The Master Guide / 02 The brief and research · Research
Research as a funnel
Research deepens from the brief to the concept to the treatment, then narrows in production. Each stage researches only what its decisions and open questions need, and writes back short conclusions.
Research deepens from the brief to the concept to the treatment, then narrows in production. Each stage researches only what its decisions and open questions need, and writes back short conclusions.
The funnel [proven for the principle; the counts are working defaults, best available, not yet validated]:
Stage | Depth | What is researched |
Stage 01 (Brief) | Facts fully; culture broadly | Brand, product, audience, market, competitors' last two years, a quick culture scan (market, dialect group, red lines, 5–10 recent local ads), award work on a similar problem |
Stage 02 (Concepts) | Broader, in 3–4 parallel strands | Category codes from 10–20 local ads, the culture and genre scene, audience tensions from real comments, the language of the locked sub-dialect, and the originality of the shortlist |
Stage 03 (Treatment) | Deepest, but only on what the concept locked | Casting types for that city and class. Wardrobe, set dressing and props for that place, period and income. Dialogue drafts. Beat timing measured from 2–3 reference clips. A risk pass |
Production (Stages 04–07) | Narrow and practical | Pronunciation notes, timing per shot, quick fact checks |
The seven areas at the brief [proven], each ending in one line on what it means for the film:
Area | Ends with |
1. Brand: history, positioning, brand book, past campaigns, sensitive topics | "So the film must, or must not…" |
2. Product: what it really does better, its exact pack and label, price tier | The one product truth the film can show |
3. Audience: who buys, who decides, how they talk; real comments in their language | The behaviour to change, in their words |
4. Market and category: the codes (what every ad shows, what looks dated, what is rising) | Where the film can stand apart |
5. Competitors' films of the last two years | The space no one owns |
6. Culture: rituals, places, phrases and seasons; red lines | What to use, what to avoid |
7. Award work on a similar problem: 3–5 films, found by problem and mechanism (see References → Which layer each stage studies) | The mechanism worth learning from |
Restricted content for the category is asked at the meeting and locked [proven].
Rules at every stage:
Every claim carries its source, date and confidence [proven]: one line per conclusion, then "Means for the film: …".
A cultural claim that shapes a concept needs two independent, dated sources [proven].
Re-check anything older than 2023 about trends, creators, slang or dialect habits (decided).
Firsthand evidence first (decided): the product, the brand's own channels, real comment threads read in the market's language. Sentiment tools trained mostly on Modern Standard Arabic misread Gulf dialect, sarcasm and code-switching.
Open every cultural research request with one line of framing (decided): market, sub-dialect, audience, red lines. Research may run in Arabic while documents stay in English.
Never reopen a locked decision because research found something new (decided). Bring it to whoever owns that decision, you or the client, as a question: what was found, and what it would change.
No invented facts. An unknown is an open question.
Parallel strands at the concept stage (decided): 3–4 on independent topics (category codes, the culture and genre scene, audience tensions, the language of the sub-dialect). Each gets a written brief: its cultural framing line, the one question it answers, its scope and what it leaves to the others, its sources, its output and when to stop. Each returns a condensed summary, with the full notes in a file.
At the treatment, research only the locked choices, ending in a risk pass: red lines, a stereotype check on every joke (who is the joke on, and would that group laugh with it?), religious and cultural facts against primary sources, and a second source for every cultural claim used. Results stay in the project file.
Sources
Tash Walker, Research Live, "The write stuff".
Ipsos on category codes.
Altinisik et al., AraBehave (2026, one study: one sentence of cultural framing raised one model's cultural score from about 3.8 to 4.57 of 5).