CREATIVE INDUSTRIES · CONTENT DECISION
A global streaming platform had received from a Korean studio the 62-minute pilot of a series combining psychological thriller and fantasy. The cast brought together an established Korean star and two emerging actors. Season 1 order: 60 million dollars, with a three-season commitment totaling 180 million over three years.
For the platform, it is a pilot. For the viewer, it is another question: do I want to spend the next episodes in this universe? The decision had to be made within three weeks.
THE PROBLEM
A pilot can be original, spectacular, intriguing. Ordering the rest requires something else: viewers must want to stay with the characters, accept the pace, tolerate unfamiliar cultural codes, wait for what comes next, then recommend the series once the initial surprise has passed.
The budget placed the series among the platform's most expensive productions in the Asian market, with limited room for error. The question before internal decision-makers was therefore not “did you like episode 1?” but: does this universe generate an engagement trajectory sufficient to justify an entire season?
The international editorial director used our system three weeks before the final decision, with results expected within four weeks.
THE THREE OPTIONS ON THE TABLE
A DISTINCTION: AN OPINION IS NOT A TRAJECTORY
Thirty-five people in California and Seoul had judged the series original but slow. Internal editorial experts were divided over the same object. A split does not indicate which of the two readings will turn into an audience.
A MECHANISM: INITIAL ATTRACTION IS NOT NARRATIVE ENDURANCE
A very strong pilot can create immediate desire without producing lasting attachment. The simulation therefore does not measure an impression; it projects a sequence of successive states: attention, emotional engagement, intellectual investment, intention to recommend.
The shift from intrigued viewer to viewer-ambassador happens after the pilot, at a precise point in the next episode. That transition, not the score of the first episode, carries the audience trajectory.
A MECHANISM: CULTURAL SPECIFICITY IS BOTH AN ASSET AND A FRICTION
Genre is not the discriminating variable: psychological thriller and fantasy work across all six markets. What diverges is the encounter between narrative pace and culturally differentiated expectations.
Korean specificity is therefore not a defect to erase. It produces rising engagement in some populations and entry friction in others. The simulation is used to determine where it does one and where it does the other.
A DISTINCTION: A MARKET IS NOT A HOMOGENEOUS CULTURE
The platform's six cultural markets — Korea, Japan, United States, France, Brazil, United Arab Emirates — each contain multiple populations. A US viewer familiar with Korean series is closer to a Japanese genre fan than to their neighbor. Within-market variation matters as much as between-market variation.
WHAT IS TESTED
These are not nine launch campaigns. They are nine object hypotheses: cut, music, voices, packaging, timetable and territorialization, crossed in varying combinations.
6 200 synthetic viewers watched the pilot, each exposed to the 9 variants in randomized order, with dynamic follow-ups on transition points identified in real time.
The reconstructed population covers 1,2 million viewers across the six target cultural markets, calibrated using public data — audiences for reference series by market, national demographic and cultural data — and the platform's proprietary data: internal viewer typologies, consumption history by market, genre preferences by cohort.
It is structured into 24 typologies of relationship with genre series, combining age, relationship with non-English-language series, narrative-pacing preference, previous exposure to Korean series and integration into cultural recommendation networks. No personal data entered the system.
RELATIONSHIP WITH NON-ENGLISH-LANGUAGE SERIES
from established habit to stated resistance: language is not a national boundary.
NARRATIVE-PACING PREFERENCE
the variable that most strongly separates reactions to the same 62-minute cut.
PREVIOUS EXPOSURE TO KOREAN SERIES
established fans, genre viewers with no prior affinity, first-time viewers.
INTEGRATION INTO RECOMMENDATION NETWORKS
what determines whether individual support spreads or remains isolated.
These configurations reveal part of the population's heterogeneity. The simulation operates on synthetic individuals, not a handful of persona types.
REACTIONS
IT IS NOT THE GENRE THAT DIVIDES; IT IS THE PACING
The psychological thriller and fantasy elements work across all 6 markets. What diverges is narrative pacing: the pilot is perceived as slow by an average US viewer, with a measurable attention drop at 18 minutes; as well-paced by a Korean or Japanese viewer, with engagement rising throughout; and as acceptable by a French viewer, with stable attention. The problem is therefore not the quality of the series: releasing simultaneously everywhere with the same cut optimizes 2 markets out of 6.
MUSIC ACCOUNTS FOR 40 % OF THE DIFFERENCE IN RECEPTION
Over the same images, the atmospheric bed produces +34 pts of engagement in East Asian markets and −18 pts in Western markets. The Western orchestral score does the reverse: +28 pts and −22 pts. The hybrid variant optimizes none of the markets. Territory-specific remixing, technically possible in post-production but never done at this scale on a series, makes it possible to optimize all 6 markets simultaneously for 400 000 dollars and a projected impact of +34 pts in average engagement.
WORD OF MOUTH IS TRIGGERED IN SCENE 4 OF EPISODE 2
Converted viewers — those who actively recommend the series within 72 hours of watching — do not become recommenders at the end of the pilot. They do so after a psychological-revelation scene that brings the pilot's elements into coherence and opens the fantasy dimension. Teasing that specific scene, rather than using the generic pilot teaser, amplifies the recommendation effect by 42 %. The viral material is not in the pilot already cut: it is in the unedited footage from episode 2.
THE 24 TYPOLOGIES DO NOT CONVERT AT THE SAME SPEED
Seven typologies — established Korean-series fans, international cinephiles, psychological-thriller genre viewers — convert in 1 to 2 weeks and feed the first viral wave. Eleven typologies convert in 4 to 6 weeks if word of mouth from the first wave is strong. Six typologies convert in 10 to 14 weeks, or do not convert. A campaign spread evenly over 12 weeks misses the first wave and never activates the second.
A COMBINED ARCHITECTURE OUTPERFORMS 7 OF THE 9 VARIANTS
Territory-specific release with unchanged images and pacing but regionally adapted soundtracks and differentiated packaging, teaser centered on scene 4 of episode 2, campaign concentrated in the first 3 weeks: this combination outperforms 7 of the 9 variants tested. Additional post-production and marketing cost of 3,2 million dollars, for a projected +34 % differential in global audience over 12 weeks versus the standard release strategy.
A MECHANISM: LOCALIZE WITHOUT DILUTING
When there is entry friction, the temptation is to simplify, explain and universalize. The hybrid variant built in post-production is exactly that: an average compromise. It optimizes none of the markets.
The effective lever touches neither the images, nor the pacing, nor the object's distinctiveness: it sits in an adaptable layer, the music, remixed market by market. Accessibility and specificity are not traded off in the middle; they are traded off layer by layer.
A TEMPORAL MECHANISM: PROPAGATION THROUGH RECOMMENDATION
Typologies do not convert at the same speed, and the slowest only convert if the fastest have recommended. The twelve-week audience therefore cannot be read from the initial peak: it is read from the sequence of waves.
A MECHANISM: VIEWER TIME IS THE SCARCE RESOURCE
A highly intense but narrow audience, a broad but lukewarm audience, and a moderately engaged but highly influential audience do not produce the same future value. The pilot's aggregate score collapses them into one number. The trajectory separates them.
This is what distinguishes the seven first-wave typologies from the other seventeen: they are not the most numerous, they are the ones whose conversion in one to two weeks determines the conversion of all the others.
Ordering a season means asking these populations for hours of attention. A pilot can be excellent over 62 minutes and insufficient over a season: engagement must be assessed against the time being requested.
COMPARISON
Only dimensions for which the case documents measurements carry numerical values. “Not published” is retained rather than replaced by an estimate.
TERRITORY-SPECIFIC RELEASE
Images unchanged, soundtracks adapted 22 pts gap between markets
STANDARD RELEASE
One object everywhere projected gap of 47 pts between markets
A DISTINCTION: GO / NO-GO IS NOT YES / NO
The committee had three options: order, reject, or renegotiate an alternative pilot with editorial changes. The answer produced was none of the three as initially framed. It was: order, without touching the cut, while changing the soundtrack by territory, the campaign material and the investment timetable.
The pilot is an option, not the final product. Its value is not only as content: it is information about the series' future potential. That shift is what makes it possible to arbitrate a 180-million-dollar investment on an incomplete object.
Most content decisions are made before the complete product exists. This case adds investment arbitration on an incomplete object to the library: pilot, populations, reactions, season hypotheses, propagation, multi-market potential, decision.
REAL-WORLD DEPLOYMENT
The international editorial committee approved the full season-1 order based on the projections, adopting the identified dominant strategy. The Korean studio agreed to remix the soundtrack in post-production into three territorial variants — one for East Asian markets, one for Western markets, one intermediate version for emerging markets — with an additional cost of 480 000 dollars integrated into the season budget.
The marketing campaign was built around the teaser for scene 4 of episode 2, with differentiated packaging by market and investment concentrated in the first three weeks after release. The global marketing budget was adjusted by 8 million dollars to incorporate that concentration.
The series was released 14 months after the ordering decision, on the original schedule: simultaneous launch across all six markets, with the three remixed soundtrack variants, differentiated packaging and priority targeting of the seven first-wave typologies.
DECISION
PROJECTION, THEN OBSERVATION
Over the first twelve weeks, the audience trajectory followed the projections with an average gap below 8 %. The first viral wave began in the second week, driven by scene 4 of episode 2, which was widely shared on social networks. The second wave began in week 4, activated by word of mouth from the first.
At twelve weeks, the series had reached 68 million household views across the six target markets, above the 62 million projection. Performance was balanced: a maximum differential of 22 points between the best-performing market, Korea, and the worst-performing, the United States, versus 47 points projected under the standard release strategy.
The conversion rate into active recommenders, measured through social-sharing metrics and in-platform recommendation, was 78 % above the average for comparable series. Season 2 was confirmed in week 8, rather than on the initial end-of-season-1 decision timetable, with a commitment to develop an adapted soundtrack from the production stage. The platform industrialized the methodology across three other high-budget international series orders over the following 18 months.
These values describe what was observed after the decision. They do not establish exclusive causality between the strategy selected and each of these movements.
TAKEAWAY
The question was whether this universe deserved the hours of attention represented by an entire season, across populations whose pacing expectations, languages and recommendation networks differ. The pilot was not the product being evaluated: it was the information available about a product that did not yet exist.
International potential did not come from cultural neutralization. The hybrid variant, the compromise version, optimized no market. It came from the ability of the distinctive object to remain legible across multiple populations, provided adaptable layers were adapted without touching the layers that define the object.
POSSIBLE FUTURES
A
DECIDE ON THE PILOT SCORE
Decide based on internal focus groups: 35 people, California and Seoul
B
UNIVERSALIZE THE CUT
Simultaneous release, one cut and one soundtrack across all 6 markets
C
DECIDE ON THE TRAJECTORY
Territory-specific release, adapted soundtracks, scene-4 teaser, campaign concentrated over 3 weeks
METHOD
1,2 million synthetic viewers across six cultural markets, calibrated using public data and the platform's proprietary data, structured into 24 typologies. 6 200 viewers watched the pilot, each exposed to the 9 variants in randomized order, with dynamic follow-ups on transition points identified in real time: attention, emotional engagement, intellectual investment, intention to recommend. Interviews were conducted in parallel across 1 200 simultaneous compute threads, compressing the equivalent of three months of international focus groups into less than five minutes of computation.
This granularity made it possible to isolate editing effects, music effects, dubbing effects and marketing effects independently, a decomposition conventional focus groups do not perform. The word-of-mouth trajectory was then projected over the 12 weeks following release for each combination, modeling recommendation networks, dominant social networks by market, critical relays and conversion effects between typologies: 216 distinct trajectories per variant, or 1 944 trajectories in total.
This case comes from a simulation carried out for a global streaming platform facing an ordering decision committing 180 million dollars over three years. The client is not named, no personal data entered the system, and the detailed results remain the client's property.
The simulation does not decide the artistic value of the work, the budget or the production schedule. It concerns the consequences, across different populations, of several hypotheses for editing, music, dubbing, packaging and release timing.
Your next decision
Describe your need. We can point you to the right level of support.
State your decision. See the future it produces.
