Research

Every study on this page starts from a sample frame fixed BEFORE collection, not selected afterward to fit a conclusion. The Reddit Community Access Index froze its 24-subreddit list, then pulled the community profile, rules, and top-post sample for each one through FORKOFF's own Reddit data pipeline. The X Post Format Benchmark froze a 17-account sample frame across developer tools, AI/ML products, SaaS, and developer-audience media, then pulled each account's public timeline through FORKOFF's own X data pipeline, filtered to original posts only.

Exclusions are disclosed rather than silently dropped. When an account in the X sample returned not-found, it stayed excluded rather than being substituted with a friendlier replacement. When an account posted only reposts inside the measurement window, it contributed zero rows to the post count rather than being backfilled. The same discipline applies to the RADAR-derived studies: every tracked launch, organic and amplified, is kept in the corpus, because dropping the amplified ones would flatter the "organic reach" conclusion the study is measuring in the first place.

Where a study classifies rather than just counts, the deciding evidence is published beside the classification. The Reddit index labels each community's promotion posture (open, conditional, or banned) from that community's own rule text, and the specific excerpt that drove the call ships in the same dataset row, so a reader can check the classification rather than take it on trust.

Every number in the Reddit Community Access Index and the X Post Format Benchmark is public platform data about public communities and public accounts, measured on a stated pull date. No client result, campaign figure, or internal engagement metric appears in either study or its downloadable dataset. That is a deliberate design choice: a private number cannot be re-run by anyone, and the value of a published benchmark is that a reader can point the same public endpoints at the same accounts or communities and get the same answer.

Both datasets ship as a full CSV under CC BY 4.0, linked from each study's methodology section, free to reuse with attribution. The ASVC Discovery Gap dataset (branded-query vs head-term citation rates across 4 AI engines) is downloadable on the same terms. Where a study cites an outside figure for comparison, industry CPQV ranges, platform view-counting policy, bot-traffic estimates, that source is named inline at the point of use rather than folded into the first-party numbers.

Three update patterns run across the catalog above, and each study states which one it follows. State of Launch Videos and the Views-to-Likes Benchmark recompute live from the public RADAR launch corpus on every page load, so the 30-launch figures on this page and on /radar are always the same run rather than a cached snapshot. The AI Citation Index re-runs on a fixed quarterly cadence against the same 50-prompt cluster and the same 5 engines, so a reader can compare cite-rate movement quarter over quarter rather than across two differently-built runs.

The Reddit Community Access Index, the X Post Format Benchmark, the Clipping CPQV Benchmark, and the ASVC Discovery Gap are dated point-in-time snapshots. Each carries its own pull or publish date on its individual page, and gets re-pulled when the underlying category has moved enough to warrant a fresh measurement rather than on a fixed clock, since a subreddit's promotion rules or a platform's view-counting policy do not change on a predictable schedule.

Each individual study carries its own stable APA and BibTeX citation block (see the "Cite this hub" section on that study's page), with the publish date, last-updated date, and canonical URL a journalist, analyst, or AI answer engine can cite directly. At the catalog level, this page is the stable index: if you are citing FORKOFF's research program as a whole rather than one specific study, cite this page (forkoff.xyz/research) and name the individual study you are drawing a figure from.