Stop counting followers, start listening to them

Read time 7min 00sec

Counting social media followers is a lot like sitting at a bar and counting the tops from the bottles of beer you've drunk. All this tells you is how many beers you've had; it tells you nothing about the quality of the conversation, how good the music was, or the general entertainment of the evening. It doesn't provide the overall context of the experience.

This was the analogy put forward by Robin Meisel, of RGT Smart, at the ITWeb Business Intelligence Summit yesterday.

According to Meisel, social media analysis is shifting away from mention counting and sentiment analysis towards understanding the importance of context, and this is achieved through linguistic analysis.

"Context is important because this is what executives are using to arrive at their decisions; this is where true BI comes into play," he said.

Social media analytics, said Meisel, is concerned with sentiment analysis (positive, negative, neutral), influencers, etc. "Effectively, it's about trying to gain insight out of measures, but very few people actually consider the context behind what is happening in the social media space."

Linguistic analysis, he said, uses insight to gain information on the drivers - what is behind the decisions that people are taking?

Meisel equated sentiment analysis to having a hammer. "If you have a hammer, every single problem looks like a nail." Linguistics analysis, he says, tries to beef up that toolbox, in a manner of speaking, to give businesses enhanced tools to better understand social media in order to make better decisions.

Decentralised vs holistic social media management

Meisel categorised businesses according to how they use social media - decentralised, centralised, central node, wheel and spoke, and holistic.

Most businesses (about 60%) fall into the decentralised category, he said. An early adopter within an organisation may "jump in, boots and all", resulting in everyone else backing off, believing it was problematic.

Businesses then decide to centralise the social media process, assigning the task to one department, such as marketing. In such scenarios, information is disseminated in a top-down, automated manner.

"What starts happening then is that other business units start getting involved, and they're saying: 'wait a minute, you don't understand my business, you don't understand how this works, I need input into this'. So it changes and we move into the wheel and spoke type of environment.

"Wheel and spoke then evolves further, where you've got the individuals within the various divisions saying: 'hang on, I've also got something to put into social media." He notes these are likely to be the Millenials, or the younger generation, for whom social media is second nature. Whereas, for example, the older generation may believe that a LinkedIn profile should be kept separate from a Facebook profile, these individuals have one personality across all profiles."

About 1% of businesses fall into the wheel and spoke category, he said, but what's missing is the data jungle, the need for social to become part of the DNA of the organisation.

"DNA is when you walk into your company and you have your social dashboards as you come in; you understand what is being said in social media; your organisation feels comfortable talking to the market, it's not relegated to one division."

The meaning behind the cat picture

"My Facebook timeline has a hell of a lot of cats on it," said Meisel. When analysing the memes people share, it's possible to "go into the bits and bytes" and determine the message structure and the meaning the particular meme conveys.

"But you're missing something. You're not getting the individual who's sharing that image's actual intent; you're getting the original creator of the message.

"But hang on a minute, the guys are using natural language when they share those images. Twitter is made up almost exclusively of natural language, so by mining into the natural language, we are able to gain insight into what people are actually thinking - the drivers."

Language analysis

Computers are poor at language, said Meisel; they rely on dictionaries. "Dictionaries are just a bunch of words with no personal meaning. Languages are constantly changing and it's difficult for dictionaries to keep up. The Oxford English Dictionary lists about 150 000 words in the English language. From a sentiment, true linguistic perspective, the words to which we can assign meaning from a sentiment analysis perspective is 975 - 0.6% of those listed in a dictionary. Computers also look at how many words are used, not how they are used, and computers don't understand sarcasm.

"Humans are just as poor at language," he said. "We need context to decide the meaning of words." Referring to a Harvard University study, he said humans only understand each other 70% of the time.

When language analysis is automated, a computer will assign the same meaning to a word each time it encounters it. He gives the example of Paris Hilton describing something as 'hot' in reference to something appealing as opposed to temperature hot. "The approach we're taking at the moment is completely broken."

How do we move away from this?

"If you are able to drill into the data and analyse it and get meaningful insight, you can better align your strategy. Once your strategy is more aligned, you can pull that into your marketing message; your marketing message becomes more aligned, the social space likes you more; they're aligned with your values and they'll start talking about you more and driving more data for you," he said.

Linguistics is not online reputation management (ORM), said Meisel. ORM involves sitting at the haystack looking for the needles, pulling them out one by one, and assigning them to the relevant department. "Linguistics analysis plugs on top of ORM. It's about taking an X-ray machine, putting it in front of the haystack, taking consistent snapshots in terms of what's happening and reporting on that." From that, one is able to tell how many needles relating to, for example, the finance department, is in the haystack and relaying that information back to the department.

ORM is a tactical approach, whereas linguistic analysis is strategic approach, he said.

More to language than words

When analysing social media, linguistics looks at function words and conjunctions (the, and, of, because, etc), which provide insight on the structure of the sentence. One can also incur demographics from this. Linguistics also looks at punctuation - the older and more educated people are, the better their punctuation is. Punctuation is also very important in terms of providing context. (Really? vs Really!), he said. Verbs, cognition, biological functions, motion, space and time, and phraseology also play a role in linguistic analysis.

The use of pronouns is telling, said Meisel. "The use of first-person pronouns implies the individual is taking ownership and is likely being truthful. Interestingly, 10% of Oscar Pistorius' affidavit has first-person pronouns in it, which indicates that he's taking ownership for what he is saying and that what he is saying is truthful."

He concluded by summarising the structure of linguistic analysis to include: summation and adding (followers, mentions), sentiment analysis, language and linguistics (to aid in understanding), psychological analysis and brand measures.

He again cautioned not to rely only on mentions, followers and sentiment analysis, as these lack the context that is critical to understanding meaning.

Have your say
Facebook icon
Youtube play icon