An independent study reference written by Dr Phuc V. Nguyen. It is not official subject material — for assessment requirements always follow your subject outline and vUWS.
Social media analytics
Social media analytics turns public online conversation into evidence a business can act on. It joins three jobs. Listening asks what is being said about a brand, a competitor or an issue. Measuring asks how far content travels and who engages with it. Acting feeds those findings into service recovery, product changes and campaign decisions. The raw material is unstructured text, images and interaction records rather than tidy rows, so volume is cheap to collect while representativeness and intent stay hard to establish. That gap is where most weak social reporting is born.
Why it matters
Picture standing in a very large, very noisy shopping centre holding a notebook. You can hear everything, but not all at once, and the loudest voices are not the average shopper. Social media analytics is the discipline of taking notes in that room. You decide what to count, whose voice counts, and what a sudden burst of noise actually means, before anyone writes it into a report for the executive team.
A brand dashboard shows mentions up 40 per cent against the prior period while average sentiment is unchanged. What is the most defensible first move?
Formulas
Worked examples
A grocery chain sees brand mentions jump from about 400 a day to 3,100 in one afternoon. The marketing team wants to write up a viral win.
Decompose the spike before celebrating. Split the 3,100 mentions by sentiment, by originating account, and by whether each item is an original post or a reshare. Here 2,600 are reshares of one customer video about a mislabelled allergen, and most of the reach comes from six large accounts. Volume is up and so is risk. The right response is product and service recovery plus one clear statement, not a campaign. The analytics job was to turn a single number into three: what was said, by whom, and how it spread.
Two campaigns are compared. Campaign A collected 9,000 likes, campaign B collected 2,400. Leadership wants to repeat A.
Likes are a count, not a rate. Campaign A reached 600,000 accounts through paid promotion, an engagement rate near 1.5 per cent. Campaign B reached 40,000 organically, about 6 per cent. B drew far more response per account exposed and cost less. A is not automatically the weaker campaign, because reach has value of its own for awareness, but any comparison has to fix a denominator and state the objective first. Ranking raw counts across different reach levels simply rewards whoever spent the most money.
Common mistakes
- ✗Rising mention volume is good news. Volume has no direction. A spike can be a recall, an outage or a joke at the brand expense, so it must be read together with sentiment, source, and whether the posts are original or reshares.
- ✗Social listening shows what your customers think. It shows what the small, self-selected group who post in public choose to say. Most customers never post, and heavy posters differ systematically from light ones, so listening produces hypotheses to test rather than population estimates.
- ✗Engagement rate is comparable across platforms. Each platform counts reach, impressions and interactions differently, and a reshare on one is not the same act as a save on another. Comparisons only mean something within one platform, one post format and one measurement window.
- ✗A sudden change in the numbers reflects a change in the market. Platforms alter their ranking, their sampling and their reporting definitions without notice, and automated accounts inflate counts. Rule out measurement causes before accepting a behavioural explanation.
Revision bullets
- •Three jobs: listening (what is said), measuring (how far it travels), acting (what changes)
- •Raw material is unstructured text and interaction records, not tidy rows
- •Engagement rate needs a stated denominator: reach, impressions or followers
- •Share of voice is relative, so it moves when competitors move
- •Volume alone carries no direction; pair it with sentiment and source
- •Posters are a self-selected minority, so listening yields leads, not population estimates
Quick check
A brand dashboard shows mentions up 40 per cent against the prior period while average sentiment is unchanged. What is the most defensible first move?
Campaign A: 9,000 likes from a reach of 600,000. Campaign B: 2,400 likes from a reach of 40,000. Which statement is best supported?
Connected topics
More in Text and Social Data
Sources
- Kaplan, A. M., & Haenlein, M. "Users of the world, unite! The challenges and opportunities of Social Media." Business Horizons, 53(1), 59-68, 2010.Widely used working definition of social media and of user-generated content as a data source.
- Ruths & Pfeffer (2014)Ruths, D., & Pfeffer, J. "Social media for large studies of behavior." Science, 346(6213), 1063-1064, 2014.Sets out why platform populations and platform sampling make behavioural claims from social data fragile.