How to Research Short Drama Audiences Before Choosing a Genre

How to Research Short Drama Audiences Before Choosing a Genre

Research your audience before you pick a genre by gathering three kinds of demand evidence: behavioral interviews with 10 to 15 target viewers, coded comments from competitor dramas, and search-intent signals that show what people actively look for rather than what algorithms push at them. Score each candidate genre on evidence strength, competition density, and your own production capacity, then test the winner with a small pilot before committing to a full series.

Why genre choice is an evidence problem, not a taste problem

Most first-time short drama creators pick a genre the way they pick a movie on a Friday night: by mood, by habit, or by whatever is trending on their own feed. That works for entertainment. It fails for production, because a genre decision locks in your script structure, your character archetypes, your episode pacing, your hook design, and ultimately your monetization ceiling for the next several weeks of work.

The cost of a wrong genre call is not abstract. In this framework, a typical AI-assisted vertical drama production runs between two and six weeks of creator time even with modern tooling, and the difference between a genre your audience actually seeks out and one they merely tolerate shows up directly in completion rate, paid unlock rate, and share rate. Creators on Lollipop Drama who validated audience demand before greenlighting a genre reported materially higher first-episode-to-tenth-episode retention than creators who chose genre by personal preference alone (Lollipop Drama internal data, as of 2026).

The trap is that every genre feels validated if you look at the wrong signals. Revenge dramas are everywhere, so revenge must be hot. Sweet romance clips get millions of likes, so sweet romance must be safe. Fantasy is underserved, so fantasy must be an opportunity. Each of these conclusions can be true or false depending on which audience you are actually serving, in which language market, at which price point, with which production capacity. The only way to know is to collect demand evidence deliberately, grade that evidence by quality, and compare candidate genres on the same scoring scale.

This article gives you the full methodology: an eight-question behavioral interview template, a comment-coding scheme for competitor dramas, a three-factor genre scoring matrix with worked examples, three rules for separating real demand from social media noise, and the minimum viable sample sizes you need before you can make a defensible genre decision. If you are looking for how to filter and shape individual story ideas once a genre is chosen, that is a separate process covered in our guide to story topic selection. This article is about what comes before that step.

The three sources of demand evidence

Audience research for short drama rests on three evidence sources, ranked from strongest to weakest in this framework.

First, behavioral interviews. Talking to ten to fifteen people in your target audience about what they actually watched, finished, abandoned, and paid for gives you primary evidence. It is slow but nearly impossible to fake, because you are asking about past behavior, not stated preference.

Second, coded competitor comments. The comment sections of successful and failed dramas in your candidate genres are a public archive of audience reactions. Classified properly, they tell you which story beats viewers explicitly praised, which they complained about, and which made them demand more episodes. This is secondary evidence: abundant, cheap, but noisy.

Third, platform and search signals. What people type into a search bar is active intent. What a recommendation feed shows them is passive exposure. Search-intent data, autocomplete suggestions, and platform keyword tools are tertiary evidence: useful for directional confirmation, dangerous as a sole basis for a genre bet.

A defensible genre decision uses all three, with interviews weighted heaviest. If your interview subjects describe pain points that match the complaints you coded in competitor comments, and search data shows people actively looking for that genre rather than just bumping into it, you have convergent evidence. Any single source alone is a hypothesis, not a conclusion.

The audience interview: an eight-question behavioral template

The most common interviewing mistake is asking opinion questions. "What genre do you like?" produces socially acceptable, lazy answers. People say they like thrillers and then binge romance for six hours. Behavior questions avoid this, because memory of what someone actually did is far more accurate than their self-image as a viewer.

Run these as one-on-one conversations of 20 to 30 minutes, in the viewer's own language, with people who match your target demographic and have finished at least one vertical short drama in the past three months (in this framework, that recency filter matters more than total viewing volume).

1. "What was the last short drama you watched all the way to the end?" This question verifies that the person is a genuine completer, not a casual scroller, and it anchors the whole interview in a specific, memorable artifact rather than abstractions.

1. "At which episode did you decide to keep watching, and what happened in that episode?" This identifies the actual hook mechanics that worked on a real viewer, so you can see whether your candidate genre's standard hooks match what retains this audience.

1. "What was the last short drama you abandoned, and at which episode did you stop?" Abandonment data is more diagnostic than completion data. The episode where they quit reveals which beats, pacing failures, or trope executions lose this audience.

1. "What were you doing when you watched it, and on what device?" Viewing context shapes genre fit. Commute viewers and bedtime viewers have different tolerance for complexity, dialogue density, and episode length, which connects directly to decisions you will make about episode length later.

1. "Have you ever paid to unlock episodes of a short drama? Which one, and what was happening in the story at the paywall?" This is the single most important question in the template. It converts vague preference into revealed willingness to pay and tells you which story moments in each genre actually trigger payment.

1. "Have you ever recommended a short drama to someone? Which one, and what did you say when you recommended it?" The exact words people use to recommend a drama are your future marketing copy, and the dramas they recommend reveal which genres carry social currency in this audience.

1. "Is there a type of short drama you looked for but could not find, or could not find a good one of?" This is your demand gap detector. Viewers who actively searched for something and came up empty are describing unmet demand in their own words.

1. "If a new drama in a genre you like came out tomorrow, what would make you tap on it in the feed: the title, the cover, the first three seconds, or something else?" This tells you which discovery levers matter for this audience, which will later inform your work on click-through rate, but for now it tells you whether the genre itself is the draw or whether packaging does the heavy lifting.

Two execution rules in this framework. First, never show the interviewee your own project or pitch; the moment you do, you are collecting politeness, not data. Second, record exact phrases verbatim. "I stopped when the male lead forgave her too easily" is usable evidence. "The pacing was off" is not, until you ask "off how?" and get the specific beat.

Coding competitor comment sections

Your second evidence source is the comment sections of five to ten competitor dramas across your candidate genres, mixing hits and flops so you capture both praise and complaint. In this framework, every comment you code falls into one of three categories:

Coding rules in this framework: code at the level of story beat, not sentiment. "I love this" is uncodeable until you know what "this" refers to; use the reply thread or the timestamp to attach it to a beat. Code each comment once, into its dominant intent. Keep a raw count per genre plus a per-episode map, because a genre whose payoff points cluster at episodes 1 to 3 but whose pain points cluster at episodes 15 to 30 has a mid-series retention problem baked into audience expectations.

When you finish, each candidate genre produces three numbers: payoff density (payoff comments per hundred coded), pain density, and demand-more density. In this framework, the demand-more density is the strongest single predictor of binge potential, because it captures viewers whose engagement was strong enough to convert into active pursuit, the same behavior pattern behind the metrics in our binge-watch retention framework.

The genre demand scoring matrix

Now combine your evidence into one comparable score per candidate genre. The matrix has three factors, each scored 1 to 5, with explicit anchors so your scoring is consistent across genres and across time.

Total score is the sum, out of 15. In this framework, 12 or above means greenlight a pilot; 9 to 11 means fix the weakest factor first or narrow the subgenre; below 9 means park it.

Worked example with three candidate genres, using hypothetical but realistic research outputs for a solo creator targeting an English-language female-skewing audience, in this framework:

The reasoning behind the scores. Revenge comeback scored 5 on demand evidence because interviewees named it unprompted, its demand-more density was the highest of the three, and search queries included specific plot patterns, but only 2 on competition because the genre is saturated with strong incumbents, and 4 on capacity because its confrontation-heavy scenes are easy to stage with few characters. Sweet romance scored 4 on demand, 2 on competition because saturation is even worse, and 5 on capacity because two-person dialogue scenes are the cheapest and most consistency-friendly format in AI production. Suspense scored 3 on demand because interest was real but diffuse, 4 on competition because the genre is underserved in this niche, and only 2 on capacity because suspense depends on precise visual information control, exactly where AI video artifacts hurt most, as covered in our guide to fixing AI video artifacts.

Note what the matrix prevents: the creator's personal favorite was suspense, and the "hottest" genre by raw feed impressions was sweet romance. The evidence says revenge and sweet romance tie at 11, and the tiebreaker is the weakest-factor rule: sweet romance's bottleneck is competition, fixable by niching into a fresher trope combination, while revenge's bottleneck is also competition but its capacity fit is slightly worse. The creator pilots a niched sweet romance first, then a revenge title with a differentiated hook second. That is a decision you can defend to yourself six weeks later.

Real demand versus social media heat

The most expensive mistake in genre selection is confusing what people actively seek with what algorithms passively push. A genre can dominate your feed because platforms amplified it, because a few big spenders promoted it, or because it is genuinely demanded. Only the third case is evidence. Use these three rules in this framework:

Rule one: heat terms must travel with specific title searches. Real demand shows up as searches for specific drama names, actor names, or genre-plus-plot combinations. Pure noise shows up as hashtag volume and meme remixes with no associated title-level search interest. If millions of people are joking about a trope but almost nobody searches for an actual drama featuring it, you are looking at entertainment about the genre, not demand for the genre.

Rule two: comment sections must show demand-more behavior, not just meme participation. Scroll the comments under the viral clips. If the top comments are jokes, reaction quotes, and arguments, the attention is social. If they are "where can I watch the full version," "what is this drama called," "update faster," and "is there a season two," the attention is commercial. In this framework, a viral clip with under roughly one demand-more comment per fifty top comments is noise until proven otherwise.

Rule three: real demand decays slowly, noise decays fast. Re-check the conversation volume two weeks after the spike. Genuine genre demand flattens into a stable baseline of searches, comments, and new viewer questions. Platform-driven noise typically collapses once the amplification stops. A two-week persistence check costs you nothing but patience and filters out the majority of false signals.

Creators who skip these checks tend to greenlight genres at the exact peak of their noise curve, then launch four to six weeks later into a feed that has moved on. The two-week check alone, applied consistently, would have prevented a meaningful share of abandoned productions we observe on the platform (Lollipop Drama internal data, as of 2026).

Sample size and weekly cadence: the minimum viable research

Audience research expands to fill whatever time you give it, so cap it. In this framework, the minimum viable research package for a genre decision is:

• Ten to fifteen behavioral interviews, spread across at least two weeks so you are not sampling one news cycle of taste. Below eight interviews, individual quirks dominate; above twenty, marginal learning drops sharply for a solo creator.
• Three hundred to five hundred coded comments, drawn from five to ten competitor dramas across all candidate genres combined, with at least fifty comments per candidate genre. That is enough to compute payoff, pain, and demand-more densities without drowning.
• One search-intent pass per candidate genre: autocomplete harvest, keyword tool check, and the two-week persistence re-check for anything that looked hot.

Time budget: four to six focused hours per week for three weeks, in this framework. Week one for recruiting and interviews, week two for finishing interviews and coding comments, week three for search checks, scoring, and the matrix. Three weeks feels slow when you want to start generating footage, but it is short compared to the cost of producing fifty episodes of a genre your audience only pretended to want. And it plugs directly into the next de-risking step: once the matrix picks a winner, test the concept with a pilot before committing to the full run.

Common mistakes

From research to greenlight

The output of this process is not certainty; it is a ranked, evidence-backed shortlist with known weak factors. That is exactly the right input for the next stage: write the pilot, stress-test the hooks using the three-second rule, keep characters stable with a character bible, and validate with real viewers at small scale before the full production budget moves. Creators who combine pre-genre audience research with a pilot test reach series-level decisions with far less sunk cost than creators who do either step alone (Lollipop Drama internal data, as of 2026).

Genre is the highest-leverage decision in a short drama's life, and it is made at the moment you know the least. That is precisely why it deserves a method instead of a hunch.

Related reading

• How to filter and select story ideas after you pick a genre
• Test your AI short drama with a pilot before committing to a full series
• The binge-watch retention framework for vertical drama
• Hook architecture and the three-second rule in short dramas
• Choosing the right episode length for your short drama
• Analytics that matter: retention, completion, and revenue