Observations first · analysis labelled · models kept apart

A public record of how discovery systems change

The Research Observatory preserves timestamped observations of search and YouTube results, publishes its methodology, and separates factual replay from analysis and simulations derived from the record.

0
published datasets
0
verified snapshots
0
days on record
5
tracked queries

No real dataset has been published yet; collection is under way.

Datasets

Try it on the demo

DEMOFictional sample data. Every site is invented on the reserved .example domain and every video and channel is made up, so the replay and analysis can be tried before real observations are published. It is replaced automatically once real data is published.
SearchDEMORULE-LABELLEDlast observed 2026-08-14

Home battery backup (demo)

Fictional sample: 45 days of top-10 results for “home battery backup systems”, with a new entrant, an exit and one turbulent day, so replay and analysis can be tried before real data is published.

45
snapshots
1
query
US
desktop
volatility, derived
SearchDEMORULE-LABELLEDlast observed 2026-08-14

Solid-state battery startups (demo)

Fictional sample: 45 days of top-10 results for “solid state battery startups”, including a late entrant that climbs quickly. Every site is invented on the .example domain.

45
snapshots
1
query
US
desktop
volatility, derived
YouTubeDEMORULE-LABELLEDlast observed 2026-08-14

Home battery install videos (demo)

Fictional sample: 45 days of YouTube relevance positions and public view counts for “home battery install”, including one breakout video. Every video and channel is invented.

45
snapshots
1
query
US
region
volatility, derived

Systems

Two discovery systems, one standard

Ready, not yet collecting

YouTube Observatory

Relevance positions for fixed queries, with each video’s public view, like and comment counts. Dataset pages replay the rankings and derive view velocity, lifecycles and cohort breakouts, with the comparison window stated beside every number.

Fixed inputs

What is being collected

The tracked queries are declared in advance and versioned with the methodology, so nobody chooses what to record after seeing the results. Collection alone does not publish anything: a dataset appears above only after a reviewed publication.

Query setSystemQueryLocationDeviceDepthStatus
Battery market discoverySearch“fastest growing battery companies”USdesktop20collecting
Battery market discoverySearch“solid state battery companies”USdesktop20collecting
Battery market discoverySearch“grid scale battery storage companies”USdesktop20collecting
Battery market discoverySearch“battery recycling companies”USdesktop20collecting
Battery market discoverySearch“sodium ion battery companies”USdesktop20collecting
Battery market videosYouTube“home battery install”US—20configured, not running
Battery market videosYouTube“solid state battery explained”US—20configured, not running
Battery market videosYouTube“battery recycling”US—20configured, not running

Truth contract

Replay, analysis and simulation are different things

OBSERVED

Replay

Reproduces a timestamped external observation, exactly as recorded and checksummed. Nothing is filled in.

DERIVED

Analysis

Numbers computed from the record by a published formula: volatility, trajectories, share, velocity. An interpretation, labelled as one.

MODEL

Simulation

The output of a documented model after you change its inputs, such as the search labs. It shows how a mechanism behaves, not what a system did.

Methodology

Changelog

  1. AI labelling 1.0

    Structured labels and weekly narratives, reviewed in a private console.

    • Each distinct result or video is labelled once with the entities it is about, its content type and the intent it serves, from a closed vocabulary (instructions items-1.0). An entity is kept only if its name appears in the item’s own title, site or channel.
    • Labels live beside the archive, never inside it. A person can correct or reject any label; a correction is never overwritten by a later model run. Names are merged into one entity per real-world entity when data is published.
    • Weekly narratives are drafted from facts computed from the record, never from raw results. Every number in a narrative must appear in a fact it cites, and only narratives a person approved are published.
    • Labels are published only for datasets whose approval allows it, with the share a person reviewed, and the published file is checksummed like the snapshots.

  2. YouTube 1.0

    YouTube collection protocol.

    • Relevance-ordered video ids for a query and region from the YouTube Data API v3, then each video’s public view, like and comment counts in the same run.
    • Hidden counts are stored as null, never estimated. A video removed between the two requests is left out.
    • No device dimension: the API does not offer one, so YouTube series are recorded as "unspecified" rather than claimed as desktop or mobile.

  3. Derived analysis 1.0

    Derived numbers published beside replay.

    • Volatility = (1 − rank-biased overlap) × 100 between consecutive snapshots of one series, with persistence p = 0.9.
    • Rank trajectories, domain share of the top ten, tenure, view velocity, cohort breakouts (≥ 3× the median velocity over 7 days) and view lifecycles.
    • An absent result has no rank: charts leave a gap rather than inventing one.

  4. Publication 1.0

    Demo sample policy.

    • A fictional sample on reserved .example domains may be shown for a system with no published data. It is labelled DEMO on every surface, excluded from search indexing and structured data, and withdrawn when real data is published.

  5. Web search 1.0

    Initial web-search protocol.

    • Fixed query sets with declared locations, devices and depth; one immutable, checksummed snapshot per query, location, device and time.
    • URL normalization: HTTP(S) only, fragments and campaign parameters removed, duplicates discarded, observed rank kept.
    • Reviewed publication with minimum history, maximum gap and provenance checks.

Read the full methodology →

Open data

Check it yourself

The catalog lists every published dataset. Each dataset’s index lists its snapshot files with SHA-256 checksums, and a CSV flattens the same observations. Nothing on a dataset page comes from anywhere else.