GHAN research
Research and written standards, published so they can be quoted and checked rather than summarised. Each piece states its methodology and says plainly where GHAN has data and where it does not.
Published
- Desktop install safety and fraud — GHAN's standard — is installer bundling safe, the prohibited-behaviour list, the definition of a cleared install, and all fourteen rejection and reversal conditions.
- Desktop software CPI benchmarks — what cost per install means for desktop software, GHAN's published rates by category and platform, and the methodology for the observed figures that follow.
Standing methodology
- No figure is published that cannot be regenerated from the ledger or the fraud flag table.
- Absence is stated, not estimated. Where GHAN has no data, the page says so and gives the date the data will exist.
- Definitions come before numbers, because most disagreements about install metrics are disagreements about what counts as an install.
- Every published aggregate has a per-app view that the affected partner can check against signatures it produced itself. See transparency.
Questions people ask about this
Does GHAN publish original data?
It will, generated from the fraud flag table and the ledger rather than written by hand, the month the first chains clear. Until then the research here is definitional and methodological - the safety standard, the rejection conditions, the published rate card - and it says explicitly where a figure does not exist yet rather than estimating one.
Machine-readable versions of this page: markdown · llms.txt · llms-full.txt · OpenAPI · AI catalog