Trendspotting with ThemeStream

December 20, 2010

About a year ago we (MotiveQuest) were talking with a potential client in the liquor business.  He asked if we could have predicted the Ice Tea Vodka Craze using our software. We thought this was a pretty interesting question and over a few glasses of Ice Tea Vodka we started to ponder.

There are lots of ways of looking at historical buzz and being smart after the fact but we thought it would be interesting to create an algorithm that didn’t just look at buzz but looked at momentum, exaggerating the effect of growth visually.  After playing for a while, a new tool was born that we christened “ThemeSteam”,  A tool that indeed could have predicted the Ice Tea Vodka Craze.  ThemeStream is now a tool that we use almost daily to see what is hot in the categories we care about.

ThemeStream determines the words most correlated with the category conversations for each period and then highlights (through exponential comparison of deviation from the average value) those words that are most dynamic.  This allows us to see emerging, waning, and seasonal trends.

For example recently a client was interested in seeing the impact of Jamie Oliver on the food conversation among parents.  Here is the chart.

Source: Data comes from looking at a combination of food and parenting datasets focused on discussion of kids lunches N=10,995

In this chart we see when the impact of Jamie Oliver begun to have an effect on moms’ lunch decisions and discussions as he is putting focus on nutrition education out-of-home.  Note that we see the Jamie Oliver conversations growing rapidly as soon as the show – Jamie Oliver’s Food Revolution begins to air.

This was flagged as a issue for our client well before they would otherwise have been paying attention to it.  We are using this new tool for early trend identification across many projects from food to pharma to consumer electronics.

Tom O’Brien


Healthcare debate analysis

September 9, 2009

We (MotiveQuest LLC) have decided to start tracking & analyzing the healthcare debate using our online anthropology tools and techniques.  The first report in this series can be found here:  The Raging Debate

For this series we will be monitoring & analyzing the online conversation around healthcare and providing weekly updates as to what is driving the conversation, advocacy for different options and the emotional tenor of the chatter.Healthcare Topics

Here is one chart from the report showing the key topics and drivers of concern – you can see the landscape has changed significantly from June through August.

You can see that rationing shows the biggest increase from June through August.

We will be posting a follow-up report next week which will reflect today’s speech by President Obama.

If you are interested in more information you can contact me at tobrien at motivequest dot com.

Thanks – TO’B

Baking a social media cake

April 27, 2009

(The other day a client was probing about our methodology. They were asking lots of smart questions and as we worked through the answers, they asked me to send a write-up. Technically, it isn’t a social media cake, but a social media analysis and insight cake. Here it is.)

During our discussion, you asked how we bake this cake – here’s my diatribe answer to that question. (BTW, baking a cake turns out to be a great analogy!)

To bake a cake you need ingredients (data), mixers and pans (tools) and cooks (strategists).  Your questions were focused on developing a better understanding of each one of these. (Disclaimer, I’m not the CTO or the head of research, so my answers may be a bit shallow.  I promise that I’m happy to get both/either of those people (cc’d above) on the phone to dive into more detail.)

Any good cake starts with the right ingredients.  For MQ this means starting with the right data.  The right data is always category, project and client specific.    Here are our thoughts about data.

  1. Including the data you want while excluding the data you don’t what is hard.  We discussed the Visa example, and you have experienced this with your brand.  De-spamming and de-duping is an important part of the job.
  2. Brand mentions range from 2% to 30% of the relevant data (depending on category, so only collecting brand mentions will miss most of the relevant conversation.
  3. It isn’t helpful to get all of the data.  (Splogs – or spam blogs have LOTS of brand mentions – but aren’t real and VERY hard to eliminate.   Focusing on and gathering the sites where the most people are participating virtually guarantees the elimination of Splogs.
  4. 5 years of expertise in solving the challenges above with a combination of technology tools and linguistic programming expertise provides MQ with a significant advantage over most of our competition.

You can’t bake a cake without measuring cups, mixers, whisks, spatulas, pans and an oven.  We can’t do our work without a broad sophisticated toolkit.  One of the fundamental challenges in analyzing large amounts of unstructured text data is that you simply can’t make sense of it in any sort of manual fashion.  Sophisticated software tools are the answer, but how those tools are developed and deployed makes a difference.

  1. Language is fluid over time and across categories.  Tools must be too.  Every one of our tools is parameter driven and allows the strategist to adjust the linguistic model and other parameters for the category and project at hand.
  2. Simple measures (counts, brand mentions, sentiment) are not useful for understanding why people do what they do and without understanding, we don’t gain the insight of what to do next.
  3. Understanding requires more sophisticated tools.   That is why we have tools for passion peaking, measuring motivations, word association, brand advocacy and many, many others.
  4. We have many tools available, but the tools used and the order of use is very project dependent.

Finally, you can’t bake a cake without a cook.  Well, I guess you can use a box mix from the grocery store and bake a cake without a cook, but if you have a special occasion and hire a chef to bake a cake, you don’t expect to get something from a box mix.  You expect a cake cooked to your specifications for your occasion.  You expect something unique, professional, surprising and delicious for your special occasion.  In our world, the dashboard and organized data providers are the box mixes.

MotiveQuest brings the experienced professional chef who will design and deliver a custom cake according to your exact interests, needs and specifications.

  1. Experience matters.  We have very sophisticated toolsets and it takes at least 6 months to get a strategist (most of whom have advanced degrees and backgrounds in consulting or planning) trained and productive.
  2. The tools are complex because the problems are complex.
  3. A single strategist is responsible for the entire project including data collection, organization, analysis, insights and recommendations.
  4. Each project has its own specifications and requirements.  Communication, solid project management and client involvement are all required to achieve good results.
  5. MotiveQuest Senior Leadership is deeply involved in every project to ensure results that meet your needs.

I will send the MQ capabilities presentation in another email.   Our case studies are very marketing (not data or technology) focused for a reason.  We don’t get hired to provide data or technology, but instead to solve real world marketing and communication problems.  I’ll be happy to provide more detail about any of them.

Thanks very much for your time on these issues, and I hope this helps answer some of your questions.


Sara Palin & SNL: Josh Bernoff has Questions . . . .

October 23, 2008

We sent out an email blast last week about our BrandAdvocacy ’08 website and got some questions back from Josh Bernoff at Forrester (blogging here).

Here are his questions – and our answers. (By the way, Josh’s questions were pretty representative of the questions we are getting from everyone.)

1. Is there a bias in people who create or react to content online towards the democratic and liberal candidates?

  • First, if you look at the trend data on Brandadvocay08 it is clear that sometimes Obama has led and sometimes McCain has led, so on a very simple level this does not seem to be true.
  • Second, when we created the dataset we were careful to create a representative dataset of both ordinary people and political commentators.  For example on the site we have links to one of each for Democrats and for Republicans We found that there are plenty of Republicans who are online also.
  • In addition we can look at the underlying constant of regression the CNN Poll of Polls and Obama’s Online Promoter™ Score to determine if there is an “online bias”.  The CNN Poll of Polls shows a constant of 50.8%, MQ’s OPS shows a constant of 49.6%.
  • MQ’s data does not reflect a particularly liberal bias, other slices of data might, if we look more recently perhaps one could be emerging but over the 9/1 to 10/20 time frame there doesn’t appear to be one.
  • This also depends on your belief in the bias of the poll of polls; I’m not a polling person and can’t say if there is a bias in the poll of polls.

2. Do people online just like to talk about the Wow activities (like Sarah Palin on Saturday Night Live) vs the real Issues?

  • Every day we highlight the words that are most highly correlated with each candidate.

  • They show what issues the nation is associating with each candidate.  Sometimes it is true that they focus on the Wow stories but we see that they are just as likely to be seriously discussing the real issues (taxes, God, wealth, etc).  Don’t underestimate the power of forums as a place everyday people share their fears and thoughts with others like them.  For example take a look at this discussion on’s forums

3. How does Online advocacy relate to traditional polls?

  • From one perspective, its apples and oranges, polls attempt to measure the outcome of the presidential election if it were held on a particular day, MQ is measuring the number of people who are advocating a particular candidate.
  • On the flip side, if you believe as we do that people advocating to each other is an important piece of the future success of brands, products (and candidates) then we hope to show that the candidate with the most advocates will be the winner of the election.
  • Important caveat, we don’t have an electoral college or battleground states so at best we can hope to use advocacy as an indicator of the popular vote.
  • Using the data from the CNN Poll of Polls as a comparison metric, HYPOTHETICALLY, let’s assume that CNN Poll of Polls represents “truth”, we can apply straight forward regression to determine if the Advocacy metric for Obama is statistically related to the Poll of Polls data.
  • Regression shows that Obama’s Online Promoter™ score is statistically significant in linkage to the CNN’s poll of Polls results at > 98%

  • Furthermore, we can look to see if perhaps MQ’s results might lead the traditional polling methods; the relation appears strongest at 4-days leading.

The strongest statistical link from Obama’s Online Promoter™ Score to the CNN Poll of Poll’s results appears when we make Obama’s

OPS a 4-day leading indicator of the Poll of Polls

Obviously this detailed post may raise even more questions – feel free to post them in the comments section.


Hertz #1 By a Mile – Thanks Walter!

October 1, 2008

On Friday afternoon I returned to the Hertz LAX location at 12:30 PM with 1 hour until my flight left. If I didn’t make the flight, I would be stuck at LAX for at least 6 hours. To make things worse, I was flying on United – which at LAX is the last of 9 stops that the Hertz bus makes.

So I got out of the car and started sprinting toward the bus. Someone called out “Sir, Sir”. I turned around, thinking someone wanted to give me a receipt when a Hertz employee named Walter told me to get back in the car so he could drive me around to United.

We had a nice chat and Walter told me that LAX has a policy of trying to help disabled people, families and those running really late (like me) get to the terminal quickly.

Well, I made my plane and got home to see my family that night thanks to the initiative and efforts of Walter at LAX. This is a great example of one person going above and beyond the call of duty to help someone – and I definitely appreciate it.

Please pass on my thanks to Walter and to his supervisor.

Tom O’Brien
Hertz Fan Forevermore

133 MM Blogs Indeed

September 23, 2008

Interesting “state of the blogosphere” report from Technorati on 9/22.

All you have to do is look at this chart to understand the essence of what’s happening. Far from being the massive phenomenon that has been shopped around for years, it is obvious that there are a very small number of influential blogs.

I mention this because it is common for companies in the social media analytics space to brag about monitoring over 100 MM blogs as part of their service. How hard could it really be when >94% of the blogs out there have not been updated in the last 120 days?

Blogs are not so much a CGM phenomena as they are a new publishing channel. If you have something interesting and compelling to say you have a chance to get a real audience. Only 0.057% will ever crack the Technorati Authority of 50+.

Not saying blogs aren’t important, but let’s not kid ourselves about there being 133 MM real blogs out there.


Everything you need to know about YouTube

July 31, 2008

“An Anthropological Introduction to YouTube”

Title not withstanding this latest gem from Prof. Mike Wesch is not only is this a great presentation (mixed ppt and video) but a must watch for anyone who needs or wants to understand YouTube in particular and Web 2.0 overall.

See Mike Weschs’ table of contents for this video here

The reasons, motivations and payoffs for why people connect online are all in here – and it is very people-centric. All marketers will do themselves a favor to sit down and watch this video.

Hat Tip to Laurel Papworth