Methodology

How Currents reads a category.

Content analysed at scale, organised into topics, measured for momentum, with evidence attached.

See it on your categoryRead the API docs
Developed with YouTube

One video, read.

Pet care is the example.

Analyzed 1,418 videos in the Pet Care category
{  "category": "pet care",  "prepared_at": "2026-09-28T06:00:00Z",  "topics": [{    "topic": "dog-anxiety-and-calming-care",    "iab": { "ids": ["425"], "path": "Pets > Dogs" },    "terms": ["dog anxiety", "calming chews", "crate training"],    "momentum_index": 92,    "trajectory": "breakout",    "suitability": { "rating": "safe", "reason": "none" },    "evidence": [      "https://youtu.be/k3FqW8dLm2A",      "https://youtu.be/Zr7pQx1VnTc",      "https://youtu.be/b9HsYe4JuKo"    ]  }]}

Eight steps from video to intelligence.

  1. 01

    Source data

    Public social video across 110 countries and 42 languages. YouTube today; the method does not depend on one source.

    yields 120M+ videos · 18 months of history per category

  2. 02

    Video understanding

    Every video is read inside, not from its title: frames, audio and transcript, in the source language.

    yields 8.5B data points

  3. 03

    Visual AI analysis

    Vision-language models describe what is shown, said and discussed, including the comments.

    yields subjects, objects, claims, reactions

  4. 04

    Topic identification

    Videos cluster into topics, named in plain language and expressed as keywords.

    yields topics · keywords

  5. 05

    Category classification

    Each topic is placed in its advertiser or editorial category and mapped to IAB Content Taxonomy 3.1.

    yields category · IAB IDs and path

  6. 06

    Momentum and trend analysis

    A proprietary momentum model scores each topic against 18 months of category history.

    yields momentum index · breakout, rising, stable, cooling

  7. 07

    Supporting evidence

    Every topic links to the videos behind it. Every placement flag carries a written reason, on the GARM framework.

    yields evidence links · suitability rating and reason

  8. 08

    Intelligence distribution

    Planning tools over MCP, DSP contextual targeting, Google Ads and DV360, the API, Wire. Rebuilt daily; high-velocity placements hourly.

    yields MCP · feeds · REST · AdCP · Wire

The signal, in numbers.

0M+
videos analysed
0.0B
data points
0
countries
0
languages
0 mo
of category history
24h
rebuild; high-velocity placements hourly

Measured, not eyeballed.

“Rising” means rising against the category’s own 18-month baseline.

IAB Content Taxonomy 3.1
IAB IDs and path on every topic.
GARM framework
Four risk tiers, every flag with a written reason.
Cookieless and ID-less
No cookies, no user IDs.
MCP and AdCP
Open protocols for planning tools and buying agents.

See the method on your category.

A live session on your categories and markets.

See your category