KEKOPEDIA
About

Archive methodology

How this archive was assembled: sources of truth, evidence versus interpretation, file reconciliation, keyframe handling and known uncertainties.

This site turns a supplied archive of campaign material into an interconnected, browsable history. It uses that archive as the only source of truth. Nothing here is drawn from outside research, and no missing facts, credits, identities, dates, quotations or interpretations have been invented. Where the evidence is silent or uncertain, this archive says so.

Sources and their roles

The material is used according to its evidential weight:

  • Videos, transcripts, subtitles and keyframes are direct evidence of what appears in an advertisement.
  • Individual advertisement analyses are the primary analytical source for each ad.
  • The master campaign analysis supports chronology and campaign evolution.
  • The “Lamb Brain” documents are a later synthesis of worldview, grammar, humour, tone and guardrails. They are treated as commentary on the corpus, not as official historical doctrine.

Every advertisement article separates direct evidence (visuals, dialogue, on-screen text, keyframes) from source analysis (interpretation), and preserves the analyses’ own labels — directly observable, likely interpretation at release, and verification needed.

File reconciliation

Corresponding files in the archive do not always share identical names, and some internal metadata disagrees with filenames. Associations were therefore keyed on the four-digit campaign year, which is unique per advertisement and is treated as canonical, following the master campaign analysis. The keyframe metadata (keyframes.json) also records the exact source analysis and video filename for each ad, providing an independent cross-check that resolved every mapping.

A machine-readable and human-readable mapping report is generated on every build (SOURCE_MAPPING.md in the project root). Uncertain or reconciled matches are reported rather than hidden. The notable cases:

  • 2006 and 2008 — the analysis files internally label the campaign year “2005”; the filename years are treated as canonical.
  • 2015 — internally labelled “2014”; filename year treated as canonical.
  • 2019 — internally labelled “2020”; filename year treated as canonical.
  • 2020 (Lambalytica) — the supplied source video was mislabelled 2026 - Happiness Index.mp4, and the analysis heading reads “Happiness Index.” The advertisement is in fact the 2020 Lambalytica campaign (confirmed by its synopsis and the master analysis); the title and year were corrected. The genuine Happiness Index is the 2026 advertisement.
  • 2023 (Un-Australian) — the supplied brief referenced a 2021 title, but the supplied video and transcript are the 2023 Un-Australian advertisement; the analysis follows the supplied evidence.

Each of these is also flagged in the relevant advertisement’s own “Verification and uncertainties” section.

Keyframes

The keyframes were generated beforehand and are not regenerated here. For each advertisement the archive supplies a keyframes.json, a keyframes.csv, the extracted frames and a contact sheet. This archive imports the existing images and uses them for advertisement lead images, the visual timeline, and each ad’s chronological scene gallery.

The archive’s own source .mp4 files are not republished. For the moving image, each advertisement article instead embeds the official version from the campaign’s public YouTube playlist. The embed is privacy-friendly and click-to-load: the local keyframe is shown as the poster, and nothing is requested from YouTube until a visitor chooses to play. Video-to-year mapping was confirmed against each video’s title (retrieved from YouTube’s oEmbed endpoint), which matched all 22 years unambiguously — including the YouTube title “2020 Lambalytica”, which independently corroborates the reconciliation described above.

One stable representative lead image is chosen per advertisement by a deterministic preference over the keyframe metadata’s shot types. Contact sheets are retained for reference on the source library but are never used as lead images, and the full set of frames is presented as a captioned, chronological gallery rather than an unexplained grid. Each gallery frame carries a concise description and, where the metadata supports it, a timestamp. Every extracted frame in the archive was imported; all carried a “high” confidence rating in the source metadata and none was flagged for manual review. Live import figures are shown below.

Content model

Advertisement articles are generated from the analyses and keyframe metadata by a repeatable import script, and are kept separate from hand-editable content so that regeneration never silently overwrites editorial work. The eras, themes, targets of satire, humour techniques and character articles are hand-authored and grounded in the same sources, with cross-links wired from a single curated configuration. The relationships between advertisements and these articles are what make the site behave like a knowledge base rather than a set of independent essays.

What this archive does not claim

It does not assert release dates beyond what the analyses state; it does not identify performers or credits the sources do not name; it does not resolve the campaign’s unresolved tensions (the vegan contradiction, the First Nations gap, the sincerity of the early satire); and it does not treat the later synthesis as fact about the past. The reader is always pointed back to the primary sources.

Import figures

22Advertisements
198Keyframes imported
14Reconciled / flagged
0Build warnings

Generated 2026-07-10 by the import pipeline. The full machine- and human-readable report is written to SOURCE_MAPPING.md in the project root.

Reconciled and flagged mappings

Every advertisement was matched to its analysis, keyframes, transcript and subtitles by campaign year. The following required reconciliation or carry documentary uncertainty:

2006 — Address to the Nation
  • Internal metadata year: 2005 (canonical from filename: 2006).
  • The analysis file's internal metadata labels the campaign year as 2005; the archive filename (2006) is treated as canonical, following the master campaign analysis.
2007 — Address to the Nation (Australia Day Party)
  • Source analysis H1 work title “Address to the Nation” differs from the canonical title “Address to the Nation (Australia Day Party)”.
2008 — Australia Day Address
  • Internal metadata year: 2005 (canonical from filename: 2008).
  • The analysis file's internal metadata labels the campaign year as 2005; the archive filename (2008) is treated as canonical. The master analysis flags the topical context as most consistent with 2007 (Sydney APEC) rather than 2004.
  • The analysis records a Helen Clark gender joke; the Lamb Brain guardrails classify this as a deprecated execution.
2009 — Address to the Nation (GFC Rally)
  • Source analysis H1 work title “Address to the Nation” differs from the canonical title “Address to the Nation (GFC Rally)”.
2010 — Sam Kekovich UN Speech
  • The master analysis flags some place-name puns in this ad (e.g. on Middle Eastern cities) as sensitive in retrospect; the Lamb Brain treats them as deprecated executions.
2014 — Generation Lamb
  • The master analysis treats 2014 as the first year of Era 2 (Transition and Expansion) and simultaneously as a bridge from the Kekovich desk era: Kekovich announces a semi-retirement here while the ensemble format has not yet fully taken over.
2015 — You Never Lamb Alone (Richie's BBQ)
  • Internal metadata year: 2014 (canonical from filename: 2015).
  • Source analysis H1 work title “Richie's BBQ” differs from the canonical title “You Never Lamb Alone (Richie's BBQ)”.
  • The analysis file's internal metadata labels the campaign year as 2014; the archive filename (2015) is treated as canonical.
  • The analysis notes the ad was criticised for lacking Indigenous representation.
2016 — Operation Boomerang
  • The structural exclusion of the vegan character is the corpus's sharpest counterexample to the 'food unites everyone' theme, and was noted as controversial.
2017 — Gods (You Never Lamb Alone)
  • Source analysis H1 work title “You Never Lamb Alone” differs from the canonical title “Gods (You Never Lamb Alone)”.
  • A spring release rather than an Australia Day release, per the analysis.
2019 — Unite Australia and New Zealand
  • Internal metadata year: 2020 (canonical from filename: 2019).
  • The analysis file's internal metadata labels the campaign year as 2020; the archive filename (2019) is treated as canonical, following the master campaign analysis.
  • This is the only advertisement in the corpus to directly reference the 'Change the Date' debate.
2020 — Lambalytica
  • Source analysis H1 work title “Happiness Index” differs from the canonical title “Lambalytica”.
  • Source reconciliation: the supplied source video was mislabeled '2026 - Happiness Index.mp4', and the analysis file's heading reads 'Happiness Index'. The advertisement is in fact the 2020 'Lambalytica' campaign, confirmed by its synopsis and by the master campaign analysis; the canonical title and year have been corrected accordingly. The 2026 advertisement is the genuine 'Happiness Index'.
2023 — Un-Australian
  • Source reconciliation: the analysis file notes the supplied brief referenced the 2021 'Make Lamb, Not Walls' ad, but the supplied video and transcript are the 2023 'Un-Australian' advertisement. The analysis (and this article) follow the supplied evidence, treating this as the 2023 ad.
2024 — The Generation Gap
  • This is the only advertisement with a machine-readable keyframe collection report in the archive (keyframes_output/collection_report.md).
2026 — Happiness Index
  • The first advertisement in the corpus to prominently display the Aboriginal and Torres Strait Islander flags in Kekovich's office. First Nations peoples remain absent as protagonists across the corpus.