Data published by Answering Service Picker

1 dataset, each derived from named sources' own pages and re-read on a cadence, latest 2026-10-21.

How these are made

Every dataset starts as a charter that names the buyer's question and the sources by URL. Our fetch service reads each source's own page, and the sentence carrying the figure is kept exactly as printed, with the page and the day it was read.

A reviewer accepts or refuses every candidate sentence with the reason recorded, and the derivation is computed over one figure per source, or one for each breakdown a source prices. A source read in full that publishes nothing is counted as not published, never filled in. Each dataset is re-read on its cadence, so the next read reports what moved.