ImpactWatch

With the month of May coming to a close, the new ImpactWatch demo about China and the 2008 Olympics has been fully operational for a couple of weeks now. The data collected already demonstrates the capabilities of ImpactWatch in real-world scenarios. Since May 6th, articles containing keywords relevant to China and the Olympic games have been aggregated and analyzed on a daily basis. Using the built-in tools of the ImpactWatch platform, I created this graph showing the average favorability (using a circulation metric) of both China and the 2008 Olympic games from 5/6 through 5/28 (today):

It’s no secret that China has had a rough time managing public-relations coming into the summer games because of media exposure related to questionable human rights practices, pollution, and the notorious Tibet fiasco. On May 12, about a week into our demo, the tragic earthquake in China’s Sichuan province shook the lives of thousands — as well as the focus of the American press, as the above graph illustrates. The average favorability of China skyrocketed on May 13, as media exposure shifted from criticizing the summer games host to covering the events surrounding one of the largest natural disasters of the year. As Economist.com has pointed out, amid enormous loss, China has largely gained positive press from this terrible tragedy. The results from ImpactWatch show that China’s favorability is holding relatively steady, at a higher level than just three weeks ago.

Here’s a breakdown of the articles that we’ve collected so far by issue:

As you can see, the earthquake and its effects is the most prevalent issue at the moment — expectedly so — and I don’t expect that to change anytime soon.

Over the coming months, the data contained within the demo will become better and better, which is the nature of aggregation and media analysis. Trends will become more long term and the quantities applied to the data will become more meaningful. In addition, I don’t have to go searching for these news clippings; one of the great things about ImpactWatch is that it automatically pulls articles based on keywords from various sources and feeds, but leaves the analytics to real, breathing humans — we’ve talked about why in previous blog posts. ImpactWatch tracks and streamlines the work/analysis without misstepping about the most important part, the actual reputation.

More China/Olympics updates to come — the summer is just getting started.

Times change, and so should ImpactWatch.  After becoming interested in all the China/Olympic drama while researching for this post on The Bivings Report, I decided that it was time for a new, more topical demo.  Replacing our current Real Estate Demo is a 2008 Olympics Demo.  This one continues to show off the great features of ImpactWatch, but the articles contained within it now concern anything and everything about the upcoming Games.

This new demo tracks media favorability for both the general Olympics as well as the country of China, specifically.  It also tracks which topics are covered in the articles surrounding these games, everything from Burma and Tibet to the worry over China’s air pollution.  OlympicsWatch does all of this while retaining the sleek and efficient design that has made ImpactWatch such a valuable resource for so many clients.  The analytics section of this new demo takes full advantage of our recently created graphing tool, so feel free to explore this very useful device as well.

To sign up for a weeklong trial of our new demo, click here.

Since 2004 – the year web 2.0 truly took the limelight – I have stood witness to the rapid evolution of the internet. An ever increasing portion of web content has come from end-users and consumers in various forms of text and digital media. This information is nothing that you, a user that navigated their way to this blog (blogs being of the first trends in user generated content), didn’t already know.

It is likely that you have probably been aware of this trend for a while – and heard about it repeatedly on blogs, in articles, and at conferences. It has taken shape into a vast, universally accessible archive of raw consumer feedback and profiling information, so it comes as no surprise that lots and lots of people are highly interested in finding ways to use this repository of dialogue in effective ways.

Competent analysis of a widely anonymous medium is not an easy agenda; there are many complications that arise when trying to arrange the clutter of user-generated web 2.0 content into coherent information that has real-world application. As Bill Tancer wrote in an article for TIME magazine last April about the demographics of web 2.0, it takes a little more than some tricky guesswork to put the content into a meaningful perspective.

In summary of his article, Tancer writes about how certain classical business principles are exhibited, and even highly exaggerated, in user-generated web content. The Pareto principle illustrates how 80 percent of results can be attributed to 20 percent of the causation. As a result, this trend adds to the distortion when trying to interpret web 2.0 data and attribute it to specific groups of consumers. He goes on to talk about other disproportionate and not-so-apparent trends in user generated sites such as YouTube and Flickr. In my research, other analytic sources support Tancer’s conclusions.

Tools like Tweet Scan are quick, inexpensive (can you say ad-supported?) and effective; lots of people are using them to find specific, raw feedback that was not possible before this wave of web technologies. It is thus important to be mindful of the limitations of said resources, and to frame their results within what is known. That’s when I think these tools become truly powerful.

Here at ImpactWatch, we have recently been conducting some closer examinations of social media. Through specific case studies, we have been conducting market research and developing new ways to aggregate and interpret social media. It was obvious to our team that social media – a medium based on actual human relationships (read: consumers) – would be invaluable to our clients who already rely on us to track their reputation in more traditional mediums.

There are countless variables and trends that I could speculate about that I’m sure, in time, some scholar of new media will base his or her dissertation and life teachings upon. Edward Bernays started a whole field of public relations and the scientific study of mass media psychology and consumer profiling, paving the way for audience identification and strategic policies. Since his time, these fields have come very far. Just as I’m sure that the current digital toolsets and comprehension therein are just a taste of things to come.

If you own a computer and have a disposable income, chances are good that you have bought something via Amazon.com, a well-known site dedicated to being the Internet’s largest store.  The site boasts many features, including discount prices, lists of recommendations for frequent users, and intuitive search features.  A past blog post on The Bivings Report highlights one of Amazon’s recent user-friendly upgrades.

An associate of mine recently praised Amazon because of the comments feature on the individual items’ pages.  At the bottom of a specific product’s page, after the official description and details are listed, comments from users/buyers are shown.  The comments deemed “Most Helpful” are the first ones that you will see, while on the right side of the page are the most recent ones.  Directly above the Most Helpful comments is a small graph of metrics, showing the number of votes, what were those votes, and the overall average of the votes.  Users can assign a numeric rating to the product-a ‘star’ system-on a scale from one to five.

Being the skeptic that I am, I questioned my friend’s praise of the comment system.  Typically, sites with comments vary wildly in usefulness.  Comments on websites can come in many forms: helpful, useless, curse word heavy, angry, naive, etc.

However, I was pleasantly surprised as I delved into the comment sections on Amazon.  Most of the posts were valuable, or at the very least, civil.  There were few, if no, drunken rants to be found.  I did notice that the overall rankings for the products seemed a little high.  I figured that it may be just a coincidence, but I decided to do a quick analysis to comfort me.

After selecting a product category that had many similar, yet different products (TV show box sets on DVD), I went through 100 product pages.  On each, I noted the number of 1 star, 2 star, 3 star, 4 star, and 5 star reviews, as well as the overall average of the reviews.  I only rated one randomly-selected season per TV show.  The complete data set can be found in a spreadsheet below.

The first detail of interest is the fact that a vast majority of votes were for 5 stars.  In fact, the number of 5-star reviews is nearly triple the amount of all the other scores put together.  The results are summarized below in a pie chart, made using ImpactWatch:

Why so many 5-star reviews?  I came up with many conclusions as to why users dominate the clickable 5-star vote button:

• Users typically rate products that they would buy or enjoy already, instead of ones that they hate and to which they’d give lower scores
• Users are more generous online than they would be offline
• TV show box sets are rated by quality of the show, and not the quality of the actual DVD, leading to higher scores
• Users think little about the rating; a 5-star rating essentially being “I liked it!”

Of course, this tendency to give out high ratings led to the majority of products being rated an average of 4.5 or 5.  Results are summarized below:

The overall average of the averages is 4.545.  I suppose that I should start saving my money, because the Amazon community suggests that I purchase every DVD box set available. 

Perhaps a different product selection would have yielded different results, but I doubt it.  While Amazon is a great online resource, its comments feature seems to have fallen prey to fanboy-ism, resulting in ratings skewed higher than they should be.

While it would not be simple to overhaul the entire site, my suggestion to combat this problem would be to expand the ratings feature, just by increasing the available ranking choices from 5 to 10.  I have found on other websites that ten-point scales lend themselves to a greater variety of ratings.  Amazon probably does not want to change a thing though, because higher ratings may lead to more sales.  Was this rating system a clever marketing idea from the start, or did the company luck out?  I guess that’s for only the CEO to know.

[This blog post has been rated 6 stars.]

Amazon Comments Data Spreadsheet 

PDF of Amazon Comments Graphs

Dell, who was once an example of how not to embrace social media, has done a tremendous job at stemming the tide of the negative conversations about them on the Internet. Other than knowing that Dell was listening and participating in the conversations with success, I had not heard the story of what they had specifically done to achieve all that success. That was until a couple weeks ago at the SNCR New Communications Forum in Santa Rosa California, where Richard Binhammer of Dell joined John Cass of the SNCR for a Keynote Conversation.

Through Dell’s blog searches and participation in numerous other social media outlets, Dell discovers about 4000 posts a day in all languages. They handle that load by doing triage on the posts, deciding which ones urgently require a reply, which need to be watched, and the ones that don’t need a reply. Out of the 4000, about 200 are addressed by the tech support team that communicates with new media and about 100 have to do with corporate brand and image.

At times it can seem overwhelming for companies to consider tracking new media and social networking. In the past I’ve suggesting to start by at least listening. This is exactly what Dell did. Also, if you consider that a huge company like Dell can narrow 4000 posts a day to only 300 that they really need to focus on should give hope to anybody trying to raise their awareness of social media conversations.

The results are certainly impressive. Dell has been able to lower the number of negative posts about them from 49% to 21%. Richard suggested that may be the best they can do, however. In the past 8 months that number has not gotten any lower. It just seems that no matter what they do, 20% of the people out there just aren’t going to like Dell. An audience member suggested that this might be a great research topic for the SNCR. Does 20% equal success or are there ways that the needle can be moved even further?

One thing that Richard noted as important to the success of this initiative was that they had corporate buy-in. Michael Dell basically directed that he wanted this to happen. Sometimes in a smaller or less technical environment it can be hard to get this kind of managerial support, but it seems to be a battle worth fighting. They also rarely involve the legal department, and as such, it was critical to get people with good judgment involved.

Two other comments were particularly interesting to me. First, they pay no particular regard to influence of who they pay attention to. The program is truly about communicating with their consumers. Dell has learned first hand how a few blog posts can snowball. Second, from their interaction with social networks, Dell is aware of emerging issues and concerns two to three weeks in advance of getting a call about it from a main stream media outlet. I have no doubt that kind of awareness helps their overall PR efforts.

One of the main features of ImpactWatch is the ability to track how positively or negatively your company or product is being portrayed in online media.  You can also track this portrayal for individual products or your competitors and their products.  In addition, all of this information can be summarized in a neat little graph under the “analytics” section.But how does the positive and negative system work?  After all, can a computer actually scan an article for words and key phrases and decide whether or not it’s overly positive or negative?  Possibly, but living in a world dominated by sarcasm and metaphors, I have my doubts that the results would be altogether accurate.  This is especially true for blog entries since they tend to be voiced by opinionated authors writing specifically to be snarky and opinionated.  An automated (i.e. non-human) sentiment rating system would certainly not work for those.  Only a human being can detect the subtleness of sardonic wit.

No, the answer to the debate between positivism and negativism is a human rater.  This rater reviews the article and marks which way it leans, as well as which indicators and specifications were reviewed within the article.  This person reads many articles in the same category and knows, from comparison, which ones are positive and which ones are negative.  Doing this greatly decreases the degree of subjectivity.  In addition, articles can be related as “neutral”.  This designation is saved for those articles that mention the company or product but do not necessarily serve to review it.  We can even delve further into these ratings by using a five-point scale (from ‘very positive’ to ‘very negative’), which I actually prefer, but then again, I’m a sucker for details.