Topic Library

• Each topic = a 30-sec reel premise
  The 1–2 sentence note IS the script kernel.
• Each sub-pillar groups by 5 canonical buckets
  Tips · Primary Q · Secondary Q · Expectation · Objection — every entry maps to a topics_bank.csv slot.
• Voice cadence
  [Diagnostic Finding] + [The Pattern] + [The Infrastructure Fix] + [Proof] — every reel opens with what you found across the audit fleet, not with "let me tell you about AI."
• Winner rotation
  When a topic earns winner status (×2 ranker multiplier), retire the weakest topic in that sub-pillar and slot a fresh angle from below.

Tips production tactics from the audit fleet

• Green screen in your garage, one camera, 30 videos in one afternoon.
  Pattern from 150+ shops: the ones who post consistently batched. Tape green sheet to a wall, film 30 hooks in one outfit, editor swaps backgrounds in post. One filming day = a month of content.
• The swipe-file rule — no AI script until you transcribe 5 winners.
  Every operator I audit who tried "AI write me an ad" got generic output. Feed it 5–10 transcribed top-performing organic posts in your niche first. Then ask.
• AI UGC for niches without a founder on camera.
  Pattern: cleaning, concrete, flooring operators usually won't film. Use Higgsfield + Nano Banana Pro to generate avatars matched to the audience demographic.
• Owner-on-camera vs. employee-on-camera — when each wins.
  Owner = identity and authority. Employee = relatability and reach. Audited 150+ shops, the best mix is 70/30 owner.

Primary Q what to actually post

• Content that books appointments vs. content that gets views.
  Pattern: shops with 200K views and zero booked work are running demand-capture content on a demand-gen channel. The metric is booked jobs, not impressions.
• Stories beat tips — pull yours from client interviews.
  Every shop I've audited with a high-converting feed has one thing in common: a 30-minute Loom from a happy contractor, transcribed and broken into content for the year.

Secondary Q sharper niche cuts

• Authority content that books before they shop.
  The pattern in the 47x ROI accounts (Riley): educate while the homeowner is on Instagram. By the time they're ready to call, you're the only contractor they trust.
• Foundation repair vs. waterproofing content — different anxieties, different hooks.
  Foundation = structural fear ("Is my house safe?"). Waterproofing = damage prevention ("Will my basement flood?"). Hooks land differently. Audited the difference across 50+ shops.

Expectation benchmarks operators ask about

• How many videos before leads start coming in?
  Pattern from the fleet: 30 posted before you should expect anything. Then re-batch and run another 30.

Objection what operators push back on

• "I posted before and got zero leads."
  Diagnostic: 9 out of 10 times, you posted demand-capture content on a demand-gen channel. Different game. Show the audit example.
• "Hey guys" and the LinkedIn voice — why corporate energy kills reach.
  Pattern: every account I've audited that pulls views opens like an operator at a job site. Every account that flatlines opens like a marketing manager at a stand-up.

Tips how to actually run qualification

• The 4-question canonical qualifier: project → zip → decision-maker → urgency.
  Pattern from every healthy bot I've audited: all four must pass. No human review until every box is green. Filters 60–75% of garbage before it hits the rep's calendar.
• The multi-LLM bot stack — Gemini warms, Claude handles objections, GPT closes.
  Audited the best-performing bots in the fleet — none use one model for the whole flow. Stack them like a sales team.
• The hard-boundary prompt: "Not trained on this — I'll get back to you."
  Pattern: every bot I've seen go off the rails was guessing. Bound the scope. Hand off cleanly when it's outside.
• Persistent memory — why your bot needs to know your business.
  Without it, every conversation starts from zero. With it, the bot answers like a 5-year employee. The difference is operationalized memory.

Primary Q the big question operators ask

• Estimate-from-photo — homeowner texts a crack, bot returns a ballpark in 30 seconds.
  GPT-4V on the inbound photo. Setter walks into the appointment with a number already on the table. Audited at 3 shops — show rate up 18%.
• How an AI inbox qualifies a lead from cold click to booked quote.
  Walk through one real thread from the audit fleet. Show the prompts, the routing, the calendar booking. Transparency = authority.

Secondary Q

• Bot vs. human — where each one wins on close rate.
  Pattern: bot wins on speed and qualification. Human wins on objection-handling and price negotiation. The shops with both win the most.
• SMS vs. Messenger vs. IG DM — which converts best for contractors.
  Pattern across 150+ shops: Messenger wins on foundation repair by a mile. SMS wins on re-engagement. Different tools, different jobs.

Expectation

• How many leads get disqualified — and why that's the win, not the loss.
  60–75% gets filtered. The other 25–40% close at 2–3× the rate of unqualified leads. The math is in the qualification, not the volume.

Objection

• "I don't want a bot talking to my customers."
  Diagnostic: bot only talks to ad clickers. Existing customers still get you. Show the architecture.
• "We tried a chatbot once and it sucked."
  ManyChat in 2019 sucked. Modern LLM-powered bots are a different category. Show one running live.

Tips Andromeda-era mechanics

• Open the targeting. Detailed audiences are dead.
  Pattern from the audit fleet post-Andromeda: every shop still using detailed targeting is paying $30+ extra per lead. Your creative IS the targeting now.
• Why $20/day per ad beats $100/day per campaign.
  Andromeda needs daily-budget-per-ad signal. Spread across 5 ads, let the algorithm pick winners. Audited the rule across 20+ campaigns this quarter.
• The 3-ad portfolio rule for any scaling campaign.
  Every stable campaign in the fleet has 3+ ads each carrying meaningful spend. If one fatigues, budget shifts automatically.
• The organic outlier → paid amplification loop.
  Pattern from the best-performing accounts: top organic Reel from the week → $20/day open-targeting Meta lead ad → algorithm picks the winner.

Primary Q

• The 97% — what it means and why it's the entire pond.
  3% are price-shopping on Google. The other 97% have the problem and don't know there's a solution yet. Audited 150+ shops; the winners all fish where it's empty.
• Facebook conversation campaign setup, step by step.
  Conversion event = qualified lead, not link click. Messenger destination. CAPI on. Pixel hot. Diagnostic checklist from the audits.

Secondary Q

• Demographic-match rule — AI avatar matches the ad set.
  Pattern: if the ad shows a 26-year-old female but the audience setting is 45+ male, relevance score tanks. Match them.
• When to use lookalike audiences in 2026.
  Sparingly. Open targeting beats most lookalikes when creative is strong. Use lookalikes only when you have 1,000+ closed-job seeds.

Expectation

• What CPL should I expect in week 1 vs. week 8?
  Pattern across launches: Week 1: $50–80. Week 8: $30–50. The algorithm learns. CPL drops weeks after you feed it, not days.
• How do I know an ad is fatiguing before ROAS drops?
  Diagnostic: frequency above 3.5 + CTR declining 2 weeks in a row = retire proactively. Don't wait for ROAS to crash.

Objection

• "Facebook doesn't work in my market."
  Audited 12 shops in markets under 200K population this year. Facebook worked in every one. Your offer or creative is broken — not the platform.
• The Lead Form trap — why FB Lead Forms don't fire CAPI by default.
  Pattern: 90% of contractor ad accounts running Lead Forms are leaking attribution. SHA256-hashed email + phone fixes it.

Tips

• The 10-second greeting — bot engages before they leave the Facebook app.
  Pattern from the best-performing inboxes: conversation starts while the homeowner is still scrolling. Don't wait for them to click through.
• Speed-to-lead by channel — SMS vs. DM vs. email vs. call.
  Audited across the fleet: SMS wins on read rate. DM wins on conversation. Call wins on close rate IF they pick up. Stack all three.

Primary Q

• The 5-minute decay curve.
  Pattern across 150+ shops: respond in 5 minutes → 80% pickup. 30 minutes → 25%. An hour → 12%. The whole game.

Secondary Q

• Why "fast" without "qualified" is wasted speed.
  Diagnostic: a fast response to a tire-kicker is still a tire-kicker. Qualify first, then sprint. Backwards-order kills close rate.
• Why your auto-responder might be killing close rate.
  Pattern: "Thanks, we'll get back to you" responses correlate with 40% lower close rate. Auto-responder should ASK A QUESTION, not promise a callback.

Expectation

• How much does speed actually move close rate?
  3–4× lift on appointments booked from speed alone. Bigger than any other lever in the system.

Objection

• "We call back within the hour."
  You're losing 70% of those leads to the contractor who called in 5 minutes. Math doesn't care about your schedule.
• "My team can't respond at 11pm."
  Right. Your bot can. That's the unlock.

Tips

• 5 things to send between booking and the visit.
  Audited the high-show-rate fleet: owner intro video, what to expect on the call, sample inspection report, 2-min testimonial, calendar reminder. Hit all 5.
• The 30-minute owner story drop.
  Pattern from the highest-show-rate accounts: Loom of the owner explaining how they got into the business, sent in confirmation sequence. Authority before they meet you.
• The post-booking video that cuts no-shows in half.
  60-second clip from the owner: "Looking forward to meeting you. Here's what I'll bring." Personal, not corporate.

Primary Q

• The 24-hour booking-to-appointment gap is your biggest leak.
  Pattern: most no-shows happen here. Most "we forgot" happens here. Most reps arriving to empty driveways happens here. Plug the gap.

Secondary Q

• How a homeowner decides whether to trust you before you arrive.
  Diagnostic: they Google you, check Instagram, watch your videos. By the time you knock, the call is half-closed already. Make sure the pre-call artifacts exist.
• Pre-appointment authority drops vs. confirmation texts.
  Confirmation texts are admin. Authority drops are content. Most contractors only do the admin. The content is what builds the trust.

Expectation

• Show-up rate for demand-gen leads vs. shared leads — the real numbers.
  Pattern from the audit fleet: demand gen = 70–80%. Shared leads (Angi) = 35–45%. The leads you generated trust you. The leads someone sold you don't.

Objection

• "We already send a confirmation text."
  Diagnostic: that's table stakes. Six other contractors are sending the same one. Differentiate or get ignored.
• "My team doesn't have time to make personalized videos."
  Owner records 30 seconds, AI clones the voice for the rest. 10 personal-feeling videos in 5 minutes.

Tips

• The outcome stack — 5 layers between estimate and lifetime customer.
  Pattern from every $1M+ shop I've audited: speed → follow-up → estimate rehash → reviews → loyalty. Each layer has a dollar number. Most contractors only have layer 1.
• The estimate rehash playbook — $30K per quarter, on average.
  Diagnostic from the audit fleet: stale quotes in GHL → reactivation sequence → bot handles replies → setter dials warm responses. That's it.
• The reactivation playbook — SMS blast to dormant DB.
  Pattern: anyone who didn't close in the last 6 months → limited-spot offer → bot fields replies → calendar fills.
• AI review responder for new Google reviews (without sounding fake).
  Pull context per client, draft replies in their voice, owner reviews + sends. Pattern: shops that reply rank higher AND book the next homeowner faster.

Primary Q

• Why 80% of your dead leads are still closeable.
  Valerie's playbook (Redeemers Group): reactivated dead leads → 80% qualified → 5-figure month. Audited her stack — your CRM is sitting on next quarter's revenue.
• Why do I lose jobs after great appointments?
  Diagnostic from 50+ audits: no follow-up plan. Rep walks out, hopes the client calls back. They don't. Build the cadence.

Secondary Q

• The 24-hour "need to think about it" follow-up template.
  Day 1: "Anything I can clarify?" Day 3: case study. Day 7: scarcity. Don't wait a week between touches.
• The 7-touch follow-up — varied channel, varied message.
  Pattern: email → SMS → DM → call. Mix them. One channel feels like a stalker. Four feels like service.
• The post-job nurture sequence — review, referral, repeat.
  Audited the highest-LTV accounts: job closes → review ask → 30-day check-in → 90-day referral ask → 1-year reactivation. Built once, runs forever.

Expectation

• How many dead leads are closeable?
  Pattern from 12-month CRM audits: 30–50% are recoverable with the right sequence. Most operators leave it on the table.

Objection

• "We follow up twice and move on."
  Your competitor follows up 7 times. They're closing your dead deals. Show the audit.
• "My team won't run the follow-up sequence."
  That's why you automate it. Bot does touches 1–6. Human only handles the warm response.

Tips

• The screenshot test — if you wouldn't send it to a partner, redesign or kill.
  Pattern from 30+ dashboard audits: the useful ones produce a "look at this" moment. The rest are database queries.
• One hero metric per role.
  Owner sees revenue. Marketing sees CPL. Sales sees close rate. Pattern: shops that show everyone everything get ignored by everyone.
• Coin your own metrics — "Qualified Hour Rate" beats "CPL."
  Diagnostic: renaming the metric makes it yours. Yours = stickier. Sticky = retention.

Primary Q

• What should a contractor dashboard actually show?
  6 numbers from the audits: leads, quotes, show rate, close rate, cost per booked job, jobs sold. Anything else is noise.
• Cost per booked job — the only number that matters.
  Pattern: CPL is vanity. Cost per quoted opportunity is closer. Cost per booked job is truth. The shops that track it beat the shops that don't.

Secondary Q

• Which ad actually booked the job?
  Multi-touch attribution + offline conversions back to Meta. Diagnostic: most agencies can't answer this. You should.
• Crew utilization as the master KPI for contractor dashboards.
  Pattern: marketing exists to fill crews. If utilization is high and CPL is also high, you don't need to lower CPL — you need more crews.

Expectation

• How long until the dashboard is reliable?
  Pattern from launches: 6 weeks of data minimum. Anything shorter is statistical noise. Don't make decisions on week 2.

Objection

• "I ask every customer where they found us."
  Diagnostic: they lie. Half don't remember. The other half default to "Google." Self-report attribution is the weakest data in the stack.
• "Your ads are doing great" — why agency reports are usually nonsense.
  Pattern: reports designed to keep retainers. Real dashboards make the agency accountable.

Tips

• Which event to fire when — qualified, booked, closed.
  Pattern: three CAPI events, three different learning signals. Fire all three or you're training the algorithm on the wrong outcome.
• The Lead Form trap — why FB Lead Forms don't fire CAPI by default.
  Diagnostic from the audit fleet: most contractor accounts running Lead Forms are missing 100% of attribution. SHA256-hashed email + phone fixes it.
• Pixel vs. CAPI vs. offline conversions — what each does.
  Pixel = client-side, cookie-blocked. CAPI = server-side, blocks-proof. Offline = closed-job revenue. Pattern: you need all three.

Primary Q

• Train Facebook to find closers, not messengers.
  Pattern: without closed-won data feedback, the algorithm finds people who'll message — not people who'll pay. Two very different audiences.
• AI is the new SEO — show up when ChatGPT recommends contractors.
  Diagnostic: when a homeowner asks ChatGPT "best foundation repair in [city]," the result is determined by structured data. Most contractors are invisible there. Fix it.

Secondary Q

• What does the algorithm learn when I send closed-won data back?
  Job value. Not lead value. Not message volume. The difference between a $9K job and a $200 service call.

Expectation

• CPL drops 30–50% after 90 days of closed-won feedback.
  Pattern across launches. The algorithm gets sharper every week you close work and feed it back. 3–4 weeks for first signal. 90 days for material drop.

Objection

• "My pixel is already set up."
  Diagnostic: pixel without CAPI is half a loop. CAPI without closed-won data is a lukewarm loop. Show the full audit.
• "I don't trust giving Facebook my sales data."
  Hashed. Anonymized. Aggregate. Same data they get from your competitor — except yours is making you money instead of them.

Tips

• What a $14M contractor marketing stack actually looks like.
  Postgres backend, N8N workflows, Meta + GHL APIs, dashboards on top. Pattern: not a single SaaS — a connected system.
• When your CRM should talk to your AI — and when it shouldn't.
  CRM passes data IN. AI never writes back to CRM unsupervised. Pattern from the audits: boundaries matter.

Primary Q

• The Super Pixel — what 10K+ appointments of audience data unlocks.
  Cross-pollinates targeting on day 1 for every new client. You can't buy this — only build it. The data moat IS the product.

Secondary Q

• Why a connected system beats a single SaaS.
  ServiceTitan does CRM. Chiirp does SMS. HubSpot does email. Pattern: none of them talk to each other. A system does.

Expectation

• How much technical setup does this actually require?
  Diagnostic: 6-week install. You don't touch a prompt, pixel, or Meta dashboard. White-glove or it doesn't ship.

Objection

• "But ServiceTitan already does this."
  Diagnostic: ServiceTitan is a CRM. The system we're describing is a marketing operating system. Different layer of the cake.
• "I just need more leads, not a data system."
  Pattern: the data system is what makes the leads compound. Without it, you're starting over every month.

• Voice cadence is binding
  Every topic kernel uses the Diagnostic Finding cadence (Pattern from X audits → The Pattern → Infrastructure fix → Proof). Voice = Systems Consultant / Asset Architect.
• Backend Infra (3.3) is deliberately smallest
  Audience appetite is lowest. Don't overproduce there.
• Inherits from Content OS
  This outline lives downstream of the Content OS. Update Content OS first if positioning shifts — outline inherits from it.
• Replacement queue
  When a topic becomes a "winner" in Topics Bank (×2 ranker multiplier), retire the weakest topic in that sub-pillar and slot in a fresh angle from above.