While both teams claim a business like mentality it’s clearly not business as usual for Giants Fans.

The emotions shown on the fan pages are startlingly similar except on the excited dimension.

Hard to say it better than Giants fans say themselves.

“I'm getting so much flack from my West Coast friends! Just because I live in Los Angeles CA - doesn't mean I can't love the NY GIANTS! What the heck!!..right!! Whatever, I'm sporting my NY BLUE BABY!!! So super stoked!!! sorry, that's west coast/surfer talk for - SO STINKIN EXCITED!!!” 

Want to explore more about the conversation on the Patriots and Giants pages on Facebook, check it out here on Fathom Analytics.

Growth, Recession, Credit, Recovery, War,  Asian Crisis, 9/11, War, Katrina.  The timeline of the words used by the fed when they talk about the economy.

The Federal Reserve makes the minutes of their open market committee meetings available online (6 years delayed because of… well I have no idea why they need a 6 year delay). 

We parsed it up to make it fit into our data model and loaded it into our system (would it kill them to edit and proof the transcriptions?, you’d think if they’re important enough to be delayed 6 years they’d be important enough to not misspell Greenspan’s name).

There’s a wealth of information here which could/should/will be analyzed for years and frankly I got my only C at Kellogg in macroeconomics (Prof Eberly was great this is just a subject that my brain is not wired for, semiconductor physics yes, macroeconomics not so much).

I read a TechCrunch article that has some interesting statistics provided by EdgeRank Checker.  I put together a little graph showing the click through rate data that is text based in the article. 

Clearly having a facebook page with a significant number of fans gives a strong base for future marketing efforts with 50% better click through than general ads and 3x better click through vs. facebook ads, brands can start to determine their ROI per fan simply by understanding the dramatic reduction in click through costs.

For the most simple example, imagine I’m an insurance company looking to grow my automotive insurance business.  This report shows cost per click of ads in auto insurance are well over $50.  So for every 1,000 fans on my facebook page I can deliver a click through and save $50.  Plus, I can do it over and over again.  With 500,000 fans on a facebook page and assuming you don’t want to inundate your fans with offers, restrict it to 1 offer every week.  Those 500,000 fans would be work $1.3M in savings from direct click through costs alone.

A simplistic example for sure but a direct savings of $1.3M is a start.  When we imagine multiple product groups with various offers, and the benefits of the impressions, marketers can likely develop very strong ROI analysis of building their fan count on facebook pages simply through reduced costs to future marketing efforts.

As I've worked the last 7 years in building metrics and tools to help companies understand the reasons why people do what they do, 3 facts have become evident:

1) Pretty much everyone is online

2) Online communities are reflective of or an enhancement to traditional social relationships

3) People I know are the best source of product recommendations

The digital trails consumer conversations leave online are a good approximation for the world at large.  The latest pew Internet research report show online population reaching  79% of all Americans.  Furthermore, rather than asking people questions about what they think, we can just observe what they say.  We don't have to ask people if they do or don’t do something we can observe them.

Online communities are as real and vibrant as any other and are built around long term relationships.  Whether it's a community like Facebook that looks like friendships I've developed throughout life or LinkedIn, the connections I've made in business, or on a forum like Rennlist, my boss's favorite Porsche aficionado site, the community is strong and vibrant, with lots of on-going relationships that grow and fade with new people coming into ask questions from experts or lurkers that keep up with the community without contributing or the occasional visitor that just wants to see what the experts think.  These communities are strong, vibrant and have their analogs in social situations in the offline world.

The impact of person to person recommendations is stronger than advertising or other methods of company sponsored communications, when you want to know what car to buy you ask a car nut, when you want to know how to get the best deals on frequent flyer miles you ask a friend with a passion for travel.  Increasingly, though you'd find your friend saying "go look on flyertalk" there's a thread for that (can I trademark "there's a thread for that?").  I was recently browsing flyertalk to figure out international fares for a 2 year old ( I've never booked a child fare before and didn't want to get screwed) while I was there I saw a new miles promo on United it had been posted and had hundreds of comments before it even reached my inbox from United later that day.

Given that people online are pretty much everyone, the communities represent real social relationships and people I know give the most trusted recommendations we should clearly expect to see that a social media metrics measuring peoples' recommendations to each other would be reflected in sales.

We measure advocacy, which is the number of people recommending a brand above any others, and have seen strong relationships between the metric and sales in a number of categories. (I'll post more in the future specific to the nuts and bolts of it)

While the scales are different we can see the positive correlation in advocacy and luxury market share for BMW. Looking at the scatter plot and doing the regression confirms the relationship and it's statistical significance with a p-value of 5% (we're 95% confident that there is a positive correlation between advocacy and luxury market share).

Given the complexity of the automotive category and the economic conditions in this time frame it's amazing that a single factor model can so nicely line up.  With an understanding of incentives to dealers/buyers, advertising spend and supply issues I'm certain the model could be improved but at the end of the day we're validating that social media does matter.

Social media clearly has an impact on sales in this category and others.  Many times the relationship will not be this clear or the category may be so complex that a single variable model cannot tease out the results or there could be errors in the metric (source spam/problems, bad language model, etc…)  but if you believe in the 3 tenets, it’s not surprising and quite intuitive that there should be a link between a social media metric and sales.

I wanted to post the full comments from a recent message board posting located at http://www.netnography.com/showthread.php?4138-Correlating-Online-Recommendations-and-Product-Sales&p=4295

First it’s a great review of our methodology developed a couple of years ago, followed by some pointed questions in the comments and my response.  I will follow up with additional examples in the coming weeks and look forward to follow on discussions with Michael and Vito.

Correlating Online Recommendations and Product Sales

MotiveQuest describes itself as “a new kind of research company that uses online anthropology to explain consumer motivations.” Although they don’t use the term “netnography,” that’s what their “anthropological research approach” is. Explaining their approach, the company says:

“MotiveQuest uses new software, combined with an anthropological 
research approach, to help you better understand online customer
 conversations about your brand.”

The company has developed a metric it calls THE ONLINE PROMOTER™ SCORE, which “measures the number of people online that are recommending a brand to others.” MotiveQuest worked with the Mini Cooper auto brand to measure the impact of online promoters to sales. They published a case study on SlideShare that includes the graph below. They say that the graph “shows the correlation the monthly change in online promoters for the previous month versus the change in sales. For example the point in the upper right is MINI’s monthly change in sales from April to May 2006 and the change in online promoters from March to April 2006. Statistical analysis gives 99.8% confidence that the metrics are positively correlated.”

Analyzing the relationship between the social media conversation and product sales is clearly something marketers in all fields are keenly interested in. It’s a positive development that research companies who use netnographic techniques and offer consulting services based on their findings are trying to find ways to do that.

Some Questions About the MotiveQuest Case Study

This case study from MotiveQuest seems like a good idea, and the results are quite interesting and worth analyzing. Given my interest in this topic, I wanted to find out more about the relation the authors of this case claim to have found between their Online Promoter Score and Mini sales, and so far I have been unsuccessful. I would appreciate it if MotiveQuest could contribute to this discussion by explaining what they did, and answering some questions that came up while I was trying to replicate their procedure.

1. I found the official sales from Mini in the U.S. for the period December ‘05 to July ’06, and the data presented in the case study is significantly different from those figures. What data did you use? I also tried using data for only North America sales, only Europe sales, entire world sales, and some other combinations. Unfortunately, the percentage variations do not match with any of the latter. Did you use official sales data or your own estimations?

2. In slide number 7, you show that you can do all this statistical analysis with only seven observations, and in later slides the presentation assures high statistical significance. What methodology are you using? I would be very surprised if by having only six observations (percentage changes) , that is, seven values minus one value to calculate the differences, you are able to back up all the results. Also, with such a small sample, is it valid to throw away the observations that you don’t feel comfortable with (i.e., outliers)? Additionally, you are working with some form of time series, are you using the proper econometric procedures to control for trends or seasonal changes? Unfortunately, this is what confuses me the most and makes it hard for me to buy the results shown in the presentation.

3. In slide number 8, the case study implies that there is some form of causality between your score and the sales. How do you know that? Isn’t it simply the case that more people buying Minis results in more posts, and that your Online Promoter Score has no predictive value? I am very interested in this, because if you actually have found a predictive measure, you have a gold product if, also, you are capable of showing similar out-of-sample results.

Finally, and assuming that the data and regressions presented in the case study are accurate, I proceeded to regress your Online Promoter Score with other economic variables, such as inflation, unemployment, gas prices in the U.S. and total car sales in the U.S. Factoring in those variables allows me to get an even higher correlation between the Online Promoter Score and Mini sales. Therefore, I could conclude that the Online Promoter Score actually has some information in it (taking into consideration my previous assumption), but under no circumstances is it able to produce the results shown in the presentation.

Some initial replies

It's been a couple of years since I did this analysis and since this study we’ve found links in a number of other categories, many parts of which are client confidential but I’m sure there are pieces which can be shared more broadly so that we can show more examples of the relationship.

I'll start with one comment. I don't really believe this is revolutionary but a natural relationship we should expect to find, based on beliefs: that the most influential factor on consumers’ decisions is each other and that people posting in social media are at least a sample of the overall world. There are many data points supporting both of those beliefs, so it’d be surprising to not find some sort of relationship between online discussions and sales. It happens that measuring advocacy through the online promoter score does that.

Addressing your points in order:
1. We used data provided by Mini for their sales for the US only, we got permission to share the indexed values but not the raw data. This may be very similar (or exactly the same) as the data commonly released through automotive news. If you'd like to send me your data, I can compare it to the old files I have.

2. The data earlier in the slide set shows a period during our first engagement with Mini and their agency during which the relationship is clearly evident, the full data set that produced the regression shown above is longer.

As far as controlling for "econometric" variables like seasonality, trending, etc... typically we look at share in order isolate seasonal and trend effects and we can compare the changes in share vs. changes in share of or raw advocacy, it really depends on the dynamics of the category.

3. Causality is a dangerous word… The relationship is leading in many cases but often times quite strong when aligned as well. We’ve increasingly worked with our clients on strategies to help them increase the advocacy and drive stronger market share and sales as a result.

That’s really interesting to see our data improving your regressions, I’m always surprised that the relationship is as strong as it is given all of the other factors that would have a strong impact on sales. If you would like to chat contact me at: bmiller at motivequest dot com and we can share more.

First, the idea for this post is based on Jeff Matthew’s post: An Open Letter to Ben Bernanke, he takes the approach of counting occurrences of how the economy was described in various earnings calls to come to the conclusion that things are improving (even if slowly).

We collected over eleven thousand earnings call transcripts from 1/1/09 through 9/30/10.

Our sentiment tool agreed with Jeff’s analysis… the economy is indeed being discussed more positively than in the past (the midpoint of the shaded area) while the width of the shaded area shows that the number of earnings calls mentioning the economy is also falling.  Overall, seems a positive indicator of the future in the transcripts of the calls.

Finally, we can look at the words most associated with economy over the time period.  Very clearly shows the changing nature of the business climate (and the despair in earnings calls in early 2009). Overall, some interesting usages of typical social media analysis tools on a different source of data to show the voice of CEOs rather than the voice of the consumer.

… but, I also think it tries to take a relatively abstract concept and makes it tangible.  The graph shows the amount of association that each brand (the drop) has with a particular concept (the blades).  I considered changing the length of the blades to illustrate the relative sizes of each concept but overall, I think this visualization doesn’t really do much so I’m pausing on the development until I think of a “cooler” way to bring this to life.

Regardless it’s an interesting look at what can be done with pretty minimal effort in the .Net framework and GDI+.