Rebuilding Growth Engine
Private strategy document. Enter hub password to continue.
That's not right. Try again.
Rebuilding Growth Engine: Combined Automation Master Plan
What this is: one plan that fuses the two we already published, the Local SEO Strategy and the Offline Marketing Automation Blueprint, into a single hands-off engine. It answers one question: how much of both plans can Charlie run so you and Bob barely touch it, and exactly where a human still has to step in.
1. The one-engine insight
The two plans read like two projects. They aren't. They're one machine seen from two sides.
Here's the hinge. The local SEO plan's single most valuable ranking lever, after the Google Business Profile itself, is local backlinks and citations: a family-law attorney linking to Bob from her "resources" page, a Daily Camera feature on gray divorce, a library listing Bob's free talk, a chamber-of-commerce membership. Every one of those is a local ranking signal Google and the AI local pack weight heavily. And every one of those is also a line item in the offline marketing plan. The attorney referral, the press hit, the calendar listing, the community talk: those live in the offline plan as trust-building outreach, and they live in the SEO plan as link-building. Same act, two payoffs.
So we don't run "SEO" and "offline marketing" as separate efforts that happen to overlap. We run one outreach engine whose every output does double duty: it warms a human relationship (the offline goal) and it strengthens Bob's local search entity (the SEO goal). Cross-referencing the two plans isn't housekeeping, it's the whole design. Nobody does the same outreach twice, and every touch compounds on both scoreboards.
The second half of the insight is infrastructure. The offline blueprint already specified seven systems built on shared plumbing: one database, one gated dashboard, one Gmail pipeline, one family of print templates, one scheduler. The SEO plan needs a town-page pipeline, a GBP posting cadence, a review loop, and rank monitoring. Those four bolt straight onto the same plumbing. The database that tracks which attorney referred which student is the same database that measures local leads. The dashboard that shows the pitch queue is the same dashboard that shows the grid-scan heatmap. Build the spine once, and the SEO layer is cheap.
The plan in one line: build one shared engine, point it at both scoreboards, and gate only the handful of moments that genuinely need a human.
2. The unified system map
Eleven components, all on shared infrastructure. Seven come from the offline blueprint; four are the SEO layer. The shared plumbing is one Cloudflare D1 database, one password-gated dashboard at annettethompson.com/rebuilding/, one Gmail account with filters, one Typst template family, Windows Task Scheduler for the cron, and the grid-scan tool that's already built and working (scripts/grid-scan/).
The offline seven (the outreach engine)
- S1, the CRM spine. D1 + gated dashboard. Tracks contacts, outreach, referrals (who sent which student), listings, collateral, and source-request queries. This is also the local-lead attribution the SEO plan asks for. Everything else reports into it.
- S2, contact database. Scrapes public directories (Colorado Bar family-law, Psychology Today, mediation.com, CDFA/IDFA, Google Places) into a clean referral list, and pulls live current bylines from local outlets (never invented names). Feeds referrals, press, podcasts.
- S3, source-request monitor (the crown jewel). Gmail ingests Source of Sources / HARO-free / SourceBottle digests, a scheduled parser filters to Bob's lanes (divorce, gray divorce, co-parenting, midlife, therapy-meets-tech), and Charlie drafts a Bob-voice answer with the right stat and credential line for each match. Every landed quote is a permanent "as seen in" asset and a high-authority backlink.
- S4, calendar + free-listing engine. A reusable listing template filled per event, pushed to Patch, library, city, county, and small-paper calendars. Every talk and cohort start gets listed everywhere. Many of those listings carry local links.
- S5, press-outreach + follow-up tracker. Mail-merge-style personalized pitches off verified hooks, each logged with a follow-up date the dashboard surfaces, because most PR wins come from the second touch nobody remembers to send.
- S6, Typst collateral generator. One
stats.toml(4.96 stars, 293 reviews, 3,000+ alumni, prices, dates) drives rack cards, flyers, the referral one-pager, sign-in sheets, and postcards. Change a number once, rebuild every asset in one command, so nothing's ever stale. - S7, seasonal PR automation. Task Scheduler fires the timely hooks (holidays-and-divorce in November, January divorce season, Best of Boulder voting in March, the cohort-fill push 8 weeks out) with a pre-drafted pitch ready for Bob.
The SEO four (the search layer, same plumbing)
- A1, town-page generator. A template plus per-town data plus
stats.tomlproduces each 1,000-to-1,400-word "areas served" page withLocalBusiness+FAQPageschema and a UTM-tagged CTA. The first four (Boulder, Longmont, Louisville, Lafayette) are already drafted by hand; this automates the rest of the tiered backlog from the SEO plan. Output flows to the site repo. - A2, GBP posting engine. The 30-post bank (already written) drives a 2x/week Monday/Friday rotation of Update / Offer / Event / Product, front-loaded with Offers before the July 24 early-bird deadline. Charlie drafts, schedules, and rotates; also drafts the 12 Q&A seeds. (Google retired the local-posts API, so the posting click is a browser step, see the ledger below.)
- A3, review-velocity loop. Monitors new reviews, drafts a Bob-voice response for each within the 48-hour window, tracks velocity against the two-a-month floor, generates the one-tap review link for graduations, and flags suspicious or competitor-malicious reviews. Honors the Shea rule (never censor a graduate's own words, never author a mention).
- A4, rank + citation monitoring. The grid-scan tool on a schedule with delta mode (show me the spread after the GBP work lands), plus a NAP-consistency audit across the tier-1/2 directories and a schema validator. This is the measurement layer that replaces the ~$300/yr Local Falcon and the ~$52/mo SE Ranking, both already dropped.
One honest gap: GA4 conversion tracking can't be automated yet because it doesn't exist, and we're deliberately not building it into GoHighLevel while a platform move is likely. It's a launch-day requirement for the next platform, not a system to build now.
3. The hands-off ledger
Here's the honest sort. Every step of all eleven components falls into one of five buckets. The first bucket is Charlie alone. The next four are the only places a human touches it.
No human, runs up to the gate. Building all eleven components. Scraping and enriching the contact database. Parsing source-requests and drafting Bob-voice answers. Building press lists and drafting pitches and scheduling every follow-up. Generating all collateral from the stats file. Firing seasonal reminders with pre-drafted shells. Generating town pages with schema. Drafting, scheduling, and rotating GBP posts. Monitoring reviews and drafting responses. Running grid scans, delta comparisons, NAP audits, and schema validation. Maintaining the dashboard and the referral-attribution ledger. This is roughly 80% of the total labor in both plans, and it's the busywork, the part that today just doesn't get done because nobody has the hours.
His name, his voice, going public. Approving a source-request answer before it sends. Approving a press pitch or an op-ed. Approving a batch of GBP posts. Approving a town page before it goes live. Approving review responses. Signing off on collateral before it prints. Each is a yes/no on something already fully drafted. Realistically minutes a week, reviewed from the dashboard.
Things with no API. Publishing the approved GBP posts (Google retired the posting API). Submitting the calendar forms (Patch, library, city, county). Posting the prepared Nextdoor drafts (their terms forbid scripting). The reviews owner-dashboard scrape. Any GHL change or next-platform work. These batch nicely: one browser session every week or two, working from a queue Charlie already filled, human-paced.
Stops and waits for a person. The one-time Whitespark citation cleanup (~$80). Any print run or postage or EDDM mailing. A standing grid-scan budget (scans are ~50 cents each). Patch's optional $1/day cross-promotion. A dedicated Google Voice number for NAP consistency (free-to-cheap). Nothing here spends a dollar without Bob's explicit go-ahead.
Bob presenting live at a talk, a podcast, community radio. The warm attorney-to-Bob introduction. The pastor-to-pastor conversation. A graduate choosing to leave a review. A professional choosing to hand a client a card. Automating any of these would break the exact thing that makes them work. Charlie's job around this bucket is to remove every logistic so these moments happen more often, never to fake them.
So the honest answer to "how much can you do hands off" is: about 80% of the labor runs with zero human input, and almost all of the rest is one-click approvals and one batched browser session. The genuinely un-automatable core, Bucket 5, is small, and it's the trust that the whole business is built on. We automate the friction, never the warmth.
What a running week looks like: Charlie has already scraped, drafted, scheduled, scanned, and queued everything overnight. Bob opens the dashboard, spends a few minutes approving a source-request answer, three GBP posts, and two review responses. Once every week or two, Annette opens her browser and clears the posting-and-listing queue in one sitting. The seasonal jobs fire themselves. The grid scan reruns itself and shows the spread. That's the hands-off state.
4. What actually lands on your plate
Stripping it all the way down, here's the complete list of what only you or Bob can do. Everything not on this list, Charlie handles.
- Bob, from the dashboard (minutes a week): approve the drafted quotes, pitches, posts, pages, and review replies before they go public. That's it for the recurring ask.
- Bob, the irreducible core (the part that fills seats): show up and present, take the warm intros, have the pastor and attorney conversations, invite graduates to review at the end of each cohort. This isn't new work the automation created; it's the work that was always the actual business, now with the logistics stripped off so it happens more.
- Annette, batched browser sessions (an hour or two, every week or two): clear the GBP-post / calendar-listing / Nextdoor queue Charlie has pre-filled, and run the periodic reviews scrape. Plus the one-time next-platform work when that decision lands.
- Annette, the money calls (rare, discrete): yes or no on the ~$80 Whitespark cleanup, the grid-scan budget, and any print run when collateral is ready.
That's the whole human footprint. Compared to running these two plans by hand, it's a rounding error.
5. Build order, cost, and what starts today
Recommendation: build it, max autonomy, phased. The build cost is almost entirely Charlie's time on cheap delegated models (OpenRouter and Codex, not the Anthropic meter). The gates protect everything that matters. And the two highest-leverage pieces, the source-request monitor and rank monitoring, start paying off immediately and independently of the rest. There's no reason to wait, and no reason to do it half-way.
Phase 1 (week 1), the spine and the crown jewel. S1 CRM + dashboard, S3 source-request monitor, and A4 rank monitoring (the grid tool is already built, this is just scheduling it). Starts producing "as seen in" assets and rank baselines from day one.
Phase 2 (week 2), collateral and the content pipeline. S6 Typst generator (unblocks four channels at once), S2 contact database, and A1 town-page generator (four pages already drafted, this generates the backlog).
Phase 3 (week 3), the outreach engines. S5 press tracker, S4 calendar engine, A2 GBP posting engine (30-post bank ready), and A3 review loop.
Phase 4 (ongoing), on rails. S7 seasonal automation. Once the systems it triggers exist, this is just scheduling, and it's what keeps the engine running without anyone remembering to.
Running cost, the whole engine: under $10 a month
D1, Pages Functions, Workers, and Gmail are all free-tier. Grid scans are ~50 cents each, a few a month. DataForSEO lookups are pennies. We already dropped Local Falcon (~$300/yr, the grid tool replaces it) and SE Ranking (~$52/mo, DataForSEO replaces it). The only optional spend is the one-time ~$80 Whitespark cleanup and a near-free Google Voice line. For a system that runs 80% of two marketing plans, that's essentially nothing.
What can start today, with zero decisions pending
Phases 1 and 2 in full. The CRM, the contact scraping, the source-monitor build, the collateral templates, the town-page generator, and the grid scheduling all get built with nobody needing to decide anything, because none of them go public. The gates only bite the moment something's ready to send or post. So the right move is to start building now while the six decisions below get answered in parallel. No time lost.
6. The six decisions that unlock full speed
Consolidated from both source plans. Only these gate the public, spending, and irreducibly-human parts. The build doesn't wait on any of them.
Green-light the build (and the ambition level). Max autonomy phased (recommended), crown-jewels-only first, or plan-only for now.
The source-request setup. Register Bob on Source of Sources + HARO-free + SourceBottle (all free), and decide whether digests route to Bob's inbox or a dedicated address Annette controls. Needs Bob's OK to create accounts in his name.
One canonical NAP. Confirm the GBP primary category (recommendation: keep "Divorce service," add secondaries) and settle one phone number: Bob's personal line, or a dedicated Google Voice number for clean consistency across the web.
The money calls. Whitespark ~$80 cleanup, yes or no. A standing grid-scan budget (a few dollars a month). Collateral print budget when the first cards are ready.
Bob's lists. The alumni list (for the referral postcard) and the graduate-town roster (so town pages carry genuinely local proof). Only Bob has these.
Platform-decision timing. Is the GoHighLevel move likely enough to hold conversion-tracking and town-page publishing for the next platform, or do we ship on GHL now? This is the one decision that changes sequencing rather than just unlocking a gate.
Answer these and the gates open. Until they're answered, Charlie builds the 80% that needs no permission.
Sits on top of the two published source plans, the Local SEO Strategy and the Offline Marketing Automation Blueprint. It doesn't repeat them, it ties them into one engine. Honors every standing rule: Bob approves everything public, sends stay human-paced, spending stops for a person, the source filter stays conservative, no Shea Karam in anything we author.