TWCPG Archive — the deal-flow corpus

A second dataset, complementary to the 419 curated breakout-list brands: the full structured deal flow underneath. We extracted every M&A and fundraising event from the This Week in CPG newsletter archive (2022-09-24 → 2026-05-07), then enriched each new brand to the same standard as the original cohort — founded year, ownership, revenue, acquirer, sources cited.

The merged dataset is 2,274 brands, 99.7% web-research-enriched. This page documents what's in the deal-flow corpus and how to use it.

If you want to slice it yourself, head to Explore. If you want every brand, Brands is the canonical merged table.

The corpus, in numbers

The TWCPG-corpus brands plus the 362 breakout-list-only brands make up the 2,274-brand merged dataset on Brands. All numbers below are raw counts from the corpus — interpret them as you see fit.

Top acquirers

By deal count across the full corpus:

Top lead investors

By rounds led:

How the dataset was built

Three extraction phases against the newsletter archive, then a merged enrichment pass against the same WebSearch-backed worker that produced the original 419-brand cohort:

1. Content sample (May 2026). 24 stratified posts → structural taxonomy of how the newsletter formats its bullets.
2. Full-archive extraction (May 2026). Per-post haiku call with bullet-level validation across 413 posts. Output: structured M&A + fundraising bullets per post.
3. Bullet enrichment (May 2026). Re-parse every deal bullet to extract founded_year, amount_usd, round_stage, normalized acquirer/investor names. Output: data/twcpg/enriched.merged.cleaned.json.
4. Merged brand enrichment (May 2026). The 1,857 brands surfaced by TWCPG that weren't in the original cohort were seeded as records in data/brands.json and run through the same enrich_brand worker that produced the original 419-brand enrichment — full WebSearch, full source citation per field. This brings the entire merged dataset to a single enrichment standard.

Specs live under docs/superpowers/ in the project repo. The original 419-brand pipeline is documented on Loop.

Caveats

• Corpus boundary. Starts 2022-09-24. Pre-2022 fundraising rounds for older brands aren't captured. Any analysis indexed on "first funding event in corpus" is biased late for any brand active before that date.
• Editorial coverage. TWCPG covers what the newsletter chose to cover. Notable deals get bullets; small/quick acquisitions and undisclosed-amount deals sometimes don't. The merged-with-brands.json view (which has research-backed acquired_by on every brand) is materially more complete than any TWCPG-only cut.
• Stage labels. "Equity" is a catch-all for funding rounds where the original bullet didn't name a series. ≈40% of fundraising events fall here, which weakens per-stage analysis.
• Attribution. All bullet text originates from This Week in CPG by Anson Frericks. The structured extractions and enrichments are derived data; the underlying observations belong to the source.

Downloads

The full per-event corpus and the derived per-brand index are downloadable as JSON. If you want to use this data in your own analysis or build on top of it, these are the canonical files:

• twcpg-enriched.json (2.8 MB) — full per-event corpus with founding years, amounts, lead investors, acquirers
• twcpg-brand-index.json (1.2 MB) — per-brand event arcs, top-investor and top-acquirer rollups
• twcpg-enrichment-stats.json (29 KB) — derived stats substrate

For the full merged brand database (TWCPG corpus joined with breakout-list cohort), the canonical view is Brands. For interactive pivoting with saved views, use Explore.

If you find something interesting in the data, ping [email protected].