How Pendulum Digests YouTube Data

Unlocking the World’s Largest Video Platform 

At Pendulum, our north star is simple: to bring every relevant public datapoint that lives outside a company’s firewall into one coherent, trustworthy lake. Today’s social media is overwhelmingly driven by audio, video, and images, so stopping at basic text analysis is no longer enough.

Nowhere is this more evident than on YouTube. With billions of uploads occurring annually—including over 260 million hours of video uploaded in 2025—you cannot vacuum up the entire platform. Instead, you have to be smart about who and what matters. 

Here is a look behind the curtain at how Pendulum ingests YouTube data, classifies it, and drives successes for some of the biggest organizations in the world.

Compliance First, Deep Intelligence Second

Serving Fortune 500 firms means passing the toughest security reviews on the planet. If a source isn't compliant, it isn't worth having. We rely strictly on public-only data—no private messages, no shady leaks, and absolutely zero impersonation via fake accounts or bot armies.

Rather than relying solely on standard APIs, we use purpose-built, Pendulum-sourced collectors that adhere to YouTube's rules while ingesting the platform's data into our unified pipeline.

Crucially, we go the extra mile by extracting the real value hidden within the video itself. We scrape titles and descriptions, we auto-transcribe speech. This means a spoken keyword on YouTube arrives in the exact same standardized stream as a news article or an X/Twitter post.

Finding the Needles Before They Are Headlines

To manage YouTube’s massive scale, our ingestion strategy relies on five key pillars:

• Head Channels First: We ingest every post from the largest creators so that when big voices move, we are already listening.
• Look-Alike Expansion: Our embedding models automatically surface and pull in fast-growing, adjacent creators.
• Topic Scouts: We seed constant discovery using a living library of over 450 global issues, scoring and promoting valuable new posts.
• Customer Triggers: The moment a brand or topic is spun up, our crawler fans out, backfilling historical data and adding new creators.
• Manual Picks: If a niche analyst is needed, dropping their handle into an Influencer List starts the pull from day zero.

Community Analytics

Our community coverage is currently strongest on YouTube, thanks to our unique ability to automatically classify channels using multiple tools. By leveraging commenter subscription data—a signal unique to YouTube—we create channel embeddings. When paired with K-nearest-neighbor algorithms and labeled data, this enables us to accurately map entire ecosystems of creators.

Our Successes: From Transparency to Corporate Strategy

We are already powering horizon-scanning for some of the largest companies on Earth, capturing more narrative-shaping content than anyone else in the market.

Our YouTube classification work began by mapping political and cultural channels, starting with an initial dataset of 750 channels that we expanded to over 7,000 for the Transparency Tube project. By late 2021, that single community alone had grown to over 50,000 channels, alongside vast datasets mapping everything from Mainstream News to Video Games and Personal Finance.

This capability has driven highly targeted, real-world success for our clients:

• A Global Bank: We developed a Personal Finance community channel to monitor key economic and business narratives.
• A Publishing House: We mapped a network of pastors spreading QAnon conspiracy theories on YouTube to support investigative journalism.

Book your custom brand briefing.