Can Social Media Monitors Predict the Future?

By Andrew Frank | October 6, 2008 | 20 Comments

There are many important tactical reasons for an enterprise to monitor social media: to protect its brand and reputation, gauge the effectiveness of its marketing, extend its customer service relationships, find and engage with influencers, and so forth.

But many social media analysts make an even stronger claim: that social media holds the key to strategic trend-spotting, providing specific, otherwise hidden, predictive insights into high-level decision-making processes. BrandIntel, for instance, claims, “we’ve cracked the code on early and predictive.

crystal_ball2_bmwpreview.jpg

Since this blog has recently gotten the attention of a number of these firms, who were quick to provide impressive proof of their listening power, perhaps a few will indulge me once more in exploring (and showcasing) this aspect of their craft.

How many social media monitors are specifically focused on extracting predictions? …and who will offer a sample prediction from the social media world? (Predictions about the future only please.)

Needless to say, predictions are inherently risky, and the most valuable the ones come with the longest odds. But understanding the kinds of predictions a firm can extract from the corners of the social web might help change perceptions about just how significant a domain this is. Think of it this way: a clipping service: a few thousand a month; actionable intelligence about the future: priceless. That’s why I’m predicting that social media monitoring will be a three billion dollar business by 2012. (Please refer to the disclaimer at the bottom of this page.)

Thanks again to all the firms who have responded over the past few days (and sorry to zap you again):

Biz360, Buzz.io, Buzzlogic, Cision, Converseon, Custom Scoop, CyberAlert, Enterprise RSS, Market Sentinel, MotiveQuest, Nielsen, New Media Strategies, Onalytica, PopularMedia, Radian6, RelevantNoise, Scout Labs, Sysomos, Techrigy, Trackur, Umbria, VibeMetrix, Visible Technologies…

20 Comments
  1. 6 October 2008 at 12:44 pm
    Tom O'Brien says:

    Andrew: you are giving us a real workout here!

    Our metric is the Online Promoter ScoreTM. It measures brand advocacy online and in several categories we have proven a strong correlation with sales. Changes in OPS are a leading indicator of changes in sales.

    http://tinyurl.com/58xm8d

    We are currently using this measure to predict the outcome (FUTURE) of the 2008 campaign. We publish advocacy for Obama and McCain every day along with the 10 words most closely associated with each candidate. Take a look:

    http://brandadvocacy08.com

    TO’B
    MotiveQuest LLC

  2. 6 October 2008 at 12:59 pm
    David Alston says:

    Hey there Andrew. I’m not sure about the word “predict” in the context of machine only situations. With the ability to cover all forms of social media and all of the conversations happening within these channels on a topic you then have the ability to look for trends for sure. Also, the ability to see which posts go “viral” and attract a lot of attention can also give someone insight into what topics are hot and of the most concern for a community. Knowing which bloggers are more influential (and their connected network of influence) on a topic than others could help someone predict which message could spread faster if it resonated. However I think the final “prediction” aspect is best suited to the human element be it analysts, agency experts or corporate owners – at least for now.

    Movie, music and financial industries have been trying to perfect the ability to predict for a long time and unfortunately haven’t landed on it yet. However, this doesn’t mean we should ever give up on trying :)

    Cheers.
    David Alston
    Radian6

  3. 6 October 2008 at 1:23 pm
    Dan Kidd says:

    Andrew,
    You have raised an interesting issue. We work with our clients on two different areas with social media. First is the ability to understand the customer experience and to provide direction on where our clients can improve that experience. The second is the impact that social media has on other potential customers that are exposed to the communication. The impact of social media is part of the mix of communications that a potential customer is exposed to in their buying process and affects their purchase behavior. We provide quantification of the impact of many types of communications that affect purchase behavior. The analysis of this impact provides our clients insight into what is shaping their prospect’s future beliefs and purchase behavior. Social media analysis in isolation of other media types and without understanding the impact of each communication that touches a potential customer can be very misleading.

    The analysis required to provide insights into what shapes future behavior is complex but is what we do for each of our clients. Unfortunately I cannot provide you with specific examples as this insight is highly valued by our clients and cannot be shared publicly.

    Dan Kidd
    VP of Sales
    Biz360, Inc.

  4. 6 October 2008 at 1:51 pm
    Jen Zingsheim says:

    Andrew,

    Hello again! You raise an excellent point, and it’s one that we make often. Monitoring is important, but what makes it truly valuable is being able to process that information and *do* something with it. Whether it’s providing customers the ability to sort through and analyze the data themselves, or having our team of analysts do it for them, turning raw information into intelligence (as you put it “actionable intelligence”) is our daily task at CustomScoop.

    Like Dan, I can’t provide you with concrete examples because our clients do consider this actionable intelligence. I can tell you that on a number of occasions our analysts have said based on past experience, we expect issue ‘x’ to take off in the blogosphere, and they’ve been spot-on. It’s one of the advantages of having smart, experienced people looking at the data, they can draw conclusions based on what they’ve seen before.

    Jen Zingsheim
    CustomScoop

  5. 6 October 2008 at 3:32 pm
    Martin Edic (Techrigy) says:

    I think its important to separate the tools from the uses they are put to. Predictive analysis is an inherently human analysis of trends in context, something even the vaunted semantics engines can’t do well. Our tool, for example, provides very complete data sets of results (text) and meta data associated with those results. Finding near future trends in those results is something our customers (service agencies, brand managers, market researchers, etc.) might build a practice around. We give you the ability to see into social media; the conclusions you make are your own.

  6. 6 October 2008 at 4:02 pm
    Dean Westervelt says:

    Hi Andrew, clever post. And although we weren’t officially invited to the party, I thought I’d drop in with a few thoughts. As someone who has a lot of prior direct marketing analytic background on my resume, I tend to agree with Martin @ Techrigy – semantic analysis of unstructured data certainly returns meaningful and actionable marketing findings (as Dan and others point out) but fundamentally is not a “predictive analysis”.

    However, when there is enough buzz surrounding an event (say the Super Bowl or the Presidential Election), a fun pre-post comparison can be made. For example, we (Collective Intellect) “predicted” three of the top four advertisers in terms of post-Super Bowl activity (blogs, boards, and online news references) associated with them. “Prediction” defined rigorously? No. But examining pre-buzz activity enabled us to predict who would continue the buzz on through and after the Super Bowl.

    Why don’t you ask all of us to predict Presidential winners for select swing states? (my team may disavow my existence now…)

    Thanks for the post!
    Dean Westervelt
    Social Media Analytics, Collective Intellect

  7. 6 October 2008 at 4:19 pm
    Leslie Bradshaw says:

    Andrew,

    The tag-you’re-it is a great collaboration/crowd-sourcing tool. Thanks for including us once again.

    In terms of predictive models, at New Media Strategies, we have a very unique take on this that you and your readers might enjoy.

    Over the course of the last 9 years of being in business, NMS has a database of almost 1,000 campaigns from which we aggregate results and provide our clients with a valuable tracking system that allows them to understand their campaign in the context of other, similar campaigns in a given category/vertical. From this, reality/context-based predictions, benchmarks and insights are gained.

    In addition to having this 9-year trove of data, we also continue to work on additional predictive models that we – like others in this comment thread – aren’t at liberty to share :)

    On a final note, like any sound shop, we also don’t over-position the ability to “predict” in light of entropic factors on campaigns – ranging from quality of asset and elasticity of client’s product in a market down-turn, to things like key talent being arrested and on/offline coordination and spend.

    Thanks again and don’t hesitate to be in touch.

    Leslie

    Communications Manager
    New Media Strategies

  8. 6 October 2008 at 5:29 pm
    Valerie Combs says:

    Hi again Andrew!

    Echoing a lot of the great comments already provided here – while the transparent nature of social media and the advancement of technology can give us unprecedented insight, I would agree that actual predictions be left to human analysis, which in our case, resides on the client side.

    That said, at BuzzLogic, we do take advantage of our platform to do a little of our own analysis (and very occasional crystal ball gazing:) on a variety of topics, across pop culture, politics, business, etc, which we publish on our blog – http://www.buzzlogic.com/blog/ – and for PR Week – http://www.prweekus.com/The-blogosheres-take-on-the-Wall-Street-meltdown/article/118345/.

    As an aside, we too are watching the Obama vs. McCain conversation. A quick glance reveals the number of sites talking about Obama (and the frequency with which content is published) outpace those focused on McCain, which positions Obama as having a stronger online grassroots presence. You can also ascertain which phrases are resonating with influencer sites as compared to all conversation participants, which tells you where and how ideas are originally seeded, and demonstrates the nature to which an influencer might generate a ripple effect for a particular meme.

    That said, all of the online discussion in the world doesn’t matter if voters don’t actually go to the polls to vote. Understanding and analyzing user generated content can get us closer to making accurate predictions, but to reiterate David A’s post, technology alone is definitely not there yet.

    Valerie Combs
    BuzzLogic

  9. 6 October 2008 at 5:54 pm
    Jenifer Manning says:

    Thanks again for including us in your test.

    RelevantNoise is not a tool that will predict the future, however it will help to spot trends and anomalies in data. RelevantNoise looks at trends and the tonality of the posts, which enables one to make educated assumptions. Also looking at the terms around a topic can give great insight as to the future of a product or event.

    Looking at the term “Super Bowl” leading up to the game will show not only the teams involved but other terms such as commercials and specific brands such as Budweiser which is expected, as always, to have great ads during the game. Go Daddy will show up as controversy does drive buzz, their ads have been controversial in the past and the pre event buzz did wonder what they had in store for us.

    Our analysts and clients use the data to understand current trends, and buzz on new and future products and garner intelligence that can be used to make actionable business decisions.

  10. 6 October 2008 at 6:36 pm
    Brett Safron says:

    Andrew,

    Interesting responses from everyone. There is no doubt that monitoring social media outlets can render insights into a company’s brand and reputation management far quicker than traditional media outlets can. This is quite evident from Cision’s perspective given our comprehensive coverage of all media types – from social media to online mainstream news; from podcasts and viral video to local broadcast news programs; from daily newspapers to trade and niche print publications. We see the natural progression stories and issues take through the various media sources and clearly see social media’s capability of “predicting the future”. Today’s post or online discussion on a company’s news or issue becomes tonight’s most viewed article on news websites, tomorrow morning’s broadcast news “What’s Hot on the Web” segment and finally tomorrow’s (or later) article on the newsstand. This process has grown over the years with the advent of each new media type with the immediacy factor increasing with each new type.

    The difference in social media and its unique ability to predict the future is based in the source of the content. No longer is there a requirement to be a paid reporter for an established publication. Social media has become a company’s mini-focus group but obviously much more public and unregulated. In the same way you pull representative consumers into a room to administer questions created to predict the future, social media outlets have become a quicker, cheaper and more honest and open forum doing quite the same exercise. At Cision, we believe there is considerable value to recognizing the full life cycle of a story or issue and tracking it throughout its course. If you can gauge not only how that story will play out in social media realm but also how it will proceed through other channels, imagine the predictive value that can be derived there.

    On a side note, we are in the midst of development of a product that can predict which company’s product will crush its competition in 2010, which candidate will be elected in 2012 and which year the Cubs will actually win the World Series. We cannot predict, though, when this product will be ready. :-)

    Brett Safron
    Cision, Inc.

  11. 6 October 2008 at 11:27 pm
    Josh Carr Superstar says:

    There was a lot of heavy conversation up above about prediction and yadda yadda yadda. I am just excited about being mentioned.

    I also liked the sound of $3 billion – I am putting it into my investment deck right now.

    Just so I don’t sound like a fool after a whole bunch of smart people filled the page with jargon, here are some unorthodox ways we have been using our engine:

    Competitive Positioning – we were asked the question “our competitor is using such and such in their positioning, we want to know how this has been picked up in the industry”

    Market Research – Many of the questions you would ask have already been answered by the web. Save some money.

    Link relationship clouds – hard to explain until you see it, but basically we can show you the hierarchy of your inbound and outbound links. It is graphic way of showing your web ecosystem.

  12. 7 October 2008 at 1:07 am
    Mark Rogers says:

    Hi Andrew,

    Thank you for including the folks at Market Sentinel in this. I have two quick observations on how live information from systems such as ours can generate real value by answering key questions:

    a) What’s working? What’s stalling? Tracking Net Promoters/advocacy/positive sentiment online can quickly give brand owners the chance to reinforce success and quit funding failures.

    b) Can I have what she’s having? Looking at patterns of language and citation (linking or mentioning) around competitor, or even unassociated brands can give marketers clues as to how they behave to position themselves in the same way.

    This can help marketers to position themselves exactly so that their businesses thrive and steer them away from bad choices. Over time it will make for better businesses, growth and profits.

    As for a prediction about the future: I would urge investors to review their portfolios looking for indicators on Net Promoters (overall positive sentiment) relative to customer service. If you cross check this metric with performance relative to the average sector P/E ratio, you can find real bargains in a market as volatile as this one. These companies will survive bad times by looking after their customers, their competitors may not.

    A UK-focussed example:

    http://www.marketsentinel.com/blog/2007/12/what-is-it-about-john-lewis-and-nationwide

    Best wishes

    Mark Rogers
    CEO Market Sentinel

  13. 7 October 2008 at 8:43 am
    Andrew Frank says:

    The quality and intensity of this dialog reinforce my belief that we’re approaching a turning point in social media intelligence…I’m guessing that a key side-effect of the explosion in global market uncertainty and volatility is driving business leaders to seek new sources of intelligence to better navigate in rough waters…Mark provides a great example…anyone else following this line of reasoning with customers and prospects?

  14. 7 October 2008 at 10:14 am
    Nick says:

    Challenging question Andrew. For years statisticians and data miners have been trying to claim predictions. Ultimately they build a model and the value of their predictions is as good as the model. At Sysomos we build models and validate them statistically towards other observable factors (such as sales, views, etc). The value to the end user is a dynamically evolving and statistically sound trending report that offers evidence of what is forming.

    Best
    Nick Koudas
    Sysomos Inc

  15. 7 October 2008 at 2:25 pm
    Giles Palmer says:

    This is fantastic – it’s like some kind of swarming game

    OK – sorry – a day late :) been busy!!!

    My view is that we should try to be as scientific about prediction as we can. That’s to say – try it, measure your success and try to understand what happened. The problem with we human-types is that we are good at telling a story AFTER the event and making the ‘facts’ fit our interpretation, but when it comes to prediction – it’s another matter altogether.

    I think prediction is very difficult, and i’m amazed at how few people try to take this predict -> measure -> evaluate approach. It’s pretty easy to do after all.

    always last to the party

    giles

  16. 7 October 2008 at 3:11 pm
    Giles Palmer says:

    one other observation i have is that once something shows any promise with prediction, lot’s of others act on the information and effectively make the prediction happen. But then reality kicks in and usually shows that the prediction wasn’t as good as once predicted! and the cycle goes back into sync.
    I wonder how many analysts consistently over a statistically relevant period out perform the stock market index trackers ? I recon none. But i may be wrong :)

  17. 7 October 2008 at 7:57 pm
    Janet Eden-Harris says:

    Andrew,

    Congratulations on starting a great dialogue.

    When the blogosphere first burst on to the scene, marketers have been looking for research that would make them comfortable using this medium as a predictor of overall consumer behavior. Bloggers talk a lot. About everything. But are they representative? Can their discussions be used as a predictor of buying habits?

    JD Power’s (formerly Umbria) latest research shows a definitive YES. The latest research, soon to be released, comparing blogger discussions over the past year discussing their intent to buy an automobile, correlating to actual sales shows a direct pattern. Intent discussions rose earlier this year; so did sales. Those same discussions began to fall in volume through the summer. So have sales. Month after month, the data tracks.

    I think the holy grail is within sight.

  18. 8 October 2008 at 11:26 am
    Simon McDermott says:

    Hi Andrew, Attentio is a leading European player with US based clients such as J&J and HP. We have predicted frequently for clients and non-clients. We have suggested that LG would have significant customer care issues because of negative sentiment in key markets (with nice phones but poor quality). We also regularly predict movies that will flop or succeed (early) based on negative sentiment and low buzz. Buzz is better at precicting sales or profits when the item is low price and easily obtainable, but the “butterfly effect” is more apparent in larger ticket items. Simon

  19. 9 October 2008 at 7:31 pm
    Blake Cahill says:

    I am glad to see so many of my peers interacting with you Andrew. Looks like we are all generally heading in the same direction. Although, I would propose that know of only one peer David Rabjohns, CEO at MotiveQuest, who is going to shave off his hair on You Tube if their election predications don’t hold true. Here to hoping that Dave has hair post election day.

    http://www.prweb.com/releases/2008/10/prweb1421754.htm

    Cheers,

    Blake Cahill
    Visible Technologies

  20. 21 October 2008 at 11:51 am
    Martin Edic (Techrigy) says:

    About the $3 billion: I’m sure everyone on this thread would like to no the thinking behind that number! Sure hope you’re correct…

    In a way this is like doing sentiment analysis (which we do). It can only be accurate when vetted by humans. Software-based sentiment is notoriously poor at context, irony, etc. Building a predictive model would have to take this into consideration also. You might combine demographics, sentiment, authority and volume into a model that creates a river indicator that says it looks like things are moving in this direction. Have to think about that. We have the ingredients, just need to find the recipe.

Comments are closed.