Netflix's Global Storytelling Expansion: Modernizing Localization Analytics

As Netflix grew to a scale of 300M+ members, 190+ countries, and 50+ languages, it had to handle more dubbing and subtitling assets, and in that process analytics workflows became fragmented, pipelines were duplicated, and dashboards became siloed.

The core challenge was questions like "Who made this dub?" Dubbing/subtitling production history has to be connected across multiple data sources using complex, constantly changing logic, and the rules also vary depending on asset type and production workflow. When this logic gets replicated across pipelines for each use case, both inconsistencies between reports and maintenance burden grow at the same time.

Accordingly, Netflix carried out modernization centered on consolidation, standardization, and trust.

• It audited more than 40 dashboards and tools to review usage and code quality, prioritizing backend pipeline consolidation over frontends.
• It consolidated 3 legacy dashboards covering dubbing partner KPIs into a single data/backend layer, building a foundation that can later be extended into various frontends.
• To reduce "Not-So-Tech Debt" — which includes not just technical debt but also interpretation difficulty and storytelling problems — it revamped the Language Asset Consumption tool to combine audio and text languages into a single consumption language.
• Through this, it distinguished between Original Language and Localized Consumption, making it more intuitive to see whether members prefer subtitles, dubbing, or a mixed form.
• It shifted to a write once, read many structure, centralizing business logic in a unified table like Language Asset Producer. This table is also reused across sub-domains such as Dub Quality and Translation Quality, so logic changes propagate across everything immediately.

The next step is event-level analytics. Netflix is building a general-purpose data model that captures fine-grained timed-text events, such as individual subtitle lines, and aims to analyze how characteristics like subtitle reading speed affect member engagement in order to provide subtitle linguists with better style guides. Ultimately, the goal is to more accurately measure the experience of localized content anywhere in the world and connect it to greater enjoyment.