Open data · CC BY 4.0

SignalPipe buying-intent corpus

An anonymized dataset of buying-intent signals from public Reddit, Hacker News, and RSS feeds, scored by the SignalPipe pipeline. Free to download, free to cite, free to train on.

What’s in it

·Hashed post identifiers — SHA-256 of the original URL. No raw URLs, no titles, no body text.
·Source platform — rss / reddit / hn (the subreddit / feed name is preserved as a coarse category).
·All five component scores — embedding match, keyword density, freshness, engagement, author reputation — each 0–100.
·content_score and signal_score — pre- and post-floor.
·Assigned role — closer / advisor / educator.
·Operator outcome (where available) — approved / rejected / sent / replied / ignored, with the rejection reason if rejected.

What’s NOT in it

· No raw post text, titles, or URLs.
· No author handles, follower counts, or any field that could re-identify a poster.
· No operator-level data (which SignalPipe customer scored which lead).
· No anchor sentences or product configurations.

Get the dataset

The first release is being prepared for upload to Zenodo with a permanent DOI. To be notified when it goes live, join the waitlist and select “dataset” as your interest, or email hello@signalpipe.io.

Notify me when available

Citation

SignalPipe (2026). SignalPipe Buying-Intent Corpus.
Released under CC BY 4.0. https://signalpipe.io/dataset
DOI: pending Zenodo upload

License

Released under Creative Commons Attribution 4.0 International (CC BY 4.0). You can copy, redistribute, remix, and build upon the material for any purpose, including commercial — as long as you give appropriate credit and indicate if changes were made.

See methodology for how the scores are computed, or /signals for the weekly aggregate trend digest.