fans.  If my social media following was larger than your social media following, then I win. Brands worked hard to generate the maximum number of fans, often by paying them with cash or other offers just to get them to sign up. So social media analytics was all about the numbers of fans and followers and we spent a lot of time trying to increase our following.

Phase 2 – Engagement. After a while, we realized that the number of fans and followers didn’t really mean anything. If my following doesn’t do or say anything, they are useless to me.  Sure, I can brag about my following but if they don’t buy anything, don’t provide any insights to me, don’t tell others about my great products and services then there is not much value. Instead, we started to measure engagement in terms such as social sharing, likes and comments.  This was much more valuable because even a smaller but very engaged following is better than a large, mostly quiet following.  With an engaged following, I can better realize the value of social media marketing in increased sales, better feedback on my products and services and so on. Forbes (Mark Fidelman) has a good article on top-10 brands with respect to engagement, 10 Lessons from the Top 25 Most Engaged Brands on Twitter. If I run a support organization, I can get users answer other users’ questions, thereby generate savings in my customer support organization. By measuring engagement, we can tie real dollars to a social media campaign. Currently, most social media analytics companies are in the engagement phase.

Phase 3 – Understanding.  Social media analytics is slowly evolving into the next phase where we are trying to understand who are fans are.  Having a fair understanding of our fans, we can better communicate with them, read ”‘we can better engage them” thus increase sales and other benefits.  The understanding is done via user and keyword analysis, sentiment and emotional analysis, and geographical analysis.  We are drilling down and analyze the keyword usage, and sentiment and emotional state of each individual fan and follower. All of this is automated and displayed on easy to read dashboard to help the marketer target users individually. For instance, a disgruntled user might post negative comments on Twitter.  Via alerts, marketers are instantly be notified and can limit the damage by directly reaching out to the user. It is all quite powerful. Full, 360 understanding of customers is the next phase in social media analytics and that is where we are heading today.

Phase one of social medi analytics was mostly about number of followers, fans, posts and tweets. The idea was that if you have a large number of fans, then you were doing well. That makes sense and often a site with more fans is doing better than a site with less fans. At the end of phase one, we begun to understand that the sheer number of fans is less important than what how these fans are engaging. To use an extreme example – a site with a thousand fans who do not engage is not as useful as a site with 100 fans who talk a lot.

As we move into phase two of social media analytics, we are starting to understand that what people are saying is more important than how much people are saying – quality vs quantity. If we continue to use the example above, we see that a site with 10 fans having a discussion with a particular goal in mind is worth more than a site with 100 fans where the discussion is not focused.

Before we continue, a word of caution – social media is large and the value of your social media effort (ROI) is directly linked to your social media goals. Sometimes you want a large group of people who talks about anything and everything, and sometimes you want smaller, more focused discussions. It is all about your goals.

In phase two of social media analytics, we are measuring and analyzing discussions. We try to understand which users seem to be leading discussions and which users seem to be following. Keyword analysis is used to increase the understanding of language. Through user and keyword analysis, together with sentiment and emotional analysis, we get a reasonable good understanding of users’ behavior, whether they are our customers, potential customers, our competitor’s customer or simply users interested in some subject where we have a stake.
Klurig Analytics have developed a phase two social media analytics service. Our technique is used in a number of different places including live sporting events. This graph shows the sentiment of fans from Boston talking about Boston Celtics versus fans from Dallas talking about their Dallas Mavericks. The data is based on Twitter postings and the graph was shown live.

In this game (Dec 12, 2012), Celtics was leading most of the game until towards the end, where Mavericks went ahead a few points. At that time – about 10:10pm on the graph – the sentiment between the Boston fans and Dallas fans changed. At regulation, the game had tied. The game then continued with two five-minute overtimes where Celtics finally stood as the winnner. Again, this is reflected in the Boston fans’ sentiment a few minutes after 11pm. Also, take a peek at the top graph at around the 11pm to see the increased number of tweets during the overtime.

As a marketer, you can not afford to miss an increase or decrease of this magnitude in a discussion about your products, services, competition and industry.

We use the social media analysis technique in sporting events mostly for fun. In reality, you would use Klurig Analytics’ technology to measure your and your competitors’ fans’ discussions for at least the following primary reasons:

1. Gain a better understanding of your user’s behavior and sentiment. Better understanding leads to better ways to market to them.
2. A solid a understanding and a real-time feed back leads to better risk management. As the saying goes: in the past, an unhappy customer told 10 of his friends. Today, that unhappy customer will tell a million of his friends. Solid risk management is critical today, and it starts by being aware of the risks in real-time.
3. Measure campaign effectiveness. Use our technology before, during and after a campaign to understand campaign effectiveness.
4. Marketing campaign trials; play what-if, test A/B testing across cities and products.

Click here to go directly to the article.  The following is the first two paragraphs.

Social media return on investment (ROI) is simply a measurement of efficiency. It’s a lot of things to a lot of people: “return on inactivity,” “return on innovation” and “return on engagement.”

However, in a stricter sense, social media ROI is defined as a measure of the efficiency of a social media marketing campaign. This definition might sound complicated, but in reality, it’s quite simple.

Read more at Social Media Examiner…

Are your executives demanding quantifiable ROI of social media campaigns? Are soft metrics such as “engagement” and “awareness” no longer sufficient to prove its value?

Join us for a free webinar, where we will show you how calculate the ROI of social media and prove its financial impact on your business. Webinar registration.

• Learn how to measure the financial value returned from each dollar invested in social media campaigns
• Compare the effectiveness of social media campaigns with traditional media such as TV, print, radio and outdoor
• Monitor KPI (key performance indicator) trends and optimize your social media campaigns to maximize returns.

Led by Dag Holmboe, CEO of Klurig Analytics, moderated by Mala Sarat Chandra, Lecturer in Social Media at the University of Washington, this free 30 minute webinar will run on Wednesday, May 25th, 2p, EDT, 11am PDT.