Detectors

the detector object

Every detector is one record with the same fields.

This is the shape a detector takes in the library and in code. One detector watches one source for one trigger. A source can hold many detectors; a trigger can have many detectors; a playbook can be fed by many triggers.

one detector, as data

{
 "id": "det.tried-it-and-it-failed.uneed",
 "version": 1,
 "trigger": "Tried it and it failed",
 "side": "cold",
 "source": "uneed",
 "kind": "threshold",
 "match": "launched in the last 7 days with upvotes < 10; a paid Spotlight slot makes it stronger",
 "strength": "strong",
 "fresh_for": "7 days",
 "exclude": [
 "agencies, templates, non-software listings",
 "ICP not-if lines",
 "lead rules: never customers, competitors"
 ],
 "emits": [
 "detector_id",
 "link",
 "date",
 "owner",
 "field_values"
 ],
 "feeds": [
 "signal-based-outbound",
 "lead-to-waitlist queue"
 ],
 "yield_target": "20+ a month"
}

the library

Every detector, as an object.

Placeholders in braces come from the brief and ICP: {job} is what the product does, {category} is what buyers call it, {competitor} comes from the competitive analysis, {role} is whoever does the job by hand.

Communities and search

Reddit — posts, comments · 8 detectors

X — posts · 12 detectors

Threads — posts · 2 detectors

Hacker News — posts, comments, Show HN · 3 detectors

Indie Hackers — posts · 3 detectors

Discord, Slack and Skool — messages, posts · 3 detectors

LinkedIn — posts, profiles · 1 detector

Google Alerts — web pages · 1 detector

Launch sites

Product Hunt — launches · 2 detectors

Uneed — launches · 1 detector

PeerPush — launches · 1 detector

BetaList — launches · 1 detector

Other launch sites — launches · 2 detectors

Review sites

G2 — reviews · 2 detectors

Capterra — reviews · 1 detector

Trustpilot — reviews · 1 detector

App stores

Apple App Store — reviews, listings · 2 detectors

Google Play — reviews, listings · 1 detector

Shopify App Store — reviews, listings · 2 detectors

Chrome Web Store — reviews, listings · 1 detector

Records

Revenue boards — records · 3 detectors

Job boards — job posts · 1 detector

News and funding — articles · 2 detectors

Meta Ad Library — ads · 2 detectors

Competitor pages — pages · 2 detectors

Your own data

Replies and mentions — your posts · 3 detectors

Forms and email list — submissions · 2 detectors

Site analytics — page views · 1 detector

Your app — product events · 5 detectors

Stripe — webhooks · 3 detectors

source directory

Where detectors read, and how you're allowed to.

Access changes, so check each source's current terms before automating it. "By hand" means a person reads it on a schedule; that's often the right start anyway, because it's how you learn what a real match looks like.

before you trust it

Test every detector by hand first.

A detector that hasn't been tested on real data is a guess. Four steps, about an hour each, before it runs on its own.

step 1 — Search it yourself · Run the match by hand on the source for the last 30 days. Use the site's own search or a site: search.

step 2 — Count the real matches · Twenty or more a month makes it a channel. Fewer than that, and each person gets sold to by hand, which is fine but doesn't need automating.

step 3 — Read 20 of them · Mark each one: right person, in the moment, fits the ICP. If fewer than half pass, tighten the match or add exclusions before going further. (A rule of thumb, not a standard.)

step 4 — Write down the misses · Every false match you saw becomes an exclusion. Every real person you found another way, which it missed, becomes a new phrase.

Strength, in plain terms

After it's running