Mining social media to boost segmentation

By Gregory Yankelovich | May 1, 2013

Reading time: 6 minutes

Abstract

This article describes how researchers can employ an algorithm to unsolicited customer-generated content from social media to gather the opinions of their users to improve products and target messaging.

Research Topics:: Consumer Research | Market Segmentation Studies | Online Research | Segmentation Studies | Social Media Research | Text Analytics
Industry/Market Focus:: Electronics | Internet/Web
Content Type: E-Newsletter Article

Share Print

Editor's note: Gregory Yankelovich is project management strategist at Amplified Analytics, a Richmond, Calif., research company. He can be reached at 415-742-2580 or at greg@amplifiedanalytics.com. This article appeared in the May 6, 2013, edition of Quirk's e-newsletter.

Marketers have used different market segmentation methods for a very long time. However, as our ability to collect and manage information continues to improve, newer methods of segmentation become available to enable more targeted marketing efforts for marketers and, in turn, better products and services for consumers.

One of the most commonly accepted strategies is a demographic segmentation based on the assumption that a specific group - based on age, gender, etc. - is the primary consumer of your product or service. Sometimes, this assumption is based on product purchase history. Regardless of the validity of an assumption, it does not often provide an insight on why this demographic segment selects the product in question or how they use it.

In other words, either a lot of guessing has to happen or additional segmentation strategies must be deployed. In my opinion, the popularity of a demographic strategy lies mostly in the low cost and the ease of access of the intelligence and information required, as behavioral and psychographic segmentation require a lot of research, which translates into high cost and time-to-market constraints.

Offer new opportunities

The advances in technology start to offer new opportunities for market segmentation based on an automated analysis of customer-generated content, which is becoming available with the proliferation of social media and the rise of the social consumer. Essentially, instead of assuming what demographic group would be the ideal target for our marketing efforts, we could look at a group that already expressed their interest by purchasing specific types of products or services and learn what elements of their experience were important to them.

Joel Rubinson, one of my favorite authorities in the field, posted this about Google+:

"As I review materials for the NYU social media class I am about to teach, I believe that Facebook will lead to the end of demographic targeting for media. Of course, content consumption and sharing behavior also enable this but Facebook will be the catalyst. Why not target on interests and actions? Thoughts?"

Most companies of any size use online survey techniques to engage their customers but the method does not support the discovery of the customer's perspective - it only validates assumptions based on questions posed and already deemed important by the company. Again, the primary driver of the survey method popularity is not the quality of the output and the ability to provide better market intelligence but the cost of implementation.

Better alternatives

There are better alternatives today, such as opinion mining technology, to gather unbiased market segment knowledge via unsolicited customer-generated content.

Here's an example to illustrate my point. Let's look at the market segment defined by a few popular products in the tablet category. (Non-like products that compete for the same share-of-wallet can also be used to get valuable insights.)

Apple iPad2 (666 customer "stories")
BlackBerry PlayBook (255)
HP TouchPad (650)
Motorola Xoom (576)
Samsung Galaxy Tab 10.1" (502)
Toshiba Thrive (433)

These products were selected based on their popularity, which manifested itself in a number of customer-generated content references available online in the form of customer reviews, forum comments or messages on the product page on various social networks.

The first level

The first level of customer intelligence gained by the opinion mining of this content is a list of customer experience attributes, sorted by their importance. Importance is measured as a percentage of the total number of unsolicited opinions expressed by the customers. This answers the question "What is important to the customers and how important is it?" (Table 1).

The next level

The next level allows us to measure the difference between: (1) customer expectations and (2) their actual experience. (We estimate the difference between expectations and experience algorithmically, based on the words and the expletive descriptors people use.) We then measure how well the customers' needs are met. We use a two-point scale to visualize that difference (0 = unacceptable, 1 = satisfactory [i.e., expectations are met by the experience], 2 = delighted).

For example, "I returned it back for a full refund..." would be scored a 0 - unacceptable. "The screen clarity is awesome..." would be scored a 1.8, while "I had to restart it few times before sound became strong enough..." would be scored 0.75.

Note that these measurements can easily be converted to any scale of choice without losing their meaning or accuracy. Table 2 focuses on the top five attributes of customer experience by their importance to illustrate this approach.

There are practical implications of these measurements as they reflect on marketing communications messaging, which has created customer expectations that a product needs to meet. In the tablet example, most of the products exceeded customer expectations (scores greater than 1.00) by a significant margin in attributes most important to them.

A higher level of intelligence

As an illustration, I would suggest that messaging regarding the usability of tablets could leverage customer sentiment to reassure the consumers who are hesitant to make a purchase and increase market adoption. That calls for a higher level of intelligence to provide an answer as to why customers feel this way and a context in which they express their opinions.

Below are sample user comments to illustrate the use of words and expressions (in italic) to describe their opinions and how they are attributed to a specific element of customer experience. In this case, that specific element is usability. These quoted words and expressions can be used to fortify marketing messaging:

"... responsive and intuitive. I've played around with Android and the...";
"...Good sound quality and easy to set up and is lightning fast...";
"...flawless integration. Intuitive navigation. Plays Flash with no problems...";
"...you use it to navigate the entire OS, really intuitive and surprisingly QUICK...";
"...surprisingly QUICK once you get the hang of it, and it's ridiculously easy to learn...";
"...apps that are useful and make it easier to use this tablet in a professional...";
"...browser is quick and easy to use. The BlackBerry Bridge is great for me...";
"...great UI - fun to use, intuitive, and utilizes everything you see...";
"...resolution brilliant, user interface reasonable intuitive so that one gets...";
"...touchscreen is very intuitive. After playing with it for a short span of time...";
"...operating system is intuitive. The glass keyboard (I'm typing this on)...";
"...pros: Easy to use Great screen resolution Love Bridge application...";
"...the whole unit is remarkably intuitive in terms of set up and use..."; and
"...is wonderful and extremely easy to set up. RIM thought of everything..."

Looking deeper, we see a lot of unhappiness when it comes to compatibility:

"...by pressing on the storage for those items but you can't view them...";
"...a couple that did not download, probably because they are not compatible with...";
"...a warning that says certain apps aren't compatible with this device...";
"...it is slow and sometimes the picture and audio are out of sync. On my Evo...";
"...should filter out any apps that are not compatible with the tablet device...";
"...the biggest disappointment is Samsung's own media hub is not compatible with...";
"...disappointed to discover that the Galaxy Tab 10.1 is NOT COMPATIBLE with..."; and
"...the disappointing realization that Skype IS NOT COMPATIBLE with this tablet..."

The flip side of the coin - an early understanding of the root causes of a customer's disappointment - can help alleviate larger problems, turn the situation around or even present an opportunity for differentiation. A deep analysis will provide invaluable context for taking advantage of the opportunity.

Difficult to ignore

To sum it up, this type of market intelligence can be produced within a few hours at the cost of a few hundred dollars without any installation, implementation or training investment. This makes it difficult to ignore as an alternative or an addition to the surveys-and-panels approach to market research. As the GPS technology taught us: The multiplicity of signal sources results in a better decision quality.