Numbers tell you what audiences do. Research tells you who they are and why they do it.
Unit C: How do we research and monitor?
By the end of today
What you will walk out knowing
01
Segment an audience
Divide an audience into meaningful groups by demographics, behaviour, geography, or psychographics, and explain why each segment matters for content, product, or advertising decisions.
Reveal
02
Build a persona from data
Combine analytics, survey, and listening data into a coherent fictional individual who represents a real audience segment, and explain what each data source contributes.
Reveal
03
Match method to research question
Identify when a survey, an interview, a focus group, or ethnographic observation is the right tool, and explain what each method reveals that the others cannot.
Reveal
04
Apply mixed-method thinking
Articulate why combining quantitative and qualitative methods produces stronger audience understanding than either tradition alone, drawing on Nightingale and the Routledge Companion.
Reveal
A Chorki product manager has one million subscribers. She knows how many they are. Does she know who they are?
Analytics gives you a number. Research gives you a person. Today we learn how to build the second from the first.
The foundation
What segmentation does
No audience is homogeneous. Segmentation is the practice of dividing a heterogeneous audience into groups that are internally similar and externally distinct enough to treat differently.
Demographic: age, gender, income, education, occupation. The oldest and most basic segmentation layer. Useful for advertising targeting, less useful for understanding behaviour.
Geographic: location at the level of country, city, or neighbourhood. In Bangladesh, Dhaka vs. district-level audiences have distinct device, income, and language profiles.
Psychographic: values, attitudes, lifestyle, motivation. The hardest to measure but the most predictive of content preference and brand loyalty.
Hoichoi's subscriber base segments into at least: Bangla drama fans, Kolkata film audience, and diaspora viewers each with different content needs and price sensitivity.
Applied segmentation
Segments in practice across sectors
Every sector uses segmentation differently. The variables, the purpose, and the stakes vary — but the underlying logic is the same.
01
News: subscriber vs casual reader
The Daily Star segments its digital audience into registered subscribers, logged-in free readers, and anonymous visitors. Each group gets a different paywall proposition. The subscriber is the revenue relationship; the anonymous visitor is the advertising commodity.
Reveal
02
OTT: by content affinity
Chorki classifies viewers by drama genre preference and device type. This drives both content commissioning and push notification strategy. A subscriber who only watches dramas on Friday evenings is a very different retention risk from a daily user.
Reveal
03
Music: heavy vs casual streamers
On Spotify and YouTube Music, Bangla music audiences segment by listening frequency, playlist-saving behaviour, and skip rates. Heavy streamers drive royalty volume; casual streamers are more likely to respond to advertising.
Reveal
04
Creators: core fans vs discoverers
A Bangladeshi street food creator distinguishes between subscribers who watch every video within 24 hours (core fans) and viewers who arrive from the recommendation algorithm (discoverers). Core fans drive comment engagement; discoverers drive total view count. Both matter but require different content strategies.
Reveal
From segment to person
The audience persona
A persona is a named, fictional individual who represents a real audience segment. It is a research tool, not a guess, built from multiple data sources.
Not invented: a well-built persona is grounded in survey data, analytics, and qualitative insight combined. Fiction in name only.
Not a stereotype: the persona should challenge assumptions, not confirm them. If your personas are all the same age bracket, your research has not gone deep enough.
A decision tool: teams use personas to ask "would Rina subscribe to this?" instead of "would our audience subscribe?" A person is easier to reason about than a demographic category.
Temporally limited: personas go stale. Audience behaviour changes. A persona built on 2021 data is probably wrong about 2024 behaviour.
Prothom Alo's digital team might maintain three personas: Shahed (urban subscriber), Nusrat (Facebook news reader), and Rafiq (district-level mobile user).
Click to build a persona
Click to build a persona
Assemble: Nusrat
Click each data card to see what input it contributes to this Prothom Alo audience persona.
01
Analytics input
Nusrat arrives 94% via Facebook. Average session: 1:42. Bounce rate: 74%. She reads on mobile, between 7-9 AM and 9-11 PM. She never visits the homepage directly.
Reveal
02
Survey input
35 years old. Lives in Mirpur, Dhaka. Works as a school teacher. Household income: lower-middle. Primary news source: Facebook feed. Does not pay for digital subscriptions.
Reveal
03
Social listening input
Her comment pattern: shares emotional reactions to human interest stories and educational content. Rarely comments on hard politics. Reacts most to stories about school curriculum and children.
Reveal
04
What she needs
Short, trustworthy, shareable stories about education, family, and local community. Content that works in a Facebook preview without a full click. She will not subscribe but she amplifies reach.
Reveal
05
What she is not
She is not a direct revenue source via subscription. But she is the organic distribution engine. Losing her costs Prothom Alo Facebook reach, not subscription revenue.
Reveal
06
The strategic implication
Content designed only for Shahed (the paying subscriber) will lose Nusrat. Content designed only for Nusrat will not retain Shahed. Both personas require different editorial products.
Reveal
Nightingale (2011) · Quantitative tradition
The twin pillars of quantitative research
Survey methods
Breadth
Standardised questions administered at scale. The goal: representative samples that allow statistical projection onto populations.
Cross-sectional: a snapshot in time
Longitudinal: change over time via trend, cohort, or panel studies
Tells you: how many, how often, what proportion
Cannot tell you: why, with what meaning
Experimental methods
Control
Isolate a single variable by holding everything else constant. Allows causal claims surveys cannot make.
Classic use: media effects, attention studies
A/B testing is an everyday experimental method
Tells you: does X cause Y?
Cannot tell you: what X means to the people experiencing it
Nightingale (2011) · Qualitative tradition
Depth over breadth
Interviews: in-depth, semi-structured conversation. Best for understanding motivation, life context, and the "why" behind behaviour. Not generalisable, but reveals the range of human responses.
Focus groups: group discussion that generates social meaning. Livingstone's research showed focus groups "identified more complex connections between text and reception" than surveys could. But group dynamics can suppress minority views.
Participant observation and ethnography: Murphy (2011) argues the researcher must make direct contact with audiences in the "normal courses and routine situations of their lives." Slow, resource-intensive, but uniquely powerful for understanding community audiences.
Creative and visual methods: Awan and Gauntlett argue for moving beyond words. Giving audiences cameras, Lego, or drawing tasks unlocks meaning that interview questions cannot reach.
Nightingale (2011) · Bryman (2006)
The mixed-method turn
The paradigm peace: Bryman's review of journal articles found that the 2000s saw "a tendency to stress the compatibility between quantitative and qualitative research" and a pragmatic orientation toward using any approach that allows research questions to be answered.
Livingstone's talk-show study: textual analysis and focus groups revealed "contradictions within audience readings." The follow-up survey then highlighted what focus groups had missed: "the importance of viewers' age compared to gender or social class." Each method found things the other could not.
The practitioner implication: analytics data (quantitative) tells an OTT platform what its subscribers watch. Audience interviews (qualitative) tell it why some of them are about to leave. Both are needed to make the right decision.
Sorting game
Sorting game
Match method to research question
Select a research question from the left, then click the method best suited to answer it.
Research questions
Sector tour
How each sector builds audience knowledge
Each sector combines quantitative and qualitative methods in its own way, shaped by its commercial logic and the questions it most urgently needs to answer.
01
News: registered user data + reader panels
The Daily Star combines logged-in user analytics (behavioural) with periodic reader surveys (quantitative) and occasional subscriber focus groups (qualitative). Analytics cannot tell you whether a reader trusts a story; a survey can.
Reveal
02
OTT: viewing data + exit interviews
Chorki tracks completion rates, device switches, and pause patterns automatically. When churn spikes, they run exit interviews with cancelling subscribers. The combination reveals not just that people left but what the tipping point was.
Reveal
03
Music: streaming data + fan community managers
Bangla artists and labels use Spotify for Artists and YouTube Studio for behavioural data. Qualitative research via fan community managers and Discord servers adds meaning: why a song resonates emotionally in a way its skip rate would not predict.
Reveal
04
Games: session telemetry + playtesting
PUBG Mobile Bangladesh operations use session-length and spending data for macro segmentation. Qualitative playtesting sessions with local players reveal UI friction points and cultural gameplay preferences that the telemetry registers as drops but cannot explain.
Reveal
Branching scenario · Research director
Branching scenario
Which method do you commission?
You are head of audience research at Chorki. Subscriber growth has stalled at 800,000 for three months. Your analytics show no change in completion rates, session length, or churn rate. The CEO asks: "Why have we stopped growing?" You have budget for one research project.
Survey choice. You get statistically robust data on barriers: price (54%), content library (38%), preference for free YouTube (61%). Useful for board reporting. But you still do not know the specific emotional and habitual reasons the 800,000 ceiling exists. The CEO has data. She still does not have a decision.
Interview choice. Twenty conversations reveal a consistent pattern: trial users loved the first drama they watched but could not find a second one that felt worth paying for. The discovery experience failed them. This is an actionable product insight your survey data could not have surfaced. Growth resumes after a content discovery redesign.
Ethnographic choice. You discover that Chorki competes not with Netflix but with family YouTube channels and Prothom Alo Facebook videos for the same 9 PM slot. The competition frame your analytics showed was wrong. This shifts your entire content strategy. The most consequential insight of the three.
Knowledge check
Knowledge check
Livingstone's mixed-method study of talk-show audiences found that focus groups and surveys produced different insights. What did the survey reveal that focus groups had missed?
B. Nightingale reports that the survey "highlighted what had been missed in the focus group analysis, namely, the importance of viewers' age compared to, say, gender or social class." The focus groups had identified complex meaning-making, but only the large-scale survey could reveal which demographic variable actually predicted viewing behaviour most strongly.
A well-built audience persona is primarily a creative exercise based on assumptions about typical users, rather than a document grounded in empirical research data.False. A robust persona is grounded in a combination of analytics data, survey findings, and qualitative insight. A persona built primarily on assumption is not a research tool; it is a guess with a name on it.
Psychographic segmentation, which groups audiences by values and lifestyle rather than demographics, tends to be harder to measure but more predictive of content preference.True. Demographic categories like age and gender are easy to collect but often poor predictors of what specific content a person will choose. Psychographic variables require richer research methods but produce segments that behave more consistently.
Your turn · 15–20 min
Activity: Build a Persona
Step 1
Choose an organisation
Pick a Bangladeshi news outlet, OTT platform, music service, or creator. This should be the same organisation you are building toward in your case study.
Step 2
Assemble one persona
Using the data snapshot on your worksheet, name and describe one audience persona. Fill every field: demographics, behaviour, motivation, and what this person needs that the organisation currently does or does not provide.
Step 3
Name the research gap
Identify one thing about your persona you could only learn through qualitative research, and state which method (interview, focus group, observation) would best surface it and why.
Today in five lines
Recap
01Segmentation divides a heterogeneous audience into actionable groups by demographics, behaviour, geography, and psychographics.
02A persona translates a segment into a named, researchable individual built from analytics, surveys, and qualitative data combined.
03Surveys give breadth; interviews give depth; ethnography gives context. Each method finds things the others miss.
04The paradigm peace means quantitative and qualitative methods are now seen as complementary, not competing.
05For Bangladeshi organisations, the hardest audience knowledge to acquire is why people choose what they choose, not what they choose.
Next class: The Algorithmic Audience. When the recommendation system, not the editor or the subscriber, decides who your audience is and what they see next.
Digital Audience · Lecture 11
A number tells you how many. Research tells you who.