The intuition: "If they pay for X, they probably make real money." Correct — but only for the right X.

The precise rule is not "pays for software → has money." It's:

That distinction is the whole game. Linktree fails the test — it's the flagship example but it's actually a weak signal. Broke hobbyists buy Linktree Pro constantly; there's no $90 tier (top is ~$24/mo), and the free tier is everywhere. It tells you almost nothing about revenue.

The strong signals are tools that are one of:

The lettered sections (6A–14C) extend the same rule beyond purchases: money leaves public exhaust — ad libraries, ticket prices, review velocity, raised dollars, credited work. Same bar applies: it only counts if it's pointless or impossible below a revenue line.

Everything below is scored against that rule.

The bio link is the single most-scrapable artifact a music business has. The tool they use, and especially the smart-link domain, leaks their tier. (Full domain→distributor map in §15 — that's the centerpiece.)

Who distributes an artist is a graded ladder. The bottom is pay-to-play (anyone). The top is invite-only (you're already making money). And it's detectable because the distributor stamps its name in public metadata.

Paying to collect publishing money means there's publishing money to collect.

Everyone in this category screams money. The seats themselves live behind logins — but they leak through job descriptions: any org hiring names its stack ("proficiency with Chartmetric required"). Detection = JD text mining (§10A). Honest coverage note: this only catches orgs actively hiring — an enrichment pass, never a primary sweep.

(Spotify/Apple for Artists: free, not a signal — deliberately absent from the catalog.)

Free tiers are noise. But SMS platforms and big email lists are enterprise-priced because they scale with a real fanbase.

If they're running a store, there's money moving through it. Detectability here is excellent (tech fingerprints).

A music business paying to run Meta ads has a marketing budget and usually the sales to justify it — nobody sustains ad spend that isn't working. Better still, Meta runs a public transparency tool (the Ad Library), so this is one of the very few "spends real money" signals that's genuinely scrapable. It's also a pre-qualifier for our own upsell: an advertiser is already sold on paying for growth.

The signal is not "runs an ad" — a hobbyist boosts one post for $6 (the Linktree trap again). The signal is sustained, multi-creative, conversion-driven advertising.

Spend-proxy tiers (how we bucket without a dollar figure):

How we detect it at scale:

Killer stack: sustained Meta ads (§6A) + Shopify (§6) + a distributor smart-link (§15) = a music business that sells product, spends to acquire customers, and is signed to a real distributor. Top-decile lead.

Same logic as §6A, different transparency surfaces. The EU's DSA forced every major platform to run a public ad repository; coverage varies by region, but presence in ANY library = active ad budget.

(Snapchat US commercial ads and Spotify Ad Studio / SoundCloud Promote spend: no transparency surface → Appendix A.)

Build note: same tiering as §6A (dabbler / active / serious) — count creatives × duration across libraries. A business active in two+ ad libraries simultaneously (e.g. Meta + Google) is a confirmed serious spender, no further proof needed.

Tour-management software is useless until you have a crew, dates, and routing. Its presence = you tour = you make money.

Tour money is computable from fully public parts. This is the closest thing to reading their P&L off the street.

For the producer segment, a stocked, Pro-tier store is the signal — not a free profile.

(Splice/LANDR subscriptions: invisible AND ⚪ weak — Appendix A.)

Service marketplaces publish both the rate and the completed-work count. reviews × listed price = a hard, computable revenue floor (reviews undercount orders, so it's conservative). This is the single best signal for the engineer/producer segment.

A producer running a course or coaching funnel is a producer making money — and the platforms are detectable.

Everywhere else we infer tiers. Crowdfunding pages print the actual number, publicly, forever.

Booking, contracts, invoicing, payroll, accounting = a business with revenue to manage.

Nobody hires without revenue. Bonus: this is where all the "invisible" software from §4/§12 leaks — job descriptions name the stack ("experience with Chartmetric/DISCO/Curve required").

The studio segment's strongest public exhaust: a commercial space generates reviews, listings, and rates.

The tier of the DAW/plugin stack, and premium formats, separate hobby from commercial.

Retainers and pro catalog tools = agency or funded artist.

(Highnote/Byta: private-by-design sends, zero public artifact — Appendix A.)

Not software — but the highest-trust public proof of professional income, all programmatically searchable. Doubles as the verification layer for self-claimed bios ("platinum producer" → check the DB).

Taking an advance requires provable royalty income — the strongest possible money proof. Quiet deals are invisible (Appendix A), but two surfaces ARE public: deal announcements and auction listings.

Protecting a brand costs money and signals they expect it to be worth protecting.

Everything here comes from one crawl of their domain + DNS. Cheap to run over an entire lead list.

Not purchases — public statuses that either require a distributor/label deal or mark a crossed monetization threshold.

Filtering out no-money accounts is worth as much as finding money. Run this pass LAST over every candidate list.

This is the best signal we have because it is both publicly visible in the bio and reveals who distributes them (= their tier = their money). One scrape yields both "here's a music business" and "here's how big they are."

The strongest money signals hide behind logins — but most of them leak anyway. The v1 draft wrote off Chartmetric seats, Kobalt admin, QuickBooks payroll, and DISCO as "invisible." The rigor pass found programmatic leak paths for nearly all of them:

The catalog now contains only programmatically detectable signals: 🟢 = direct fingerprint, 🟡 = multi-step or sparse leak path. The handful with genuinely zero public artifact live in Appendix A, banned from the pipeline so no build-hour is ever spent on them.

At-scale discipline: 🟢 direct fingerprints are primary sweeps over the whole candidate pool; 🟡 leak paths are enrichment passes on survivors only (see the funnel corollary at the top).

Grouped by method, because Block #2 is built method-by-method, not signal-by-signal.

The only signals that go into Block #2's first scraper. Ranked by ROI = (money strength × detectability × how much it narrows to our customer).

Then the §14C negative-signal pass runs over everything — free-tier subdomains, dormant catalogs, dead funnels, bought audiences get culled before a human ever sees the list.

Not in the top-16 but still in the catalog as enrichment passes (programmatic, just multi-step/sparse — run on funnel survivors, not the whole pool): PRO-repertory admin records (Songtrust/Kobalt/Sentric), JD mining for pro stacks (Chartmetric, DISCO, QuickBooks, Master Tour), disco.ac/f.io share-link crawls, gear-list parsing, Royalty Exchange auction earnings, announced advances. The truly-undetectable residue lives in Appendix A and stays out of the pipeline entirely.

And §19 is the calibration layer: the Burn County archetype (money exhaust × presentation debt) converts these raw signals into the profile of the one lead that actually closed.

Why this section exists. Ty Weathers (Burn County / Sand City Studio, North Myrtle Beach SC) went from IG-discovered lead → paying Studio Raine client → the source of the RIAA-Gold and Billboard credentials now displayed on studioraine.art. He is the one fully-closed loop we have: every signal below was ON him at lead time, so this is the empirically-validated fingerprint — not theory. Everything here was live-verified July 24, 2026 (raw fetches of linktr.ee/burncounty, all four of his domains, DNS/MX lookups, revenue-link status checks) and cross-checked against the July 22 deep-research audits in research/.

The two-axis insight this client proves: the lead is not "has money" alone — it's "has money AND has a visible, provable presentation gap." Ty was Top 1% on paper and invisible online. Every gap below is programmatically detectable, which means the outreach hook itself is scrapable (we opened with his own broken funnel).

GoHighLevel ($97–$497/mo) is the highest-value single software signal this client validated, and it's detectable three ways:

These cannot be programmatically detected by any known path. Listed only so nobody re-adds them or burns a build-hour trying. If a leak path ever appears (a transparency law, a public library, a share-link format), they graduate back into the catalog.

Every row that reaches John's call list must clear both:

The phone isn't just a contact detail — a business that publishes a phone number for calls is self-selecting as easy to sell to. If they take calls publicly, they take sales calls. That makes "how the phone is exposed" a qualifying signal, not just a lookup:

Two routing facts that matter for John's outreach:

Across this niche the contact chain is nearly always the same, and every hop is machine-followable:

Both things John needs — the IG handle and the phone — fall out of walking this one chain. The system below is just that walk, hardened, deduped, validated, and run in batch.

Input is the money-qualified candidate pool from Block #1 (the §17 scraping bridge output / §19 Burn County archetype). Output is a complete, deduped call list. Stages run as a funnel — cheap steps on everyone, expensive steps only on survivors (same discipline as softwaresignals.md).

Every candidate must be reduced to two anchors: an Instagram handle and a website domain. Where each starts depends on how it entered from Block #1:

Join key for dedup: normalized domain first, else IG handle. The same operator often appears via IG and Maps and domain (Burn County was on all three) — collapse to one record here so John never sees a dupe.

IG-handle and domain are effectively free to carry. The paid enrichment (profile scrape, Maps place, site crawl) happens in Stage 2's waterfall and is ordered cheapest-first so most leads resolve before the expensive calls fire.

Try sources top-to-bottom; first valid hit wins, but capture every source that returns a number so we can cross-validate (agreement across two sources = high confidence). All are batchable at 100+/day.

Coverage reality (honest): no single source covers everyone. IG-phone only exists for professional accounts with a contact button; Maps-phone only for businesses with a listing; website-phone only if they list one. That's why it's a waterfall — the union of sources is what gets us to near-100% on the archetype. For the Burn County profile specifically, sources 1, 3, and 4 all hit (IG business acct, GHL site tel: button, Linktree). Redundancy is the norm for this niche, not the exception.

IG is required output, and in most paths (§5.1) we already have the handle. Confirm it's usable with one profile scrape (reused from Stage 2 #1):

John sells website creation, so a great site disqualifies. Score website brokenness; pass = broken/weak/absent. This reuses the §19.2 problem-signal detectors, as a machine rubric:

The non-negotiable filter, run last before validation: DROP any record missing phone OR IG. Only complete {phone, IG} rows survive. This is what guarantees John's promise to himself — every single lead on the list has both.

All confirmed live against the Apify store (July 24, 2026), with real fields/prices. We already have the apify MCP wired in.

Bad numbers waste John's call time, so every surviving number is cleaned:

The studio segment is the beachhead for contactability: Maps gives phone+site+IG together, and local studios have the most broken websites — both gates clear at once.

Not every money-qualified candidate clears both gates, so we over-feed the funnel. Rough, honest math (tune with real run data):

The bottleneck is Block #1 candidate supply, not contactability tooling — the actors easily handle thousands/day. So Block #3 (the actual scraper) should target ≥250 archetype candidates/day to keep this funnel full.

Every row is complete by construction (both gates passed). This is the deliverable format:

The last two columns are why this system compounds: the outreach hook is carried straight from Block #1's problem signals (the 404, the missing gold record) — so John opens every call already knowing the lead's exact wound.

The system John asked me to create, in one paragraph: take Block #1's money-qualified candidates, resolve each to an Instagram handle + a website domain, then run a phone-acquisition waterfall (IG professional-account phone → Google Business phone → website tel:/contact scrape → Linktree tile → Facebook → WHOIS), keeping the first valid number and cross-validating any others. Confirm the IG is live and public. Gate on serveability (website must be broken/weak/absent — John sells websites) and then on contactability (must have phone AND IG). Validate/normalize the number, tag it Call vs Text, dedupe, and emit a row carrying the phone, the IG, the money signals, and the outreach hook. It runs as a cheap-first funnel on ~250 candidates/day to net 100 complete leads for ~$3–7/day, using the verified apify actors in §6 (logical_scrapers/instagram-profile-scraper for IG phone, lukaskrivka/google-maps-with-contact-details for studios, purple_beep_boop/bulk-website-contact-scraper for the site hop). The design deliberately fuses the two gates: the broken website that qualifies them for John's #1 service is the same artifact that hands us their phone number.