← PublicationsMethodA reproducible measure of platform extraction
GET /v1/publications/reproducible-measure-of-platform-extraction
kind Method
published 2026-02-20
board /v1/boards/platform-capitalism-audited
author Naoko Yamamoto
cite_as Xooplab (2026). "A reproducible measure of platform extraction." xooplab.com/publications/reproducible-measure-of-platform-extraction
Machine abstract · key claims
- Extraction ratio = platform-side value capture / gross transaction value, sourced from filings + public datasets.
- We publish the formula, inputs, and a cell-by-cell working notebook so anyone can fork and re-run.
- Across three sample platforms the spread between highest and lowest extraction was ~6×, larger than expected.
- Public filings undercount, so the score is a floor — not a verdict.
Most platform-economics writing reaches a conclusion before it touches a spreadsheet. We're trying the other direction: an open scoring function applied to three two-sided platforms with very different public profiles, and a method anyone can re-run.
The score
Extraction ratio = (platform-side value capture) / (gross transaction value), measured from each platform's published filings and supplemented by public datasets where available. We publish the formula, the inputs, and the cell-by-cell working in a notebook anyone can fork.
Why the obvious objections are the right objections
- "Public filings undercount." Yes — we say so, and treat the score as a floor.
- "Cross-platform comparison is unfair." Often, yes. We score by quarter and within sector first.
- "This will be gamed." The numerator is hard to game; the denominator is hard to inflate without lying to investors.
First three platforms
Numbers and methodology in the board. The ranking is less interesting than the spread: the highest-extraction platform in our sample is roughly 6× the lowest, which is bigger than any of us expected before we ran the numbers.