The Outbound Machine
One system: 30,000 HVAC owners → conversations → triage calls → partners. This page is the whole machine at a glance. Skim the headlines and you have the system; expand anything for depth. Chris presents from it live — then it's yours to come back to.
Ten clients by October 31. The constraint we're fixing is zero market contact.
This machine is the spears leg of the Category King campaign — a 90-day sprint (July → Oct 31, 2026) to put "Revenue Engine" in thousands of HVAC owners' heads and sign 10 clients by October 31, with 2–3 in the first 30 days for runway.
Why outbound, why now. The July 2026 constraint review named the bottleneck plainly: the company has been polishing instead of contacting — zero market contact. Predictable Revenue's own math says outbound works profitably only above $10k–$20k lifetime value; at ~$15k upfront and $50k–$100k LTV we clear that bar 2.5–5x. A downsell ladder ($8k · $5k · $500/mo) catches the rest — we'll happily close a smaller shop that fits the $5k or $8k tier.
The locked category line. AnchorWorks is not an agency and never sounds like one — every touch this machine sends obeys that.
Where this knowledge came from.
Nothing on this page is opinion. A fleet of 17 research agents ran ~401 primary-source web lookups across 7 research lanes + 2 gap lanes — then every load-bearing claim was independently re-verified by adversarial fact-checkers, and the execution plan itself was torn apart by 3 more critics before v1.1 shipped.
A fabricated case study
A widely-circulated "HVAC 2.1%→8.7% reply rate" case study could not be verified anywhere. The fact-check pass killed it — it appears nowhere in our numbers.
A famous misattribution
The template everyone quotes as "the Predictable Revenue email" is actually Bryan Kreuzberger's Breakthrough Email — a descendant, not Ross's original. We teach the verified mechanics instead.
Stale prices
Fact-checkers caught and re-priced stale 2024-era vendor figures. Every provider price in the plan is a July-2026 published number.
Mission
Extract the first principles from Aaron Ross's Predictable Revenue — what to copy, what aged badly, and whether the economics validate this motion at all.
Headline finding
Copy the system, not the numbers. The $10k–$20k+ LTV viability bar is cleared 2.5–5x. But the 8–12% reply era is now 0.45–5%, and "50–100 emails/day per rep" is a deliverability suicide note — our locked 30/inbox ramp is the correct 2026 translation.
Mission
Verify the locked send numbers against 2026 deliverability reality — Google/Yahoo/Microsoft mandates, warmup pools, tracking, sequence architecture.
Headline finding
The locked ramp sits dead-center of 2026 consensus (30/inbox is the middle of the 15–50 band). Free wins: open tracking OFF + plain-text ON = 1.3–2x more replies (Reply.io's own data). First email earns ~58% of replies; ~93% of achievable replies arrive by day 10.
Mission
Design the Clay enrichment waterfall for small US HVAC owners — providers, order, hit rates, real per-record cost at each tier.
Headline finding
Go public-records-first, LinkedIn last. License boards + a local-business specialist (Openmart, ~$0.01–0.06/rec) beat generic B2B vendors 5–50x on cost. Base tier lands at $0.10–0.60/record — BYOK keys are the 6x lever. Catch-alls bounce 27% vs 1% for valid verdicts: hence the two-layer verifier.
Mission
Verify LinkedIn per-account limits, safe tooling, warmup ramps, and how to wire the locked "24h no-reply → LinkedIn" trigger.
Headline finding
LinkedIn is a booster on ~a third of the list, not a second channel. The ~100 invites/week cap is confirmed into 2026; browser extensions carry ~60% higher detection risk than cloud tools; owner profiles findable for maybe 40–60%, genuinely active 10–25%. No-note invites accept best.
Mission
Map exactly what the reply-gated SMS touch must contain to be clean under TCPA, the state mini-TCPAs, and A2P 10DLC in 2026.
Headline finding
The reply-gate is necessary but not sufficient. A positive email reply is consent to email — texting an enriched mobile is the exact fact pattern serial plaintiffs settle at $2,500–$10,000 a pop. 2025 was the worst TCPA litigation year on record (~2,628 cases, +60% YoY). The clean pattern: the owner gives us the number, or gives a written yes to a specific number.
Mission
Reverse-engineer how HVAC-marketing vendors actually win and lose contractors — and find the ground nobody premium holds.
Headline finding
The ground we want is empty. Premium vendors (RYNO/Blue Corona, Scorpion, CAMP) buy trust channels — ACCA, Nexstar, Pantheon booths — and never cold-email owners. The outbound players are the most hated in the industry: the FTC hit HomeAdvisor for $7.2M in redress. ServiceTitan: ~9,500 customers vs a ~117k HVAC TAM — a tag, never a filter.
Mission
Find the cheapest reliable path to 30,000 base-tier-complete records — and how to merge the ~60k Chamber dataset without polluting it.
Headline finding
The data play of the whole project: state license boards. TX, FL, and VA publish bulk license files with the owner/qualifier's legal name — downloadable same-day, ~free. Coverage reality: 169,827 US HVAC businesses, 40.3% with websites, 77.1% with a Google Business Profile. Chamber merge: one-time local DuckDB script, zero Clay credits.
Mission (gap lane)
Can we tag ServiceTitan usage and ad presence across all 30k without burning the credit discipline?
Headline finding
Both tags are buildable for roughly $450–900 total. ServiceTitan: regex the scheduler embed — near-zero false positives, but only 40–60% recall, so "no fingerprint" means unknown, never "no." Meta's official Ad Library API excludes US commercial ads entirely — third-party scrapers do it for ~$60.
Mission (gap lane)
HVAC owners live on Facebook — is there a compliant way to reach them there at scale?
Headline finding
No compliant automation path exists. Pages can't initiate conversations; every "FB DM automation" tool puppets a logged-in profile against Meta ToS during the most ban-happy era yet. The workable play: manual, founder-profile, groups-first — ~10–15 Page DMs/day per operator. Now flag #2, tentatively approved.
First principles: copy the system, never the numbers.
Aaron Ross built Salesforce's outbound engine and wrote it down as Cold Calling 2.0 — "prospecting into cold accounts without ever making a cold call." Short referral-seeking emails to a targeted list; scheduled calls only with people who responded. The book validates our whole premise. Its arithmetic is dead.
Seeds · Nets · Spears
Seeds = referrals. Nets = one-to-many marketing. Spears = targeted outbound. This machine is spears — and each funnel gets its own math. The classic error is judging spears by nets metrics.
Specialize by function
Specialization is the #1 sales-productivity multiplier. At our size the four roles are calendar blocks: the machine + Ethan/Les = SDR function; Chris = triage + close; delivery = farmer.
10–20 SQLs / rep / month
A ramped, full-time outbound rep generates 10–20 qualified opportunities per month, 2–4 weeks from first response to qualified opp. That's the output bar we instrument against.
The $10k–$20k bar
"Outbound can work with smaller deals, but not profitably." Below $10k–$20k LTV, kill it. At $15k upfront / $50k–$100k LTV, AnchorWorks clears it 2.5–5x.
4–6 months of patience
"It takes 4–6 months to have consistent, quality leads coming from your outbound team." The #1 named failure mode is quitting at day 30. The machine is judged on this clock — never at week 3.
Owners reply the most
Founders/owners are the highest-replying seniority in cold email: 0.57% vs 0.42% for C-suite (Belkins, 7.5M emails). We are literally targeting the segment that answers.
What aged badly do not copy
- ✕The 8–12% reply era is gone — 2025 averages run 0.45–5% depending on study.
- ✕50–100 emails/day per rep from one domain is now a deliverability suicide note.
- ✕The SDR-factory model — quotas down 29–40% since 2018; meetings-as-KPI produces "high-activity, low-intelligence outbound."
- ✕The famous "who's the right person?" referral email breaks on owner-operators — a 5-truck owner IS the org chart, and the line reads as a mail-merge tell.
What survived we run this
- ✓The physics of the referral email: 2–4 sentences, plain text, one easy-to-answer question, lost-puppy pressure level.
- ✓The owner-adapted ask: "Are you the best person to talk to about this?" / "Want more information?" — same physics, aimed at the business, not the org chart.
- ✓ICP-first list building, qualify-before-pitch calls, seeds/nets/spears separation, the deal-size bar.
- ✓The modern volume translation: 30/inbox × more inboxes and domains — exactly our locked ramp.
- ✓Founder eyes-on is the success factor. Week 1 at 30/day with Chris reading every send is the book's prescription executed literally.
Why this will work: the ground we want is empty.
The premium HVAC vendors — RYNO/Blue Corona, Scorpion, CAMP — win through trust channels: ACCA partnerships, podcasts, Pantheon booths, ServiceTitan certification. They never cold-email owners. The outbound-heavy players poisoned the well: the FTC hit Angi/HomeAdvisor for $7.2M in redress, and the #1 documented sin owners cite is persistence after no. A respectful, founder-signed email in insider language has no premium competitor.
Words that are burned in this market
Angi/HomeAdvisor and the LSA resellers burned these. Using any of them puts us in the mental folder owners reserve for the companies they hate.
Insider vocabulary that lands
"The same call-board math Nexstar coaches preach" reads as inside the trust channels. Trade language — service calls, callback rate — never ROI or synergies.
Owners live on Facebook
HVAC owners run their Page inbox from their phones and barely touch LinkedIn — profiles findable for maybe 40–60% of owners, genuinely active for perhaps 10–25%. The channel mix is built for that reality.
Multichannel = 3–4x
Tight-ICP teams running email + LinkedIn report 3–4x single-channel reply performance. The 24h second touch exists because of this number.
Not selling leads = credibility
The fastest credibility move in this market is explicitly not selling leads. "We do not sell marketing. We build Revenue Engines." does that in one line — followed by something concrete within one sentence.
Run the funnel yourself — at the documented working rates.
- Leads sequenced 4-touch sequence · 5 send-days/week 13,500
- Positive replies × 0.6% pilot 81
- Booked triage × ~55% pilot 45
- Completed triage × ~75% pilot 33
- Strategy sessions × 75% — Chris's number 25
- Clients × 25% working close rate 6.3
Month-1 honesty. At the base rates, cold alone yields ~0.5–1 client in month one — so the 2–3-in-30-days runway target leans on warm doors, not this machine. The checkpoint is written down: if the trailing positive-reply rate is <1.5% at the end of W2, month-1 expectations re-anchor to warm doors while the machine keeps building on its 4–6-month clock.
The 5-star list goes first. The Command Center list is 5-star-heavy and curated — it should beat the base rates the simulator uses.
Multichannel lift. Email + LinkedIn teams report 3–4x single-channel replies; the 24h touch rides on every sequenced lead with a profile.
The downsell ladder. $8k · $5k · $500/mo tiers raise the effective close rate — a shop that can't do $15k can still become a client.
Faster Sept–Oct inbox scaling if placement stays green — the spare domains are locked burn insurance either way.
The Machine, touch by touch.
Email is the only cold touch. Every touch has one goal: a positive reply, earned through curiosity and free value. Click any stage to expand it. First email captures ~58% of replies, follow-ups the other 42% — and ~93% of achievable replies arrive by day 10.
Two qualifiers. Only two. (1) Are they an HVAC company? (2) Can they handle 10–20 jobs/month? Both are confirmed on the triage call — never in the data. Everything else is a TAG, never a filter. No-Facebook is not disqualified. No-ServiceTitan is not disqualified — often a better fit.
The first touch teaches — and every T1 must pass the Stadium Pitch filter: it teaches something the whole stadium of owners needs, it never pitches the 3% who are shopping.
- Closes with a referral-style ask: "Are you the best person to talk to about this?" / "Want more information?" — never a CTA, never links.
- Under 100 words, ≤2 sentences per paragraph, lowercase-casual subject, trade language, geographically specific, seasonally aware.
- This touch alone earns ~58% of all replies.
No reply in 24 hours fires the second channel. The LinkedIn URL is already on the row (base tier, locked) — for non-connections the default is a no-note invite → message-on-accept (flag #8, confirmed at the W1 review). Skip only if the row has no URL.
- Runs through HeyReach cloud (dedicated residential IPs) — never a Chrome-extension executor on a founder profile (flag #5, implemented).
- Rows with no LinkedIn: manual FB Page DM from a founder profile, ~10–15/day per operator, group-context openers, never automated (flag #2 — Chris: "HVAC guys live on Facebook. I love it.").
- Caps are hard: 100–150 LinkedIn actions/week/account.
A "Re:" bump that floats T1 back to the top of the inbox — with one new thing added: a benchmark stat, a nearby-market example, a proof point. Follow-ups collectively earn the other 42% of replies; follow-up #1 alone lifts replies up to ~49%.
A genuine angle change earns a genuine new thread. The angle comes from the row's tags: no-FB-ads ("noticed you're not running ads"), LSA-present, reputation-pain, ServiceTitan-detected, chamber-tenure, license-age ("licensed since 2009"). By day 10, ~93% of achievable replies are in.
A short, human close-out — the exact inverse of the Angi persistence machine owners hate. Every email in the sequence carries the postal footer and the reply-to-opt-out P.S. ("If this isn't relevant, reply 'no thanks' and I won't follow up").
A positive reply is any human reply that engages — questions, "who is this?", "send me info", "not now but…". When unsure, treat as positive and flag for Chris.
- GHL warm tag fires → the lead exits Reply.io, HeyReach, and GHL sequences simultaneously (seed-tested before the first real send).
- Reactive tier-1 enrichment fires now — and only now: owner mobile + personal email, validated through Twilio Lookup before any dial or text.
- Same-business-day response from the row owner. Speed-to-lead is our own doctrine.
- Intent or a business question → booking link. Curious/lukewarm → the free-value asset with a soft booking CTA. When unsure, push booking.
- SMS only with real consent: a number the owner gave us, or a written yes to a specific number — never merely because they replied.
The triage call confirms the only two qualifiers: HVAC company? Can it handle 10–20 jobs/month? Reactive tier-2 fires right before it — one Claygent pass produces the one-screen pre-call card (reviews, ServiceTitan signal, ad presence, LSA).
- 75% of completed triage advances to a strategy session (30–60 min) — where deals close.
- Chris runs strategy calls 1–10 recorded; Ethan takes 11–20. That's how the pitch gets documented before anyone is hired.
30,000 owners. Tags, never gates.
The master list targets 30,000+ HVAC companies — roughly a quarter of the ~117k US TAM — with no gates. One-truck, broke, no-ads startups belong on the list. The whole credit-control play is the enrichment ladder: enrich broad and cheap once, spend real money only as a lead earns it.
Everything the cold machine needs to run — on the row before sequence entry.
- Company name
- Owner first + last name
- Verified email (sourced preferred; patterned flagged)
- LinkedIn URL — required before entry, because the 24h trigger can't find profiles at scale
The ~10% premium data — self-justifying because the lead already raised a hand.
- Owner mobile / personal phone
- Owner personal email
- Twilio Lookup validation ($0.008) before any dial or text
One Claygent pass → the one-screen pre-call card.
- Google + Yelp review counts and stars
- ServiceTitan fingerprint (low-confidence tag, locked)
- Ad presence: Meta, Google, LSA
- Truck / revenue signals where they exist
The anti-pattern this ladder exists to prevent: tagging all 30k on every field upfront = 150k+ lookups and a hemorrhaging Clay bill for data most leads will never need.
State license boards: owner names, ~free.
TX (TDLR), FL (DBPR), and VA (DPOR) publish bulk license files carrying the owner/qualifier's legal name — downloadable same-day, essentially free. That's why LinkedIn-first B2B providers are the last step of this waterfall, not the first. License age is itself an angle: "licensed since 2009."
Catch-all routing — defusing the landmine
The landmine: a patterned (guessed) address on a catch-all domain that "passes" verification. Catch-alls bounce 27% vs 1% for valid verdicts — and many small HVAC shops run catch-all hosting.
- 1
MillionVerifier on every found email ($0.0037). Valid → sequence-ready. Invalid → next waterfall step.
- 2
Catch-all/risky → BounceBan (~$0.0034). Resolved-deliverable → the lower-priority track, capped at ≤4 sends/day/inbox.
- 3
Unresolved catch-all + anything patterned-on-catch-all → hard quarantine: 1 send/domain/day from the sacrificial inbox; re-verify at 30/60/90 days; drop at 90.
- 4
email_source+email_verdictstay visible all the way through Reply.io — locked.
The 500-row pilot — the gate on everything
500 stratified Command Center rows through the full production waterfall, $500 hard cap — stratified ~100 each across TX bulk-file · FL bulk-file · CA portal · NY weak-licensing · NV-or-AZ, so every source type gets tested. A passing gate table is necessary — Chris's sign-off is the freeze.
| Gate | Threshold | If failed |
|---|---|---|
| Owner name found | ≥70% | Add/reorder Openmart–Coldlytics seed layer |
| Name correct-when-found | ≥90% | Source audit; drop the offending provider |
| Sendable email | ≥60% | Add waterfall steps; site-scrape recall check |
| Blended base-tier cost | ≤$0.60 | BYOK more providers; reorder |
| Pilot spend cap | $500 | Stop, review, re-scope |
Re-test rule: re-run only the failed field on the same 500 rows, $250 once. A second fail on the same gate → stop and escalate to Chris.
Thirteen states. Three operators. One master list.
Each operator owns a territory end-to-end — build, enrichment QA, self-checks, and the replies from your rows. Three parallel Claude agent sessions, each against its own staging tables; the shared master sits behind append-with-dedupe, and dedupe is its gatekeeper.
Sun Belt West
Carries the two monster states — CA's portal-join is the hardest pull and Chris is the most agent-fluent. Wave-1 bulk file: TX (TDLR).
Southeast + NY
Includes Pantheon's home state. Wave-1 bulk file: FL (DBPR). All Command Center reply inboxes default to Ethan.
Mid-Atlantic / Midwest
Lighter and more mechanical (roster/GBP states) while Les ramps — his LinkedIn channel goes live ~W6 after profile warmup. Wave-1 bulk file: VA (DPOR).
Anthony
~2h/day of triage calls plus pilot verification calls ("trying to reach the owner — is that still [name]?"). Triage runs 15 minutes, ~3/hour.
15–24 triage/day absorbable
Roughly 10x expected demand at base rates (~1 completed triage/day in October; upside case ~3–4/day). Call capacity never binds — call quality is the constraint that matters.
Standby setters & closers
Three of each on standby. They activate only when the weekly review shows >8 completed triage/day sustained for two consecutive weeks AND delivery confirms capacity. Nobody gets hired before the founder-run motion is documented.
The verified arsenal, and the discipline of the ramp.
Audited live on Jul 7, 2026 — not assumed: 3 sending domains connected in Reply.io (anchorworkshq · getanchorworks · goanchorworks) with 9 warm inboxes, warmup ON, and zero emails ever sent. Five spare domains sit as burn insurance; two (workwithanchorworks + anchorworksteam) start warming now to land ready ~Sept 1.
3 domains · 9 inboxes
anchorworkshq.com · getanchorworks.com · goanchorworks.com — 3 Google Workspace mailboxes each, DNS configured, warmup running, zero send history. The W3–4 inbox target already exists.
2 spares in the oven
workwithanchorworks.com + anchorworksteam.com begin their ~2-month warmup this week — landing exactly when the W9–12 scaling needs them. Three more spares stay as burn insurance.
The protected list
myanchorworks.com (the website) and anchorworkscollective.com (login/ops identity) never enter the cold pool. Ever.
30/inbox is the permanent cold ceiling, including follow-ups — dead-center of the 2026 consensus band (15–50). Gates live between every step: placement >90%, Postmaster complaints <0.1%, trailing reply rate ≥1% (≥2% = mature placement proof). Placement fail (<90%): hold volume flat, fix, re-test in 3 days — two consecutive fails on one domain → rotate it out and tell Chris.
Standing rules at every stage: plain text · no links, images, or attachments in touch 1 · open + link tracking OFF (no-tracking senders get 1.3–2x more replies) · sentence-level spintax, never word-level · CAN-SPAM floor in every email — the postal footer plus the reply-to-opt-out P.S. · warmup never turns off · @gmail.com recipients tracked as the monthly canary segment.
The rails aren't red tape. They protect the pipeline.
2025 was the worst TCPA litigation year on record, with serial plaintiffs specifically hunting small senders. Every rail below exists so this machine — and the $50k–$100k LTV pipeline it feeds — can never be taken away from us.
SMS — consent or nothing
A text goes only to a number the owner gave us, or after a written yes to a specific number — never merely because they replied to an email. Quiet hours 8am–8pm recipient-local by company state (never area code). STOP = global, same-day, across every channel.
Why: $500–$1,500 statutory damages per text, uncapped, no proof of harm required — and a demand-letter machine targeting exactly our size of sender.
Email — the CAN-SPAM floor
Accurate From and subject on every send. The physical postal address in every email. A working opt-out — our reply-to-opt-out P.S. — honored fast and suppressed centrally across all domains.
Why: cold B2B email is fully legal under CAN-SPAM — as long as the floor is never missed. One sloppy template risks every domain.
LinkedIn — the hard caps
≤100–150 actions/week per account, ~20–25 invites/day max. Scale = more SDR profiles (created 5 weeks ahead), never more actions. Founder profiles run only through HeyReach cloud — never a Chrome-extension executor.
Why: extensions carry ~60% higher detection risk, and a banned founder profile is an unrecoverable brand asset.
Facebook — human hands only
Founders' profiles are never automated. Ever. The Page DM lane is manual, founder-profile, groups-first — ~10–15/day per operator with group-context openers and no links in the first message.
Why: every FB DM automation tool puppets a logged-in profile against Meta ToS — during the most ban-happy enforcement era yet, with mass ban waves hitting even legitimate accounts.
Week one, day by day — after Chris hands over five things.
Nothing in the stand-up starts while a prerequisite row is blank. Chris's checklist, in plain language:
Share the keys
Domains + registrar/DNS logins, Google Workspace super-admin, Reply.io / Clay / GHL credentials, the company card, and the vendor signups it pays for.
Protect the list
Export current clients + open opportunities as the suppression seed — loaded into Clay before any row is marked sequence-ready. Confirm the Command Center link and the Chamber 60k location.
Legal facts
The A2P registering entity + EIN exactly as IRS-registered (mismatch is the #1 rejection cause), and the CAN-SPAM footer postal address that goes in every signature. A2P approval can take 10–15 days — register now.
The words
Interim T1+T2 email copy written or approved by Chris by Day 3 — sends slip if missing. Nobody drafts-and-sends. Plus the free-value asset repliers receive.
Small stuff
Pick the shared build-log location, and book the weekly review as a recurring event (suggested: Friday 3:00–3:45pm) on Day 1.
Week 1, day by day
Infrastructure
SPF/DKIM/DMARC verified on every domain, Postmaster Tools registered, Reply.io configured (tracking OFF, plain text ON, 30/inbox cap incl. follow-ups), warmup history verified — ≥8 weeks or the inbox is treated as new.
The pilot
500 stratified Command Center rows through the full waterfall. Ethan tracks the gate table daily; Anthony makes verification calls; a Tavily head-to-head runs alongside.
Funnel live
Round-robin triage calendar + booking form (unchecked SMS-consent checkbox), pixel firing, cross-tool reply sync seed-tested — a planted reply must exit all three tools before any real send.
First sends
Only if DNS is verified, placement tests >90%, copy + postal address are in hand, and ~100 verified rows are ready. 30/day — Chris reads every single one.
The weekly review — the ritual that fixes the old failure
The 140-days-unread lesson: the prior outbound attempt didn't fail for lack of a spec — target, bonus, and daily + weekly reports all existed. It failed because the reports went unread for 140 days. The fix is a standing review, not more reporting.
Friday, 45 minutes, Ethan pre-fills the numbers by Thursday EOD. The 6-point agenda:
- 1Self-check rollup — fail-rule days first
- 2Funnel vs target — incl. the W2 month-1 checkpoint
- 3Deliverability panel — complaints, bounces, placement, reply rate
- 4Message iteration — replies per 100 delivered, by template and tag
- 5Spend vs gates — Clay action burn, $/sequence-ready record
- 6Flag register — anything ready for a ruling
The daily self-check — a bad run announces itself
list_build_selfcheck: operator: chris # chris | ethan | les date: 2026-08-04 state: TX sources_used: [tdlr_file, gbp_outscraper, site_crawl, clay_waterfall] rows_added_today: 1240 rows_target: 1000 volume_vs_target: 124% base_tier_completeness: # % of today's rows, present + verified company_name: 100% owner_first_last: 91% verified_email: 63% # sourced 41% / patterned 22% linkedin_url: 34% patterned_on_catchall: 87 # hard-quarantined today duplicates_blocked: 212 sequence_ready_rows: 780 flags: ["GA roster still not delivered"]
Pass rule: volume ≥80% of target · sequence-ready ≥50% of rows added · patterned-on-catch-all <15%. Any fail → fix the pipeline before adding volume.
The rewrite floor. Under 1% positive replies after 500 delivered on a template → that template gets rewritten. Templates are judged weekly.
The placement proof. A trailing reply rate at or above 2% is the real evidence we're landing in primary inboxes — better than any warmup dashboard.
The machine's clock. The machine itself is judged on the book's 4–6-month clock — never at day 30. Quitting at day 30 is the #1 named failure in outbound.
The flag register: only Chris rules.
Anything research argued should change a LOCKED decision lives here — never baked into the plan. Eight flags. Treat anything not marked implemented or approved as not approved.
franchise_national, C-tier — per the locked "everyone, no gates."Go deeper: the four documents.
This page is the map. These are the territory — all in the private repo at github.com/anchorworkshq/outbound-machine. Read them in this order.
START HERE — Agent Handoff
Ground rules, the file map, company context, current state, the flag register, execution tracks, and the shared glossary. The boot document for every session.
Open in the repo Source of truthOutbound Machine SOP
Every LOCKED decision: the two qualifiers, the 30k list rules, the enrichment ladder, the sequence, the ramp, the LinkedIn model, accountability. LOCKED is final.
Open in the repo The operating planExecution Plan — Phase 2
v1.1, adversarially reviewed. Week-by-week timeline, prerequisites, the 500-row pilot spec, the exact Clay waterfall with prices, the 13-state build SOP, the ramp with gates, the budget.
Open in the repo The evidence layerResearch Report — Phase 1
~23k words, fact-checked. Nine full research sections with every source: the book, deliverability, the waterfall, LinkedIn, TCPA, HVAC GTM, the 30k build, detection, Facebook.
Open in the repo