Knowledgeistics

Archive for the ‘Alex’ Category

Quantitative metrics are key to understanding one’s market. Now, more than ever, marketers have access to a wealth of data to help make decisions. However, I’m going to argue that this new resource is often emphasized at the expense of qualitative knowledge, which is instrumental to creating effective strategy.

Today in the New York Times, they covered the efforts of big firms like M.R.I. to better quantify the impact of advertisements in magazines. A small detail stuck out as I read. Ad revenues at magazines are not falling as fast as they are at newspapers. One obvious explanation for this is that magazines come with built-in psychographic and demographic targeting. Want to reach snowboarders? Why not try Transworld Snowboarding?

The underlying message of the article was also pretty clear. Quantitative metrics are indispensable for marketers and advertisers. Online advertising and marketing offer quantitative analysis undreamed-of before the advent of the Internet. Print advertising offers metrics that have resisted improvement for decades- ergo, print advertising loses.

At Compete, Stephen DiMarco underscores this point by drawing a line between marketing undertaken on the basis of intuition, “Powerball Marketing” and marketing informed by substantial statistical analysis, “Moneyball Marketing”. His key points, roughly paraphrased, are that:

1) Desired outcomes of marketing campaigns should be specifically quantified

2) Advanced quantitative measures should primarily guide marketing initiatives

On point 1, I agree with Stephen wholeheartedly. All projects need clear-cut objectives, and woe to the firm that allocates hard money to soft, unquantified “improvements” or “increases”.

On point 2, I have to disagree, in part. Success in marketing-by-numbers can’t explain the persistence of magazine ad revenues over newspapers’, given that both media lack robust quantitative measures. Online marketing and market research provides a veritable fire hose of consumer data to the savvy marketer. This leads to a tendency to ignore or discount qualitative knowledge, despite its necessity, as demonstrated by those plucky glossies.

Quantitative data can accurately describe consumer behavior, and be used to predict it to some extent. Qualitative data is valuable when the quantitative data you have doesn’t provide the ability to rationally explain a behavior, or make rational strategic decisions. When the beliefs, attitudes and motivations of one’s customer or consumer are well-understood, it’s possible to make good decisions when quant data can’t do it for you.

For example, if your analytics show that conversion rates have fallen off sharply after introducing a new tagline, the most the numbers can do is identify that the tagline is the problem. With a robust qualitative dataset, and a good understanding of the beliefs, motivations, and other relevant pyschographic factors expressed in that dataset, you’ll be able to identify the unintentional use of a new slang word used to describe terrible body odor in the tagline. The quantitative data would allow you to replace the tagline and fix the problem. The qualitative data would allow you to replace the tagline, as well as spin the faux-pas, make a knowing joke with your consumer, and salvage brand image.

Part of a marketer’s job is to understand their customer or consumer. Clickstream data, correlates of conversion rates, or even surveys won’t really illuminate human beliefs, attitudes, or motivations. SMMR provides a unique opportunity to collect, oftentimes, both quantitative and qualitative knowledge simultaneously. It’s important for marketers to remember that both types of information are essential. Remember: Pizza and beer separately are great, but together they make a balanced meal.

We all know that social media are a great source of market insights. All you have to do is log on, see what people are saying about your brand or product, and act accordingly. When you pay close attention to your customer, you can’t fail. Right?

Wrong. SMMR is like any other research activity – it demands not just data-gathering, but the knowledge to interpret that data correctly. If you take the time to understand your brand, product, and target market well enough, SMMR will return great ROI. On the other hand, if you simply take online commentary at face value, you’re playing Marco Polo in an empty pool.

Take the example of the 2006 film Snakes on a Plane. Early teasers generated a great deal of buzz in social media, with commentators practically salivating over the expected camp extravaganza. The studio went so far as to re-shoot portions of the film to meet (some might say pander to) fan expectations. Despite the huge buzz and apparent addressing of needs uncovered through (crude) SMMR, the film grossed a disappointing $62MM.

Buzz aside, SoaP was probably never going to be a truly great film, in terms of acclaim or revenue. However, it is clear that the studio was badly misled by internet buzz. A better understanding of the dynamics of online conversations and their target market would have prevented this misapprehension.

In contrast, the site Think Geek takes a very sane and profitable approach to integrating online conversations into their product and marketing strategy. Although (or maybe because) Think Geek is a niche retailer, they have exhibited a much better understanding of their target market.

Each year, they feature April fool’s products which are meant only as satire of the merchandise they actually carry. However, sometimes these fictional products strike a chord with their customer base and spark a lot of conversation and outright demand. One year, the “Personal Soundtrack Shirt” created such an outcry and later saw production:

April Fool’s Joke Turned Real
Yep. This unusual shirt was originally a joke product for April Fool’s day. But after your overwhelming positive response and hundreds of e-mails screaming to “make the damn shirt already” we went ahead and made the damn shirt… please enjoy.

While this might seem like an obvious choice – ‘give the people what they want’, it required careful consideration on the part of TG. First, they had to find a way to produce the shirt in appropriate quantitites at a reasonable price. Just as importantly, they had to use knowledge of their customer base to determine what that price could be. They had to estimate propensity to buy, rather than simply exclaim about the desirability of the shirt. They had to whether there was genuine demand, or simply a vocal minority, as in the case of SoaP. TG considered all of these variables and introduced a successful product.

Moral of the story: If you can’t tell the difference between real demand from your target market and loud noises from an unruly mob of bloggers, step away from the keyboard, take a deep breath, and look a bit closer.

On May 18, the much-hyped new search (er, computational) engine Wolfram Alpha launched. My first impression is that it’s a great tool, but is nearly useless for market research of any sort.

That’s a pretty bold statement, but the key here is that Wolfram Alpha is emphatically NOT a search engine as we’re accustomed to. WA refers to itself as a Computational Knowledge Engine, which is actually pretty accurate. While that distinction gives it some exciting new capabilities, it also means that it won’t be replacing your favorite search engine anytime soon.

Wolfram Alpha is useful for doing things like pulling a quick poop sheet on a geography or population, or doing some relatively sophisticated calculation, correlation, and graphing of quantitative data. What it’s not good for is learning about any topic not already coded into the WA system. For example, while it can easily compare the nutritional content of Hot Dogs and Hamburgers, it chokes if you wander even a short way off the beaten path, having no idea what Deep Dish Pizza or Philly Cheese Steaks are.

The implications for market research, then, are essentially nil. While WA might be a good first stop to do groundwork on demographics or basic statistics, market research is not a computational exercise.

Wolfram Alpha is an awesome technology. The promise of Wolfram’s ambitious goals and substantial progress is exciting. Being able to compute basically anything without special software or tools would be of huge value. Unfortunately, WA is not going to save me much work when I need to learn qualitative facts about a market, company, product or brand. Now, Wolfram Gamma or Delta? I wouldn’t count them out.

Continued from http://www.knowledgeistics.com/2009/05/how-big-of-a-problem-is-astro-turfing/

The Bad News:

Information sources sometimes have a conflict of interest
Sites that feature reviews or consumer information are known to accept payment for manipulation of supposedly “user-generated” content, or allow third parties to do so. Heavyweights like Amazon, Yelp, and Ripoff Report , among others, have been implicated as such. These outlets have an economic incentive to manipulate the content, as well as prevent users from detecting the manipulation.

Conclusive detection of astro-turf can be hard or impossible
Supposing an information outlet would only allow the posting of positive user reviews after payment, proof would be hard to come by, and the content itself would be genuine, if not representative. Further, without serious technological sleuthing, isolated and credible instances of astro-turfing may be hard or impossible to detect. When the total set of reviews is small or non-existent, one convincingly written advertisement-cum-genuine testimonial can effectively stuff the ballot box. In cases such as these, without telltale signs of astro-turf, the misrepresentation may pass.

Nothing is stopping the astro-turfers
An unsuccessful astro-turfing attempt often results in a cascade of bad publicity. However, these instances are relatively rare. Furthermore, a sincere-seeming apology to users can sometimes make “better to ask forgiveness than permission” an apt motto. While astro-turfing may violate the TOS of some sites, and be removed when detected, this is by no means a sure thing. Most likely, an astro-turfer won’t run across any resistance, and in the unlikely event that they do, they may be able to salvage their reputation regardless. Unfortunately for those interested in unbiased information, the costs are low and the benefits potentially large for astro-turfing.

So, is astro-turf a real problem? It certainly complicates things, and requires a careful researcher to take all “user generated data” with a grain of salt. By injecting bias into the information, marketers remove value for users and make researchers’ jobs harder. That said, it doesn’t warrant full-scale panic or despair. The bias of astro-turf is unlikely to generally subsume true consumer opinions, by virtue of sheer numbers, combined with careful detection of fishy information.

While it would certainly be desirable to do away with astro-turf, expanded legislation is unlikely to produce effective enforcement, consumer pressure is unlikely to increase the cost of dishonesty to the point where it is not a viable policy, and consolidation of media into electronic formats will only increase the potential value of the practice. While watchdog groups like the BBB and Consumerist.org are able to provide a check on bad behavior, caveat emptor remains good advice.

Depending on your point of view, astro-turfing is a great marketing tool, or a pernicious and nefarious tool for deception. “Astro-turfing” is the practice of ‘building artificial grassroots’, and is commonly used in a political context, but also applies to similar commercial activity. For this post, I’m going to focus on astro turfing done for marketing or sales purposes, not political or reputational. Astro-turf is different from plain old spam in that it is advertising deliberately misrepresented as genuine user-generated content.

Astro-turf poses a uniquely sticky problem for both consumers and market researchers. Consumers can find themselves stymied when considering a purchase if they believe that user reviews within a given category are untrustworthy. Researchers will be uncertain about the validity of their conclusions if their data is compromised by covert marketing.

The appeal of astro-turfing to marketers is obvious. Because of the semi-anonymous nature of online commentary, marketers are able to pass off what amounts to advertising as the genuinely expressed, unbiased opinions of users. While it may be effective, the practice of fudging reviews and commentary is clearly unethical. Unfortunately, it is also seriously widespread – it has been suggested that trusted sources of user reviews like Amazon and Yelp, [2] not only fail to prevent astro-turfing, but may condone or even encourage it.

When you consider the kind of content that gets slyly slipped into wikipedia, the covert ad campaigns launched on Youtube, and the company flacks surreptitiously patrolling various user forums, it starts to seem like astro-turf and its variations are unavoidable, and maybe insurmountable. But how big of a problem is it really? Looking at it from the consumer and researcher’s point of view simultaneously, here are some factors to consider:

The Good News:

The marketer’s reach may exceed their grasp.

A diligent researcher or consumer should be able to get a reasonable idea about a product or service by examining a broad swath of sources. Will marketers astro-turf Amazon? Sure – high traffic and visibility mean good return on astro-turf investment. The costs outpace the benefits when you try to infiltrate every small-time forum or message board that features a discussion of the product. Not every consumer will put in the time to research those small, low-potential sources, but the dedicated should be able to spot discrepancies.

Users in niche markets actively defend their information sources.

In message boards and forums, expert users often have a nose for the plastic scent of astro-turf, and are eager to sound the alarm if they catch a whiff. Marketers would face high time costs in building a reputation and relationships with forum users, only to risk it when they need to advance a suspect opinion. Instead of taking that approach, in some niche communities companies will go above-the-table and assign support staff as a liaison to the community to openly share information with users.

It’s easy to spot a rat.

Unsophisticated astro-turfers will often devote nearly 100% of their reviews or comments to building up their product or brand, and sometimes to tearing down competitors’. A quick review of their posting history should reveal this phenomenon pretty quickly.

If it’s too good to be true, it probably is.

In some cases, an astro-turfed review may have the ring of ad copy rather than extemporaneous opinion. When a “review” seems eerily reminiscent of a brand’s key positioning statement, or lavishly praise every aspect of a company – including unrelated products or services – it might be a plant.

Technological traps catch the lazy.

If marketers aren’t careful to make every aspect of their astro-turf seem natural, a tech-savvy investigator can catch them. For example, if a post or review has been partially re-used even once, Google should quickly turn up the offense. Multiple hits for a sentence or part of a paragraph in topically co-relevant, but disparate sites will reveal astro-turfing, even if perpetrated under different user names. Similarly, if user IP information is available, it’s possible to reveal whether multiple posts or reviews have come from one source. By examining the numbers in their IP addresses, commonalities can be spotted and astro-turfers outed.

To Be continued later this week, with … The Bad News.

You may have noticed the soft launch of Hunch.com, a site that “helps you make decisions and gets smarter the more you use it.”Hunch is very intriguing, and not just from a SMM standpoint.  While it’s still in pseudo-beta phase (the creator eschews the term “beta” but acknowledges some rough edges), the site is fully functional.  I gave it a spin and noticed a few interesting things.

Hunch delivers value by giving advice on a range of topics.  It claims to deliver accurate answers by creating correlations between answers to simple questions.  An example they use is a supposed Republican/Democrat, Fiji/Evian correlation.   A new user is asked to answer at least 10 questions of the type “Do you use a Mac or a PC”,  “Do you think global warming is a seroius threat”, “Are you employed”, and so on.

From the Hunch site, about how their algorithms become “smart”:

Hunch is powered by collective user knowledge, decision topics mature over time. Newly submitted topics often won’t be very smart at first, but as more and more people train and refine them, the topics will get much smarter. Second, Hunch’s decision outcomes will become increasingly customized for you the more Hunch gets to know you. How does that happen? By your trying many topics and also answering the ‘Teach Hunch About You’ questions which appear on the top right of the homepage.

Once initiated, you can dive in and query this web 2.0 magic 8-ball.  Examples of available topics include serious quandaries like “Am I Adopted?” and less weighty ones like “Should I go to Bed?” (My answer – yes.)

It’s an interesting site for a typical user, but it should be even more interesting to anyone involved in market research or SMM.  With proper analysis, the correlations and answers in the Hunch database could deliver extremely precise demographic and psychographic insights. Hunch claims to make all their money from referrals – basically affiliate links stemming from “What sort of Item X should I buy?”  They also claim that they don’t sell user data:

“Will Hunch share my individual answers or user data with other companies? No. That’s simply not the business we’re in.”

Well, that’s pretty clear.  But it doesn’t preclude using the data for targeting, branding, or other marekting purposes – not necessarily.   My interest was piqued by one of the introductory training questions: “Do you typically pay more or less than $7 for shampoo?”  If you compare the correlates of my answer with the consumer profile of any given shampoo brand, you’ve suddenly produced an immediately valuable (and marketable) bit of information.

Hunch seems to have put together a highly efficent system for correlating psychographic, economic, geographic, and demographic data.  In my opinion, however, Hunch has an edge mostly because they’ve come up with a way to collect a great deal of consumer data on the cheap.  The real breakthrough is that they’ve made it fun (and social) for consumers to turn over their data en masse.

While their user base currently seems to be quite skewed (currently, 41% of their users own an iPhone, 66% identify as liberal, and the great majority live in urban areas,) their data on each user is so robust that this type of bias should present an unusually small obstacle.   Hunch is a good system for generating referral sales, but I expect that it might be a truly great system for generating market insights.  It will be interesting to see how their team utilizes this tool in the next 12-18 months.  Furthermore, it will be interesting to see how marketers and researchers find ways to leverage the site from the user end.  Twitter provides a level playing field and delivers value to businesses and consumers alike – will Hunch allow emergent uses for business to develop?