Marketing and Research Consulting for a Brave New World
Subscribe via RSS

Tuesday, the results of a landmark study were made public proving that the quantity of social media conversations about a brand has a statistically significant relationship to changes in its sales.

“Researchers today announced the results of a landmark study that measured the impact of “consumer word of mouth” in six diverse categories, finding that online and offline consumer conversations and recommendations account for 13% of consumer sales, on average…About one-third of the sales impact is attributable to word of mouth acting as an “amplifier” to paid media, such as television, with consumers spreading advertised messages. The study was based on sophisticated econometric modeling of sales and marketing data.”

–Word of Mouth Marketing Association (WOMMA), sponsors include AT&T, Discovery Communications, Intuit, PepsiCo, and Weight Watchers.

This industry learning comes on top of an academic paper by Prof. Wendy Moe at the University of Maryland that showed a correlation of .8 between social media listening data and brand equity metrics derived from survey questions.

So now that we know that social media data are truly DATA…with predictive value, how do we act on this?

First, research needs to take social media listening seriously

As I said in an earlier post, “…finding the prediction question”, research needs to become an equal opportunity employer.  If the data has predictive value, it should be hired! Traditional survey researchers need to come to grips with the proven predictive value of social media data.  We need to stop treating social media listening as a hobby and find its mainstream roles alongside surveys and other important data streams such as clickstream and transaction data.

Second, we should create new brand metrics from social media data.

In my last post about the marketing ATOM, I demonstrated how building brand audiences is the key to success in a digital age. People become part of an audience for a brand that is significant and relevant to them and audiences talk about the brands they join.  Hence, it is no surprise to me that social media would provide an important set of brand metrics.  Once researchers enrich brand KPIs beyond the venerable survey tracker with social (and other digital) data, they will become an agent of change for the enterprise.   Social KPIs will encourage marketers to focus on building their audiences, creating content that is worth sharing, and tracking advertising and promotional campaigns through peoples’ willingness to talk about them and share them. In fact, turning social media data into must have metrics has already been done via Social TV ratings that both Nielsen and Rentrak offer and it affects pricing of TV spots.

Third, we extend.

I plan to investigate if social media listening can replace continuous tracking of attributes.  I am optimistic that we can do dipstick studies with attributes but track brand perceptions throughout the year via social data, creating a leaner, more agile, and more effective tracker program.

I would like to see us begin to partition social media conversation by client segment or audience.  To illustrate, it is now possible to match social media profiles to customer lists using machine based logic that matches on name, e-mail, etc.  As such, for example, Verizon could create a segment of customers called “On the bubble” who are more likely to defect and, as an aggregated segment, their social media conversation could be monitored.  What are they saying about Verizon, competitors, life, TV programs, etc.?  Where are the conversations occurring?  This would be very powerful and the technology in fact does exist.

I’d like to see research turn report card trackers into predictive engines built from time series data that includes social media, digital, weather, survey tracking results, etc.  Our goal is to get ahead of future trends for a brand so we can influence these outcomes positively before they happen.

I urge research panel providers and brand websites to encourage social log-in so the power of Facebook and Twitter profiles can be harnessed.  In this way, interest profiling and ad targeting merge into one thing.

Yes, the genie is out of the bottle but as you head into this world of integrative measurement, please be mindful of rigorous practice for social media listening.  Different providers can actually produce very different data streams for the same brand, depending on whether they access the full Twitter firehose, include all social channels, how their semantic engine works etc.  To understand the complexity on the last point, consider social media listening for Target the retailer.  Extracting meaningful conversations on the retailer “Target” rather than Seeking Alpha talking about a company hitting its financial targets is not trivial. Also some conversations map to brand preferences while others map to a hot promotion or topic.

This is a significant stage in the journey the ARF started in 2008 when I was Chief Research Officer.  We began to explore how social media listening could become a valuable partner or even partial replacement for surveys.  The first meeting included Unilever, Procter, and General Mills…a highly unlikely event…but we all agreed that social media listening had tremendous potential for insights value creation.  This then became the big springboard into the ARF Research transformation super-council.

Now that we know that social media data are quantitative and predictive, we must create research protocols to harness their full transformative power.

Note: For both studies, the social media data streams were provided by Converseon to whom I am a strategic adviser.

 

 

 

Tags: , , , , , , , ,

Comments

11 Responses to “Conclusive proof that social media data predict sales…now what?”

  1. Congratulations on the research. You may find our research with Northwestern from 2008 interesting to review. http://www.businesswire.com/news/home/20080521005011/en/MotiveQuest-Develops-Metric-Correlate-Online-Brand-Advocacy#.VG3pCJPF-ag

  2. Joel Rubinson

    thanks David. Yes, Motivequest has been a driver and thank you for adding to the citations.

  3. Great post, Joel! Do you know if there’s a technical paper available describing the WOMMA study?

    Haven’t there been studies demonstrating that user generated content like tweets can predict things like box office sales? I seem to recall some papers from 7 or 8 years ago.

  4. You are gushing with excitement for your client Joel, but I don’t get it. “Conclusive proof that social media predict sales”, where

    Study one: “A similar finding for online word of mouth requires more comprehensive data on impressions and/or mentions, which are not yet publicly available”.

    Study two: “future research may also investigate the ability of changes in brand sentiment inferred from social media to predict shifts in key performance indicators such as market share or sales”

    • Joel Rubinson

      Good morning Byron. The proof comes from marketing mix modeling that Analytic partners conducted on 6 brands using multistage modeling. Social media conversations were strongly associated with sales change. the online WOM comment refers to making a statement about the IMPRESSION impact vs. a paid impression. The problem is that Twitter does not release IMPRESSIONS. However, the finding regarding the impact of conversations is incontrovertible. study two is one I hope to do. Got funding? want me to be part of a consortium?

  5. Joel, great article. In response to Byron’s questions, here is a link with a bit more detail, with more to be uploaded soon: http://www.womma.org/ReturnOnWOM. The study is based on Structural Equation Modeling by Analytic Partners. They were able to explain an average of 13% of sales, of which two-thirds was based on offline WOM/social influence, and one-third based on social media. The impressions issue related only to the social media component. Researchers found a wide range of impression value from offline WOM, ranging from 5 times the sales impact of paid media to 200. I know that the researchers, and WOMMA, will be eager to share more of the results with interested parties. In the interest of full disclosure, I’m the current chairman of WOMMA and my company, Keller Fay Group, supplied the offline WOM data to the study as Converseon contributed the social media data.

  6. jonathan sinton

    Hi Joel,
    TNS has been doing this work commercially for a number of years now. We’ve done this successfully in a range of categories and countries so know it can be done, but is clearly not easy. Larry Friedman spoke at TMRE about some of this work recently. I completely agree with your 3 steps and we’re already doing much of this – suggest you trade notes with Larry when you see him. more here: http://www.tnsglobal.com/intelligence-applied/marketers-future-ready-you-now
    Jonathan

  7. This is true on the B2B side, as well. The algorithms powering most predictive lead scoring solutions, including SalesPredict, take into account these social signals when calculating lead scores. More at http://www.salespredict.com/our-approach

  8. An exciting declaration, with many exciting developments on the horizon. Studying the amplification factor social media has on mainstream media is a key element for the strength of the correlation you’re looking at. We’ve also found that census data for the geographic area being studied can further improve reliability. Thanks for sharing your news. Great post… and it is exciting.

  9. This very much confirms what we’ve been saying for some time based on our own research findings and I’m delighted to see it.

    I totally agree that the market research industry has to start taking social insight seriously, which with a few exceptions it has failed to do to date. My view is that sooner or later evidence of the power of social insight will overcome the rhetoric of those who wish to rubbish it and this is exactly the sort of thing that’s needed to achieve this.

    Congratulations to all concerned!

  10. […] As Joel Rubinson points out, the two studies together demonstrate word-of-mouth offers not just data power but also prediction power. “If the data has predictive value, it should be hired!,” he says. […]