Thinking about social media monitoring

Thanks to the rise in influence of social media, a multitude of bloggers, fans and online communities now collectively determine trends and brand perception, driving sales day in, day out through their positions of influence. The uptake in online conversations, comments and reviews has been explosive and the importance of these conversations is growing by the day. Among many other things, people are discussing brands, describing their purchase intentions and asking for assistance in making buying decisions or product support.

Monitoring this activity has become a critical business issue. 

A range of services have recently emerged under the banner: Social Media Monitoring. Through these services brand managers are able to gain insights from the conversations people are having online every day and to make improvements to products, customer service and marketing as a result.

Two of the most widely recognised companies to offer this service are Radian 6 and Scout Labs. You can listen-in on the conversations of their customers, potential customers and other stakeholders in one place through your 'social media dashboard'. As far as presentation of data goes I was very impressed. It's certainly an attractive proposition for over worked brand managers!

Both of these providers make a big play on sentiment analysis which attributes a positive, negative or neutral score to each conversation. This is really useful in helping you determine the themes and topics that are driving both good and bad conversations about your brand, in addition to tracking the overall impact of marketing campaigns or news about your brand.

The obvious weakness to this is the lack of context and 'human' interaction in their analysis of each post, instead utilising a set of complex algorithms to determines the sentiment. I think this is dangerous and misleading.

Sentiment is more nuanced than a simply positive, neutral or negative — using an automated tool to assess how people feel puts too much faith in the today’s software. An ongoing human interaction and interpretation are essential to get real value — fundamentally I don’t believe that anyone will ever nail sentiment analysis as the human brain will always be smarter than any computer.

After digging around their websites for an hour or to and sampling the mood of the blogosphere, I concluded that beyond the corporate comfort of seeing numbers, graphs and sexy charts I doubt there is any real value in this service. 

"We don't live in a binarian world, in which people would be only pro-something or against-something: in opinion surveys, there are more grads, more details, more levels of understanding. Summarizing social media opinions into this pattern is not only wrong, but is a professional mistake: you give fake overviews, fake trends, fake insights, that can lead to marketing disasters. Moreover if you use "neutral" just to sort all the conversations that don't match your sentiment automatic filtering, you're just wrong! Neutral MEANS something, it does not mean "NOT UNDERSTOOD"

Thinking about the how to address this issue I went back to the basics of content analysis to better understand the issues. According to Dr. Klaus Krippendorff, six questions must be confidently addressed in a successful piece of content analysis:

1. Which data is analysed?
2. How are they defined?
3. What is the population from which they are drawn?
4. What is the context relative to which the data are analysed?
5. What are the boundaries of the analysis?
6. What is the target of the inferences?

Each question raises critical points for me about the quality of data and validity of insights from the a computational approach to understanding sentiment.

During this research that I came about an alternative method to Social Media Monitoring which takes a 'community-centred' approach. The critical difference is it uses a different data set and relies on humans to interpret the data — Reflecting on the Dr. Klaus Krippendorff's analysis about effective content analysis, there seemed more rigor and deeper evaluation. 

The community approach analyses the lines of sociality focusing on the link structure of the internet, intensity of peoples conversations, direct and indirect interactions, how people imitate each other and the context this takes places. Leading practitioners like Linkfluence, then evaluate this data using a series of metrics appropriate for social marketing. These include:

1. Influence: Shows the likelihood that a message published on a site will be broadcast.
2. Penetration rate: Measures the level of presence of a brand, a company or a product per community.
3. Share of voice among competitors: Measures a brand, company, or product’s presence in each community compared with that of other players in its industry.
4. Repetition: Measures a message’s exposure given its repetition rate.
5. Engagement: Measures the intensity and diversity of exchanges around a brand, company, or product.
6. Return on investment (ROI): Shows a message’s exposure and broadcast rate compared with the objectives set upstream and compares it to market standards.

As a result, this approach enables you to identify and rank the most influential blogs, fans and online communities that impact your business. Reading about there service I see you can create 'hit-lists' of key influencers, lead users and trend-setters relevant to your brand then evaluate every conversation, hyperlink and comment relevant to your brand, products, trends and competitors.

The sum total of this is a richer, more relevant set of measures and data, all collected, evaluated (by humans) and transformed into deep insights about the underlying issues, sentiments and influencers driving the conversations and how they evolve over time.

My conclusion: Take the time and go for a deeper, more time consuming approach as it ultimately yields more profound understanding and insight into your market.