Thesis

It began with a personal pain point: we love watching movies, but finding time for a 2-3 hour sit is tough in daily life, and pausing mid-movie makes for a poor experience. Looking deeper, we saw many users struggle to start or finish films and long episodes, whether due to busy schedules or ever-shorter attention spans, while streaming services face fierce competition and rising content acquisition and production costs. We realized there was an opportunity to unlock value from underutilized content by reimagining movies as mini-series – dividing them into bite-sized episodes, making them easier to consume. Our vision was to use AI to split and edit films into natural-feeling episodes with recaps, opening titles, and end credits, giving viewers flexibility not just in what and when they watch, but also how – as a movie or as a series. This would reduce the friction users face with long-form content, while also helping streamers extract more value from their catalogs. 

We estimated our solution could drive significant profit for streaming companies by improving content utilization (thus reducing spend) and boosting user retention and acquisition through a unique value proposition. In 2024, for example, a 1% cut in Netflix’s $17B content budget equals $170M in annual profit, while a 1% lift in its $39B subscription revenue adds $390M (Source). We believed our solution could capture these gains and more by partnering with an exclusive client, enabling them to deliver differentiated value before competitors caught up. We further believed that a superior user experience could shift viewers away from piracy, helping recover part of the $30B-$70B the industry loses annually (Source).

Validation

To validate, we began with user research:

We discovered countless examples of people breaking movies into parts themselves, asking others in forums to do it for them, or even publishing guides on how to watch films as mini-series. These posts, threads, and articles attracted dozens to hundreds of comments, likes, and upvotes – with some single sources drawing nearly a thousand users. This provided strong validation of the concept and clear evidence of user demand.

See example

The re-release of Quentin Tarantino’s, “The Hateful Eight”, as a four-episode miniseries on Netflix was a strong proof point, as were other films successfully re-released in segmented formats.

By analyzing ad placements on U.S. streaming platforms (including free-to-watch services like Tubi), we saw that current breaks were often poorly timed. We thought that our solution could create more natural insertion points that improved both viewer experience and monetization.

To test the concept, we manually analyzed and chapterized well-known films to derive principles for segmentation. Around 30 friends and family who viewed these chapterized versions gave highly positive feedback, saying they enjoyed the experience, asked for more chapterized movies, and even wanted the same option for series episodes.

For the algorithm itself, we experimented with different approaches and concluded that combining video, audio, and subtitles was essential. To build a foundation, we created a manual annotation methodology based on film theory (including scenes, sequences, acts, plot points, and editing cues) that later guided our AI-driven solution.

As we engaged with several streaming companies and analyzed their needs, we learned our solution must align with content licensing legal constraints, minimize content manipulation to address creators’ concerns, and require minimal labor and implementation time, ensuring much higher efficiency than manual operations.

Meanwhile, we also immersed ourselves in film theory, narrative design, editing, and screenwriting. This deep study sharpened our product and algorithmic decisions and ultimately shaped the POC.

Throughout the validation process, our idea evolved from re-editing movies into mini-series to a lighter in-player approach. Instead of offering a “watch-as-series” experience with independent episodes, we introduced chapters displayed along the movie timeline, with an “end-of-chapter” screen marking each break. This version was easier to implement, sell, and integrate with existing formats and licensing agreements.

We then built and published a Chrome extension that brought the in-player concept to life. It let Netflix users watch movies in chapters directly within the platform, allowing us to test and validate the concept with real users. Starting with 20 chapterized movies, we shared it with friends and family for feedback, but the planned public release was halted after our shutdown decision.

Why it didn’t progress 

1. Dual Offering Challenge: We offered both the concept of chapterized content and the software to create it. Once the concept was validated, streamers could replicate it themselves, reducing our value proposition to faster implementation and reduced time to market. This is a limited business opportunity we recognized early on.
   
2. Misjudged Sales Cycles: We critically underestimated how long enterprise sales cycles are (often up to two years) and how slow decision making is within streaming giants. This made a quick exit impossible and undermined our value proposition to reduce time to market.
   
3. Existing Data and Solutions: We realized that major streamers already possess segmentation metadata and internal tools that support in-house implementation, further weakening our value proposition. 
   
4. Threat from Foundational AI: We were worried that a new foundational AI model could replicate our solution’s functionality and erode defensibility, as multimodal LLM reasoning and native context length continue to improve.
   
5. Lack of Patentability: The concept wasn’t easily patentable, which weakened potential defensibility.
   
6. Unpromising Pivots: Financial analysis for a B2C pivot revealed high upfront costs, low margins, and limited economies of scale, making it unviable. We ruled out ad-tech and other entertainment-tech pivots as they did not align with our vision and values.
   
7. Expected Value & Time Horizon: We aimed for speed and a clear exit. With low expected value, a 6-18 month push into enterprise sales or a B2C pivot did not make sense.

Insights

---