Course
L10 · Listening to the Audience
Unit C
Digital Audience · MSJ2252
10

Listening to the Audience

Analytics dashboards tell you what audiences do. Social listening tells you what they say.

Unit C: How do we research and monitor?

A newsroom editor refreshes a live traffic board every few minutes. Is she listening to her audience, or watching a number someone else designed to shape her decisions?

Today we move from counting the audience to listening to what it says and does, and we ask what that listening costs.

By the end of today

What you will walk out knowing

01
Read a dashboard critically
Identify what each analytics metric actually measures, what it obscures, and who it was designed to serve.
Reveal
02
Distinguish behavioural from conversational data
Behavioural data tells you what audiences do. Conversational data tells you what they say. Neither is complete without the other.
Reveal
03
Explain social listening
Social listening is the practice of scraping and classifying public online conversation to infer audience sentiment and emerging themes.
Reveal
04
Recognise the clickbait tension
Analytics create pressure to chase metrics. Understanding this tension is the first step toward critically evaluating how organisations respond to it.
Reveal
The analytic shift

From counting to listening

Before digital

Retrospective counting

Audience research happened in retrospect: circulation figures, television ratings diaries, quarterly reader surveys.

  • Print circulation audited monthly
  • TV ratings from sample diaries
  • Feedback loop: weeks or months
  • Data primarily for advertisers and managers
Digital era

Real-time listening

Platforms made audience data continuous, granular, and instantly available to editorial teams as well as business sides.

  • Clicks, scroll depth, and time-on-page live
  • Social comments scraped continuously
  • Feedback loop: minutes or seconds
  • Data flows directly into editorial decisions
Carlson (2018)

Measurable journalism

Platform case · Carlson ethnography

Chartbeat and the habit-forming dashboard

The lesson: a tool designed to help can also discipline. The dashboard makes the editor's choices visible to themselves, without anyone issuing a directive.

Core distinction

Behavioural vs conversational data

Behavioural data

"What they do"

Captured automatically from platform actions. Precise and continuous, but silent on motivation.

  • Clicks, page views, scroll depth
  • Time on page, bounce rate
  • Referral source: Facebook, search, direct
  • Device type, session frequency
Conversational data

"What they say"

Scraped from public posts and comments. Rich with meaning but harder to quantify and classify.

  • Comments, reactions, shares
  • Public posts and hashtags
  • Sentiment: positive, negative, neutral
  • Themes, keywords, emerging concerns
Social listening

How platforms scrape the conversation

Social listening tools aggregate public posts and classify them algorithmically. The audience speaks; the listener is invisible.

In Bangladesh: Facebook comment sections on Prothom Alo posts are the most accessible social listening surface for local editorial teams.

Sector tour

Who listens, and how

Every sector practises audience listening. The tools and metrics differ; the underlying logic is the same.

01
News: the traffic board
Prothom Alo and The Daily Star monitor Facebook reach and engagement per post in real time. Stories that land on the algorithm get immediate recirculation. Those that do not disappear within hours regardless of editorial quality.
Reveal
02
OTT: retention graphs
Hoichoi and Chorki track where in an episode viewers drop off. A sharp exit at minute 12 of a drama tells editors the cold open was too slow. The listening is visual, not verbal.
Reveal
03
Music: skip and save rates
Bangla artists on Spotify and YouTube track save rate, skip rate, and completion. A track skipped within 30 seconds earns no royalty. Listening data directly governs creative decisions.
Reveal
04
Creators: YouTube Studio
Bangladeshi street food creators check average view duration and click-through rate on thumbnails. The dashboard shows exactly which 10-second segment caused half the audience to leave.
Reveal
05
Games: session and spend data
PUBG Mobile and Free Fire operators monitor daily active users, session length, and in-game spending signals. A spike in early session drop-off triggers content or difficulty adjustments within days.
Reveal
Critical perspective

The resistance response

Analytics adoption in newsrooms was never smooth. Journalists built a counter-discourse to protect their professional autonomy.

01
The alignment argument
Chartbeat's pitch: engagement metrics reward good journalism. If readers return, your calculations align with quality. The data and the editorial judgment point the same way.
Reveal
02
The resistance argument
Critics asked: what if Watergate had been run through Chartbeat? Traffic to corruption investigations would have told you to stop. Popularity is not the same as public importance.
Reveal
03
The discipline effect
When the dashboard is always visible, self-monitoring replaces managerial coercion. Workers regulate themselves against the number. The tool extracts productivity while feeling like personal choice.
Reveal
04
The clickbait risk
Optimising for engagement creates structural pressure toward sensationalism and outrage. Not because anyone decides to lower standards, but because the metrics reward it incrementally.
Reveal
Click to decode
Click to decode

Annotate the dashboard

Each tile is a real metric from a newsroom analytics dashboard. Click to find out what it truly measures and what it conceals.

Concurrent visitors
4,832
right now
Engaged Time
2:14
avg per article
Recirculation
18%
read a second page
Bounce Rate
67%
left after one page
Referral: Facebook
58%
of all traffic
Loyal readers
9%
visit 5+ days/week
Branching scenario · Night editor
Branching scenario

What do you promote?

It is 10 PM. The Prothom Alo traffic board shows two competing stories. Story A: a minister resigns. Engaged Time 1:20, shares 340. Story B: an investigation into factory safety violations in Gazipur. Engaged Time 4:10, shares 18. Your editor says: "Give me more of what the data says people want."
Outcome A. Traffic spikes tonight. By morning, three outlets have run the same political story. Your investigation sits unread. A factory fire occurs two weeks later. A rival outlet publishes the safety story first. Readers ask why Prothom Alo did not warn them.
Outcome B. The safety story earns a slow burn of shares over 72 hours. Three NGOs cite it. It reaches a parliamentary committee. Traffic was lower that night, but engaged time was the highest of the week. The editor-in-chief sends a note: "This is what we are for."
Outcome C. A pragmatic middle path. The political story draws the audience; the editorial note on the investigation signals values. Neither story is sacrificed, but the hierarchy still had to be decided.
Sorting game
Sorting game

Classify the comment

Select a comment from the left, then click the correct sentiment column. These are the kinds of posts a social listening scan on a Bangladeshi news page would surface.

Comments to classify
Stance spectrum
Stance spectrum

Optimising for engagement serves the audience.

Strongly agreeStrongly disagree
Drag the slider to take a position
Knowledge check
Knowledge check

A newsroom dashboard shows a story with 4 minutes Engaged Time and only 18 shares. What does this combination most likely indicate?

Social listening tools scrape public online comments and assign sentiment scores algorithmically, without users knowing their posts are being analysed.True. Services like Radian6 and Crimson Hexagon use web scraping and language processing to classify social media posts at scale. The analysis is invisible to the people who wrote the comments.
For Bangladeshi publishers, Facebook referral traffic is a minor source of digital visitors compared with direct and search traffic.False. For most Bangladeshi news publishers, Facebook is the dominant traffic referral source, often accounting for well over 50 percent of digital visitors.
Your turn · 15–20 min

Activity: Listen In

Step 1

Choose a subject

Pick any Bangladeshi organisation: a news outlet, an OTT platform, a YouTube creator, or a brand. Go to their most active public platform, likely Facebook or YouTube.

Step 2

Classify 15 comments

Read the 15 most recent public comments on one post or video. Write a keyword summary for each in the positive, neutral, or negative column of the worksheet.

Step 3

Name the insight

Identify one thing this comment section reveals that their behavioural analytics alone could not have told them. That gap is your case-study building block.

Today in five lines

Recap

01Measurable journalism is simultaneously a material, organisational, practice, professional, and cultural shift in how newsrooms work.
02Behavioural data tells you what audiences do. Conversational data tells you what they say. Neither is complete without the other.
03Social listening is the practice of scraping and classifying public online conversation, but the methodological gaps and surveillance dimensions are significant.
04Chartbeat's dashboard creates self-monitoring without coercion: journalists regulate themselves against the number.
05The clickbait pressure is structural: popularity and public interest are not the same, and optimising for the first can damage the second.

Next class: Knowing the Audience. We move from listening to researching: segmentation, personas, and the mixed-method tradition.

Digital Audience · Lecture 10

The click tells you what.
The comment tells you why.

This week’s reading

Reading list

PreviousL09: Counting the Audience NextL11: Knowing the Audience
Back to top ↑