Spiral Marketing: The More You Know, The More You Can Know

June 24, 2009

Sbarro Isn’t Listening To Its Customers (A Cautionary Tale)

There is really no specific reason this blog post HAS to be about Sbarro Italian restaurant – there are plenty of other brands out there that are ignoring a great deal of interesting and even actionable information about brand perception, consumer experience, quality, etc.  It’s just that, coincidentally, several factors converged today to bring Sbarro to mind: 1) I used to like Sbarro’s quite a lot; 2) I have noticed that the quality of Sbarro’s has declined considerably – undoubtedly a result of the “franchise” effect – and so I almost never go there anymore; and 3) I noticed the following post on Twitter:

“Sbarro’s – never, ever, ever, EVER again. Never, not ever. EVER! #cardboard.”

So I searched Twitter to see if Sbarro has a Twitter account – apparently not.  I then searched Twitter for “Sbarro” and found comments that ranged from good, to OK, to bad – some examples:

mulebennett: @williamfleitch [Michael] Send ’em to Sbarro‘s for an authentic NY slice [/Scott]”

mohalen: @Arrens i thought i heard sbarro‘s was going out of business; i dont believe I have seen any in Ca.”

tastynsweet: sigh im at the mall right now eating sbarro they got good pizza but exspensive lol”

lynnsmithtx: Sbarro‘s pepperoni stromboli has 44g of fat & 2,470mg of sodium – a day’s worth of each.”

tiffany_dorrin: Ate at Sbarro for the first time in a very long time. I felt like I was in high or middle school again.”

BabyFresh360: Just ate some pizza from Sbarro Ewwwwww it was Sofa-King Disgusting…I Feel Sick my Fellow Twits : (“

GoodStuff4Free: Free pizza at Sbarro‘s for students with good grades June 23 http://linkbee.com/70IT

gadbearr: @joshuaharrell Sbarro…. yummmm!!”

So since Sbarro doesn’t have a Twitter account, I went to their corporate website, finally found the “contact us” link in the bottom navigation bar, and clicked on it to get a contact form.  There were 3 choices for my stakeholder role – Consumer, Franchiser and Real Estate Developer.  So I selected “Consumer,” added contact information, and in my message said that I noticed they were missing out on a lot of very useful information, including the original post: “Sbarro’s – never, ever, ever, EVER again. Never, not ever. EVER! #cardboard.”

This is where the story gets very interesting: I posted this comment, and got a rejection message back from the Sbarro’s website saying something to the effect that the “Contact Us” page should not be used to send messages it perceives as negative.  Are you KIDDING me?  Just tell me before I spend the time writing a comment that “We are very interested in anything you have to say, as long as it conforms to our vision of ourselves.  Any observations that do NOT conform to this viewpoint are out of line, and we will reject them out of hand.”

Good to know . . . now.

I am, as I would guess are the majority of you, of the school of thought that believes 1) ANY information or feedback is useful; and 2) I would rather hear negative comments that will help me improve my product or service, rather than flowery platitudes.  I JUMP on any hint of negative commentary, and try to get to the bottom of it.  In fact, some of the strongest client relationships I have ever developed have had a critical juncture, where the client perceived us in a negative way, and by rapidly addressing their concerns in a proactive, forthright and honest manner, I’ve ended up with a much better client relationship and a much stronger advocate of our services as a reward.

Now, it is clear Sbarro doesn’t belong to this school of thought, and while ultimately I believe that to be a fatal arrogance, I am not going to try to fight through the layers of bureaucracy they’ve built to ensure they don’t hear the voice of their most important asset: their customers.  I think, though, that there is a clear lesson for companies a little less vested in viewing the world through their own pre-determined lens.  While the dynamics are different for various sectors and industries, customers keep you in business; in the consumer world, you have an incredibly diverse, mobile and judgmental group of stakeholders that you ignore at your own peril.  And through emerging social media, the ability for those consumers to share good or bad information or experiences is incredibly rapid and remarkably efficient – the tweet that started this whole situation was sent by someone with 750 followers; through that one tweet, she told 750 people to never eat at Sbarro’s.  And that is just one of the 200 Million+ Facebook users and the 35 Million+ Twitter users (to name only two of the social networks out there).

The lesson of this cautionary tale?  There is more information about your brand than you could possibly imagine being shared on these social platforms.  You have two choices as a consumer retailer (or any business, for that matter): you can adopt and adapt your customer-facing processes to gather, assess, and act upon this vast reservoir of information, and through that process refine your view of your strengths and weaknesses, your opportunities and threats, the perceptions and misperceptions about your brand that you need to identify and act upon.

Or you can adopt the second option – the Sbarro option: only listen to what you want to hear, and filter and tune out the rest.  I would not recommend this option, but if you take this course, understand that you do so at your own risk.

Kev

April 8, 2009

Twitter: What Does “Deeper Analytical Tools” Mean In Twitter Context?

Mark Davidson, a consultant specializing in Micro-Blogging, Social-Web Strategy, and Social Media Marketing (www.twitterstars.com / @markdavidson or www.twitter.com/markdavidson) posed the question – on Twitter – “I think that as more businesses become serious about using Twitter, there is a definite need for deeper analytical tools. Agree or disagree?”

I’ll be posting more on Twitter as I can (stealth start-ups are not conducive to a lot of blogging activity – I may also be the world’s laziest blogger), but for those 17 people in Outer Mongolia that have yet to hear of Twitter, it is a “micro-blogging” site, allowing you to share 140 character thoughts (a “tweet”), forward others’ tweets, and reply directly to people.  You may “follow” people, in which case you see their tweets; others may follow you as well, in which case they see your tweets.

There are really two “vectors” in the physics of Twitter – one is the reach of a given Twitter user – I believe Mark is well over 32,000 followers, and some have a quarter million or more followers.  On a good day, I probably have about 8 (more or less).  As with anything we measure, I suppose, there are those that seek to acquire followers simply for the sake of an audience – more must equal better, even though the content they deliver doesn’t warrant this assumption.  Others – informed and experienced individuals such as Mark, @chrisbrogan, @stejules, etc. have a clear focus on their audience and what they wish to deliver or portray to that audience.  This is the second vector in Twitter-physics: the messages themselves, composed most reliably of words and phrases (although it is possible to link to external content).  For the sake of completeness, it should also be said that there are a broad spectrum of other users – individuals with focii on very targeted subjects, such as the Myelin Repair Foundation; networkers such as @zaibatsu or iJustine, growing broader reach; businesses (and mercifully, we won’t mention Mars’ Skittle brand debacle here); and of course, we pedestrian social butterflies just looking to connect and communicate – with a dangerous and disruptive anarchical substrata typified by people like @Irant (that would be me).

So getting back to Mark’s question: “I think that as more businesses become serious about using Twitter, there is a definite need for deeper analytical tools. Agree or disagree?” – I don’t see how it would be possible to disagree with this statement.  But the question itself begs further analysis: what do we mean by “deeper analytical tools?”  If you buy into my hypothesis that there are two vectors that drive Twitter – the audience reach of a given user, and the content of the messages sent via Twitter – then that would suggest that there are two, frequently interrelated, categories of analysis.

The first: what is the breadth and quality of my reach as a Twitter user?  If I am a company account (or represent a company), how extensive is my reach?  Am I reaching the right people – the target market for the company?  What degree of influence do I exercise over this audience?  If I am in a consumer setting, seeking to influence through PR the perception of a particular brand, then the targeting may be less important than the size of or influence on the audience.  If I am Bugatti, however, then the size of the audience is clearly (at least in my mind) secondary to how well targeted my audience might be, and my influence on that audience; if I am talking to tens of thousands of individuals, of which none can afford a Bugatti, then my impact on the business is far less than reaching a much smaller group of wealthy individuals who can.  So in terms of Audience Analytics, one must build a model that explains and informs the business regarding these two factors: Targeted Reach (rather than absolute audience size) and Influence. Now, this may or may not be a good example for Twitter, but is meant to illustrate an important consideration when evaluating the first Twitter vector – Audience Reach.  As with anything in business, you have to know who you are trying to reach, and why, to be effective.  What it is difficult to argue, however, is that the degree of influence you have on your target audience is never unimportant, which leads to the second Twitter vector: the content itself.

Twitter itself is a vast cocktail party, where all and sundry might be the topic of discussion at any given point in time.  One group may be talking about social media techniques; others about graphic design considerations; others about various companies, brands, products, social causes; and yet others “riffing” on some silly topic or another.  All of these conversations occurring simultaneously.  And all happening in 140 characters or less, which leads to a great deal of “shorthand,” grammatical omissions, typos, and acronyms.  In an unstructured conversation, it is difficult enough to develop analytics meaningful to describe the “essence” of each conversation if, miraculously, everyone spelled every word properly, abbreviated nothing, used no shorthand or jargon.

At this point, one might be tempted to say “well, search engines do it all the time,” to which my response would be to throw the “bullshit” flag.  Search engines substitute speed and volume for meaning – if you don’t believe me, search for a topic – say “U.S. foreign policy in the Middle East” – and see what you find.  I suspect you’ll find first that Google can return millions of results in something less than a few tenths of a second, and you’ll be tempted to say this is good.  Then read through the links, and you will find that those actually dealing with the topic are some subset of the entire set of links returned from your search, and you can only determine which apply by looking at all of them.  Yes, that’s correct – all of them.  Further, even if you wade through thousands of pages of results separating the wheat from the chaffe, you will find conflicting information of varying quality – how do you decide the relative merit of one result contradicted by another?  How do you synthesize this information?

And yet, by listening in on millions of conversations, it is possible to find a wealth of incredibly useful information – much better information than has been available to businesses to date.  That information will span the breadth of customer interaction with a particular company – awareness and preference building, brand perception, pricing, quality, customer service and support, R&D, etc.  But to discover that information, a “semantic web” – a contextual framework for understanding the language relevant to a given business – is necessary.  I’ll give you an example from my past.

The beer industry is a tad bit more complex than Homer Simpson’s preference for Duff Beer in a can would suggest – for example, most people become aware of and develop a preference for their beer not from the liquor store, but from trying it at restaurants and clubs.  Some prefer bottled, some canned, and some draft.  Some prefer a type of beer – a lager, pilsner or stout; others a brand – Coors, Miller, Modelo.

So let’s assume that I am using Twitter to understand the conversation relevant to Coors beer.  It should be as simple as searching for “Coors,” right?  Well, even prior to Coors merging with Molson and then Miller, the answer was “not even close.”  To really understand the conversation, I would have to search for all of the information regarding the company and all of its brands: Coors, Coors Light, Blue Moon, Killian’s, Henry Weinhard’s, etc.  Some of that may have devolved into a common shorthand or synonymous phrases for certain products – “CL” might be Coors Light, as may “Silver Bullet.”  And some may misspell certain words – “Kilian” rather than “Killian’s.”

If that isn’t complex enough, you also need to have some way, from millions of comments, of evaluating the context of a comment – not just that someone is talking about Coors, but also HOW they are talking about Coors.  How do I evaluate on some scale the degree of positivity or negativity regarding the company or brands?  More difficult, how to I discern useful feedback that may help improve the product or extend the brand?  At some point, human beings will have to look at and synthesize these “deeper analytics,” but given the explosive growth of Twitter (and other social networking platforms), the amount of information will rapidly overwhelm the resources available to analyze raw data.

Having said that, let’s make the problem more complicated, by understanding that it is insufficient to simply understand what people are saying about your company and your brand(s).  You also have and should take advantage of the opportunity to understand what the Twitterverse is saying about your competitors – Budweiser, imports, micro-brews, etc.  So not only are you listening for product feedback, you are also listening for competitive intelligence, as a gauge of consumer sentiment as well as a source of R&D innovation.

As an extension of this model, which effectively becomes a learning model, are some measures of its accuracy and sensitivity to changes in the conversation – you have to know how well you are picking up trends, which are NOT captured reliably through keywords, hashtags, or other constructs, and ARE reflected in specific terms and phrases, and their juxtaposition to other specific terms and phrases.  If I see a comment that contains the word “Coors,” what have I learned?  If I see a comment that contain the words “Coors Light” and “Bud Light,” I know a bit more – at least what one Twit perceives as a competitive product to Coors Light.  If I see the juxtaposition of “love” and “Coors Light” (which can only be a complete hypothetical in my mind at least), and “Bud Light” juxtaposed with “sucks” in the same comment – now I’m getting more out of this conversation.

So it would seem a core of any analytical system to evaluate the content of this Twitter conversation is a “learning” semantic web that grows and evolves with feedback from analysis of previous information.  This sort of “fuzzy logic,” artificial intelligence and neural networks are theoretically possible; it seems clear when analyzing a completely unstructured conversation that these techniques will have to be applied to content analysis to provide an analytical basis for the content and tenor of the conversation relevant to your competitive landscape.

Part and parcel of such a system is not only that you can categorize comments, but also some measure of how reliably and accurately your analysis is, which would typically be expressed in terms of miscategorizations and false positives and negatives.  For example, if my analytics determine that a certain comment is a value statement and it is positive – i.e. Coors Light is perceived positively – then how certain can I be of that result?  How often is a comment properly categorized – such as a value statement rather than a feature request – and how certain can I be that a comment perceived as a positive statement is truly positive?  I may evaluate a comment as a positive value statement that may in fact be a feature request: “It would be great if Coors Light came in a self-cooling can.”

To date, “analytics,” which are rapidly proliferating on Twitter, have been based on counts, ratios and value judgments masked as “analysis.”  It may well be true that the ratio of tweets to retweets is a valid measure of something, but I don’t think anyone has done any extensive, longitudinal studies that would provide evidence of this fact – Twitter is simply too new and early in its evolution to have developed true science.  Clearly, if Twitter is to become a line item in a company’s expenses, it will need to be justified in some way: public relations, marketing, product development, whatever.  Reliable analytics to support this expenditures do not exist, and thus, “deeper analytics” must be developed to support the business case for using Twitter.

These analytics will measure the two primary drivers of Twitter’s value: Audience Size and Content Analysis.  The key components of Audience Size will be Targeted Reach (size of audience that corresponds to company’s target market) and Influence (to what degree does the company’s messaging drive quantifiable behavior – awareness, preference, selection, sale, follow on, referrals).  The components of content analysis, requiring a fairly extensive “learning” contextual framework for that analysis, are categorization of content (and reliability factors related to that categorization) and some measure of sentiment along some spectrum from very negative to very positive (again, incorporating some measure of accuracy and reliability).  The categories of comments should foot to some measure of value – awareness/recall, new customers, positive brand associations and preference, conversions, repeat sales, referral sales, new product features or brand extensions, etc.

These “deeper analytics,” however, while necessary, will never be sufficient to truly understand and influence market sentiment.  While these deeper analytics will allow for more effective and efficient analysis of an exploding conversation, they cannot replace trained minds – only help to distill thousands of comments down to dozens, and through proper categorization and measurement, help signal these analysts of emerging trends (not keywords or phrases!) that can be used to influence the success of the business.

The fundamental set of equations that drive every business are not going to change because of Twitter, but Twitter could become a remarkable source of competitive intelligence for companies that understand and invest in making it a key component of their strategic arsenal.  The successful companies that use Twitter to competitive advantage will commit to Twitter, will invest in Twitter and the analysts required to understand this flow of information (people like @danzarrella), and they will utilize “deeper analytics” aligned with strategic goals and objectives to which these analysts’ efforts can be employed.  There is nothing I have seen that comes anywhere near approximating the deep analytical capabilities I have sketched above, but I feel confident that somehow, some way these tools will evolve.  They must: the massive cocktail party that is Twitter holds too many veins of pure gold to be ignored for too long.

KB

Blog at WordPress.com.