A complete, no-hand-waving accounting of all eleven Auto Marketing sites: the onboarding upgrade Beau called for, the five gaps the engine finds on every site, and a per-site spec measured down to the H1 count. Every claim is a live engine reading. Every fix is a copy-ready prompt. Nothing here is estimated.
Beau, testing the engine as a first-time user would, made two clean calls. First: when someone types their website, the AI should fill in the basics for them — name, what they do — editable, of course. Typing a business description from scratch is exactly the friction that makes people abandon an onboarding. Second: on the flagship, the “set up your department” call-to-action was too quiet — he had to hunt before realising he could continue.
Both shipped, live across five sites. The auto-fill reads the real page and returns the site's own words to edit — never an invented sentence, and only into empty fields. It reuses the deterministic scoring engine, so there is no per-message AI bill; it stays inside the flat-subscription rule the whole network runs on. Verified end-to-end on a real site (Taylor F. Jones), editable, on-brand, with the “nothing was overwritten” guard proving itself on a repeat run.
/onboard/ form got a “✨ Fill from my site” button (fills business name + what-you-do).Two prompts finish the job: carry it to the last core site, and start measuring whether it actually helps.
You are a front-end engineer on the WholeTech droplet. deptmatic.com runs its own control-panel app at /var/www/deptmatic.com (app.js), separate from the flagship, and its onboarding does NOT yet auto-fill the basics from the analyzed site. Add the behaviour the flagship and the /onboard/ mirrors already have: 1. When the user enters their website URL, add a "Fill from my site" affordance that POSTs the URL to /api/analyze (shared autoengine backend, :8932). 2. From the response fill ONLY empty fields: business name <- data.brand; what-you-do <- data.analysis.signals.meta_desc || og_desc || data.niche. 3. Never invent; never overwrite a typed field; tint filled fields and show "Filled N fields from your live site." 4. No LLM/metered call — reuse the deterministic analyze fetch only. Back up app.js (timestamped .bak), validate with node -c, deploy, confirm live. Match deptmatic.com's own design — do not import flagship styles.
You are a product analyst. We just shipped auto-fill onboarding across five sites but have no read on whether it lifts completion. Add lightweight, privacy-safe, no-cookie event counting (server log lines or a tiny same-origin /api/ev endpoint — NO third-party analytics, NO metered service) for: onboarding_started, autofill_clicked, autofill_filled_N, onboarding_completed. Then build a one-page internal readout at automarketingengine.com/onboard-metrics/ (noindex, auth-gated) showing funnel counts and the autofill-used vs not completion rates. Do not fabricate data — show real counts only, and label "n too small" until there's enough.
The Auto Marketing family is the engine. WholeVoyage is one of the biggest real businesses to point it at — a whole side of the network of booking-grade villa and travel sites, many of them Robbie Dolgin's. Everything in this report — the five cross-cutting gaps, the department, the 52 prompts — applies to them directly, and the numbers say the opportunity there is large.
Here is the villa/travel side as the engine reads it today. Same instrument, same honesty. The hub, wholevoyage.com, sits at 55 — squarely mid-pack, held back by exactly the family-wide gaps (AI 45, Social 42, no JSON-LD). But the spread is the story: a real high (smallhomevillage.com at 73, with a rare AI 90) and two genuinely broken pages that need triage before anything else.
| Villa / travel site | Overall | SEO | AI | Soc. | Tech | Reach | Note |
|---|---|---|---|---|---|---|---|
| smallhomevillage.com | 73 | 68 | 90 | 42 | 90 | 63 | family high — AI 90 is a network-best |
| smallhomevillages.com | 70 | 64 | 77 | 58 | 90 | 50 | strong sibling |
| bnbhot.com | 66 | 54 | 77 | 17 | 90 | 68 | best Reach; Social 17 |
| staysandalwood.com | 64 | 50 | 61 | 58 | 90 | 48 | Robbie · Sandalwood flagship (SiteMinder) |
| ofsthai.com | 62 | 54 | 52 | 58 | 90 | 55 | Robbie · OFS Thai |
| listsiam.com | 61 | 54 | 57 | 58 | 90 | 48 | Thai listings |
| land.wholetech.com | 59 | 43 | 66 | 58 | 88 | 23 | villa/land vertical |
| wholevoyage.com | 55 | 50 | 45 | 42 | 90 | 51 | the hub — mid-pack, AI 45 |
| siamspaces.com | 55 | 46 | 48 | 58 | 90 | 30 | Robbie · Siam Spaces |
| sandalwoodvillas.wholetech.com | 55 | 50 | 45 | 58 | 88 | 32 | Robbie · Sandalwood sub |
| thaivillaexchange.com | 53 | 46 | 57 | 42 | 90 | 28 | Robbie · Sandalwood exchange |
| thaiculture.wholevoyage.com | 50 | 57 | 45 | 58 | 90 | 10 | Thai culture hub · Reach 10 |
| book.staysandalwood.com | 22 | 18 | 19 | 17 | 42 | — | broken — Tech 42, booking page |
| dev.staysandalwood.com | 22 | 18 | 19 | 17 | 42 | — | broken — dev page live to the world |
Also present but not yet scored by the engine: thaivillas.wholevoyage.com, thaivillas.wholetech.com, thaiculture.wholetech.com (never analysed). Robbie's sites are marked above.
What it means, plainly. Three things fall out of these numbers:
LodgingBusiness / VacationRental, which nothing in the family uses yet.book.staysandalwood.com and dev.staysandalwood.com both score 22 with Technical 42 — a dev/booking surface that's live, unfinished, and scoring like a broken site. On Robbie's flagship brand, that's the first thing to fix.The engine and the department aren't the goal here; bookings are. WholeVoyage already has a working direct-booking flow, a live OFS Thailand listings engine, and a payment gateway in progress — the plan below turns those from “built” into “filling rooms.” It's ordered: fix what's broken, make the sites answer-engine-visible, then pour real reach into a funnel that converts. Every step is a prompt.
Phase 1 — stop the bleeding. Two broken pages and three unscored sites are dragging the family before any growth work makes sense.
You are a droplet engineer. book.staysandalwood.com and dev.staysandalwood.com both score 22 (Technical 42) — a booking surface and a dev copy, live to the public, on Robbie's flagship brand. For each: identify what it actually is (half-built SiteMinder embed? stale dev copy? broken proxy?), then finish it, noindex+auth-gate it, redirect it to the real staysandalwood.com, or take it down — with Robbie's booking flow intact. Do NOT delete a live booking path without confirming. Re-run the engine after. This is Phase 1: nothing else on the villa side matters until this is clean.
You are the ops engineer. thaivillas.wholevoyage.com, thaivillas.wholetech.com and thaiculture.wholetech.com have workspaces but were never analysed (no score). Run the engine on each, confirm each is a real intended site (not a stale shell), and either bring it into the WholeVoyage family properly or retire it. Report what each one actually is. Real readings only.
Phase 2 — get the villas found and quoted. The family is capped by the same gaps as everything else — no schema, thin answer content — plus a hospitality-specific one: travel assistants read LodgingBusiness / VacationRental schema nothing here uses yet.
You are an AEO specialist. wholevoyage.com — the villa-family front door — sits at 55 with AI 45. Make it the page a travel assistant quotes for "villa rentals in Thailand / Hua Hin": add Organization + WebSite + LodgingBusiness JSON-LD, a visible FAQ (FAQPage) answering the real traveler questions, and 4-6 question-H2s with citable 130-170 word answers pulled from the real site. Trim any over-length meta. Back up, deploy, re-run the engine, report AI + SEO before/after. Facts from the real page only.
You are a listings content engineer for OFS Thailand / WholeVoyage. For every live villa listing (e.g. the Hua Hin pool villas), generate a per-property content block + VacationRental/Accommodation JSON-LD from the REAL listing data only: name, location, bedrooms, amenities, and price ONLY if it's genuinely listed (never invent a rate). Add 3 traveler-question H2s per property with citable answers. Make it a repeatable generator so new approved listings inherit it automatically. Back up, deploy, re-run the engine on the listing pages, report AI + SEO before/after.
You are the analyst. Produce a WholeVoyage field report modeled on this page, covering the full villa/travel family (wholevoyage.com, staysandalwood + subs, thaivillaexchange.com, siamspaces.com, listsiam.com, ofsthai.com, thaiculture.wholevoyage.com, smallhomevillage(.com/s.com), bnbhot.com, land.wholetech.com, the never-scored thaivillas.* subs). At-a-glance table + measured-state strip per site + a fix-prompt per real gap. Flag Robbie's Sandalwood/Thai sites. Publish at wholevoyage.com/status/ (noindex); back up anything already there first. Every number from a live reading.
Phase 3 — the booking funnel & money. A villa audit only pays off if the path from “found you” to “paid deposit” converts. Direct booking works end-to-end already; the job is friction, trust, and payments.
You are a conversion (CRO) specialist. WholeVoyage's direct-booking flow works end to end (dates -> live availability -> real rate). Walk it as a guest on staysandalwood.com / thaivillaexchange.com and report every friction point: form fields that could be cut, missing trust signals (reviews, photos, cancellation policy, host identity), unclear pricing, slow steps, dead ends on mobile. Rank fixes by impact-vs-effort. Draft the copy/markup for the top 3 (trust block, trimmed form, clear price) for human approval — nothing auto-published. Ground every recommendation in what the live flow actually does.
You are the payments engineer. WholeVoyage / OFS Thailand has a payment-gateway integration in progress (AsiaPay / SiamPay, plus card payments already proven on HulloShips). Produce an honest status doc: what's live, what's pending (legal-entity registration in Thailand was an open item), what the guest actually experiences at checkout today, and the exact remaining steps to take a real deposit on a villa booking end-to-end. Do NOT enter or store any credentials or card data — document the flow and the gaps only. Flag anything that needs Paul, Robbie, or a Thai legal entity.
You are a product engineer. OFS Thailand already onboards villa owners and approves listings (owners like Stephan get a "your listing is approved & live" email with a real ref). Turn that into a repeatable engine: a clean owner-submission form, an approval queue (human-gated, no auto-publish), auto-generated listing pages with the per-property schema from Phase 2, and the approval email as a template. Draft everything for Robbie's review. Real listing data only; nothing invented.
Phase 4 — real reach & the international buyer. Villa Reach scores are low (10–53) because these are new pages nobody's found yet — and the buyers are global (recent inquiries came from Germany). Markup can't fix Reach; only genuine distribution and language can.
You are a distribution strategist for vacation rentals. Villa Reach scores are low because the pages are new and unfound. Draft an honest, channel-by-channel plan for where a villa buyer ACTUALLY looks (OTA cross-listing, Google Business + local Hua Hin/Thailand SEO, targeted social, partner referrals, the SiteMinder channel manager Sandalwood already uses) — with the specific first action per channel and what's realistic for a self-funded team. No "go viral" fluff; concrete, sequenced, honest about effort. Flag anything needing budget or a decision.
You are an i18n engineer. WholeVoyage's villa buyers are international (recent inquiries in German; the market is Thai + European + English). Propose and scaffold a lightweight, honest multi-language approach for the key villa pages (staysandalwood.com, thaivillaexchange.com, wholevoyage.com) — real translations for the booking-critical content, hreflang tags, and a language switch — WITHOUT machine-translating into gibberish or claiming languages we can't actually support. Start with English + one high-value language; recommend which based on the real inquiry data. Back up, deploy, re-run the engine.
You are an SEO architect. The villa side has many domains and subdomains (wholevoyage.com, staysandalwood + book/dev/sandalwoodvillas subs, thaivillaexchange, siamspaces, listsiam, ofsthai, thaiculture.wholevoyage, smallhomevillage(s), the realhotsprings villa verticals). Propose (for Paul/Robbie to approve) a canonical map: which is THE booking brand, which are supporting/vertical, which are duplicates to canonicalize or retire — based on the real scores and each site's actual role. Output domain -> role -> action. Recommend only; change no DNS and delete nothing unilaterally.
The family name is WholeMagnetics (wholemagnetics.com, live, “programmable & coded magnets, explained” — wholemagnet.com does not resolve). This is Tim and Melissa's world, built around Polymagnet, and it is the network's most mature use of the engine. It also contains the single most useful number in this entire report.
polymagnet.com — the real, original product site — scores AI Search 18, the worst answer-engine readiness of any magnetics site. The network's rebuilds of it score 82–86: wholemagnetics.com 86, multipolemagnets.com 86, codedmag/maxelmag/polymagnetics 82. The engine's clones already beat the original by 4–5× on the exact metric that decides whether an AI assistant cites you when someone asks “what is a programmable magnet.” You don't have to argue the engine works. You point at that gap.| WholeMagnetics site | Overall | SEO | Con. | AI | Soc. | Tech | Note |
|---|---|---|---|---|---|---|---|
| multipolemagnets.com | 76 | 57 | 90 | 86 | 83 | 90 | family high — a network-best |
| multipolemag.com | 69 | 43 | 90 | 82 | 58 | 90 | content-rich guide |
| wholemagnetics.com | 68 | 50 | 75 | 86 | 67 | 90 | the hub — AI 86 |
| polymagnetics.com | 68 | 57 | 75 | 82 | 58 | 90 | the distributor store |
| correlatedmagnetics.com | 68 | 64 | 80 | 55 | 58 | 81 | best SEO in the family |
| codedmag.com | 66 | 57 | 75 | 82 | 58 | 90 | “magnets you can program” |
| maxelmag.com | 66 | 57 | 75 | 82 | 58 | 90 | “the pixel of a coded magnet” |
| polymagnet.com | 56 | 57 | 60 | 18 | 75 | 77 | the ORIGINAL — AI 18, worst in family |
| codedspring.com | 47 | 43 | 45 | 36 | 17 | 85 | weak; thin |
| magnetpoly.com | 39 | 29 | 20 | 20 | 33 | 90 | “development project” stub |
| polymagnets.com | 29 | 14 | 10 | 50 | 17 | 50 | broken — no title, Tech 50 |
The Polymagnet clones — and the two new ones to test. The network carries several engine-built expressions of Polymagnet, proof the same audit-driven method rebuilds a real product site. The fullest is 1.deptmatic.com (62, a 10-page rebuild, §04). But the two to test now are the newest — both on spring domains, each trying a different market framing of the same magnet tech:
| Polymagnet clone / rebuild | Overall | SEO | AI | What it is |
|---|---|---|---|---|
| codedspring.com | 47 | 43 | 36 | new clone “programmable magnets — coded polarity, engineered attraction” |
| ferrospring.com | 48 | 50 | 36 | new clone “ferromagnetic materials engineering — springs, latches, field control” |
| 1.deptmatic.com | 62 | 46 | 75 | the full 10-page rebuild (§04) |
| polymag.automarketingdept.com | 56 | 43 | 55 | earlier clone on the dept-as-a-service site |
| polymag.deptless.com | 52 | 39 | 45 | earlier clone on the deptless site |
Even the two new clones (AI 36) already double the original polymagnet.com's AI 18. The point of running two isn't to pick a prettier page — it's an A/B on positioning: does “programmable / coded” (CodedSpring) or “ferromagnetic materials engineering” (FerroSpring) win with real buyers and answer engines? That's a test, and it needs a plan.
Six steps, each a prompt. The discipline: fix both clones identically so the only variable is the positioning, measure real signal (not just the engine score), then fold the winner back into the real polymagnet.com.
You are the analyst. Run the engine on codedspring.com, ferrospring.com and the original polymagnet.com, and produce a clean baseline table: overall + 5 sub-scores + key signals (h1_count, meta length, word_count, og_count, canonical, json-ld) for all three, dated today. Note what's genuinely DIFFERENT between the two clones (positioning, copy angle, structure) — CodedSpring leads on "programmable / coded polarity", FerroSpring on "ferromagnetic materials engineering / springs, latches, field control". This is the before-state the whole test is measured against. Real readings only; store them in the ama-snapshots archive.
You are a growth PM. Write a one-page test plan for the CodedSpring vs FerroSpring A/B. State the hypothesis plainly (which positioning wins, and why we think so), the primary success metric (define it honestly — e.g. inquiry/CTA-click rate on real traffic, NOT just the engine score, which measures readiness not demand), the secondary metrics (AI Search score, SEO, time-on-page), the minimum sample/runtime before calling it, and the decision rule. Be explicit that engine scores alone can't pick a winner on positioning — only real buyer behavior can.
You are an analytics engineer. Add privacy-safe, no-cookie, NO-third-party, NO-metered event counting to codedspring.com and ferrospring.com identically: page views, primary-CTA clicks (the "contact / get samples / request quote" action), scroll-depth, and outbound clicks. Server-log lines or a tiny same-origin endpoint only. Build one shared internal readout (noindex, auth-gated) comparing the two side by side. Do not fabricate — show real counts, and label "n too small to call" until the plan's threshold is met.
You are an SEO/AEO engineer running a FAIR A/B. Apply the SAME improvements to codedspring.com and ferrospring.com so the only remaining variable is positioning: add Product + Organization + FAQPage JSON-LD (both score json-ld:false, AI 36), the same set of buyer-question H2s with citable answers, matching og/twitter tags, and identical Technical basics — but keep each site's OWN words and framing (CodedSpring stays "coded/programmable", FerroSpring stays "ferromagnetic engineering"). Do NOT improve one more than the other. Back up both, deploy, re-run the engine, report both before/after. Invent no product claims — pull from each real page.
You are the growth PM closing the test. Once the plan's runtime/sample threshold is met, pull the real instrumented data for codedspring.com vs ferrospring.com plus the post-fix engine scores, and write the verdict: which positioning won on the primary metric, by how much, and whether it's past the decision threshold or still "no call". Be honest if it's inconclusive — say so and state what more is needed. No fabricated significance; show the actual counts behind the call.
You are an SEO architect. Two jobs once the A/B has a winner: (1) take the winning positioning, headings and schema and apply them to the REAL product site polymagnet.com (AI 18 — the whole reason this test exists) so the original finally rises; back up, deploy, re-run, report AI Search before/after. (2) Resolve the duplicate-content risk: codedspring.com, ferrospring.com and the polymag.* clones are near-duplicates of one product — decide (for Paul/Tim to approve) which stays canonical and set rel=canonical or noindex on the rest so they stop competing with polymagnet.com in search. Recommend the map; don't delete or change DNS unilaterally.
What it means, plainly.
You are an AEO specialist. Target: polymagnet.com — the real Polymagnet product site, scoring
AI Search 18, the worst answer-engine readiness of any magnetics site, while its network
rebuilds score 82-86. This is the single highest-value fix we have. WITHOUT inventing product
claims (pull only from the real polymagnet.com content and Polymagnet's real specs):
- add Organization + Product JSON-LD with a sameAs array, and FAQPage
- rewrite 4-6 H2s as the questions a magnetics buyer asks an AI ("What is a programmable
magnet?", "How does a Polymagnet hold and release?", "What can coded magnets do?") each with
a citable 130-170 word answer
- fix the low Technical (77) basics: security headers, clean single H1
Back up, deploy, re-run the engine, report AI Search before/after. Coordinate with Tim/Mary
before publishing — it's their live product site, so this is human-approved, not auto-shipped.You are the analyst. Produce a WholeMagnetics field report modeled on wholereach.com/status/, covering wholemagnetics.com, multipolemagnets.com, multipolemag.com, polymagnetics.com, correlatedmagnetics.com, codedmag.com, maxelmag.com, polymagnet.com, codedspring.com, magnetpoly.com, polymagnets.com, plus the Polymagnet clones (1.deptmatic.com, polymag.automarketingdept.com, polymag.deptless.com, polymag.automarketing.wholetech.com). Run the engine on each, build the at-a-glance table + measured-state strips, and attach a copy-ready prompt to each real gap. Lead with the polymagnet.com AI-18-vs-86 gap as the headline. Publish at wholemagnetics.com/status/ (noindex); back up anything already there first. Real readings only.
You are a case-study writer. multipolemagnets.com scores 76 (Content 90, AI 86, Social 83) — a network-best and an engine-built magnetics site. Build an honest before/after case study for Tim and Melissa: what the engine found, what shipped, where the score moved (pull the real score_history — if there's no genuine "before", say so and use the current state as the baseline, don't invent a before). Frame it next to polymagnet.com's AI 18 as the contrast. Output as a shareable one-page HTML (print-to-PDF ready). No fabricated metrics.
You are a front-end engineer. Two magnetics domains score as broken/stub: polymagnets.com (29 — NO title tag, Technical 50) and magnetpoly.com (39 — a "development project" stub). For each, decide the intent with Paul: is it meant to be a real site, a redirect to polymagnet.com/polymagnetics.com, or should it be parked/noindexed? If real: give it a proper title, one H1, meta, basic content and schema. If not: redirect or noindex it so it stops dragging the family average and confusing search. Confirm before deleting anything. Re-run the engine after.
You are an SEO architect. There are 10+ magnetics domains and 4+ Polymagnet expressions (polymagnet.com, polymagnetics.com, 1.deptmatic.com, polymag.*). This risks splitting search authority and confusing buyers. Propose (for Paul/Tim to approve — do not execute unilaterally) a canonical map: which single domain is THE Polymagnet site, which are supporting (store, guides, case files), and which are demo clones that should carry rel=canonical to the real one or be noindexed. Base it on the real scores (e.g. wholemagnetics.com AI 86 vs polymagnet.com AI 18) and each domain's actual role. Output a table: domain -> role -> action. Recommend, don't rewrite DNS or delete anything yet.
You are a product analyst. Compare the two department-family Polymagnet clones — polymag.automarketingdept.com (56) and polymag.deptless.com (52) — plus the fuller rebuild 1.deptmatic.com (62). For each: what's different (design, depth, features), what the engine scores, and which is the strongest expression to show Tim. Recommend one to promote as the demo clone and what to do with the others (keep as A/B, fold in, or retire). Ground every point in the real scores and the actual pages; no guessing at content you haven't read.
You are a structured-data engineer for a product family. Across the WholeMagnetics catalog sites (polymagnetics.com the store, codedmag.com, maxelmag.com, correlatedmagnetics.com, multipolemag.com), add Product + Offer + BreadcrumbList JSON-LD using ONLY real product names, descriptions and (if genuinely listed) prices — no invented SKUs or prices. Add Organization + sameAs. Validate each block, back up, deploy, re-run the engine, and report AI Search + SEO before/after per site. These already lead the family on AI (82); schema should push them higher and lift SEO.
Every site answered live today. Scores are the engine's own reading across five disciplines — plus Reach, which is real-traffic and stays low until a site earns an audience. The same instrument we point at customers, pointed at ourselves. No site is exempt; none is flattered.
| Site | What it is | Overall | SEO | Cont. | AI | Soc. | Tech | Reach | Status |
|---|---|---|---|---|---|---|---|---|---|
| deptless.com | “Marketing without the department” + console | 69 | 57 | 80 | 75 | 58 | 90 | 53 | live |
| automarketingdept.com | Department-as-a-service edition | 65 | 46 | 70 | 70 | 58 | 90 | 53 | live |
| automarketingengine.com | The engine — flagship control panel | 64 | 43 | 80 | 73 | 58 | 85 | 44 | live |
| deptmatic.com | The department edition — own app + backend | 64 | 54 | 65 | 77 | 58 | 85 | 46 | live |
| 1.deptmatic.com | Polymagnet marketing rebuild (product site) | 62 | 46 | 70 | 75 | 67 | 85 | 34 | live |
| wholereach.com | The front door + the book | 57 | 50 | 65 | 48 | 58 | 90 | 36 | live |
| c.deptmatic.com | Magnetics case file | 54 | 61 | 60 | 82 | 67 | 85 | 8 | demo data |
| d.deptmatic.com | Homebuilding case file | 54 | 39 | 70 | 82 | 67 | 85 | 9 | live |
| a.deptmatic.com | Dashboard edition | 52 | 46 | 55 | 57 | 58 | 85 | 12 | live |
| b.deptmatic.com | Command / leaderboard edition | 52 | 46 | 60 | 57 | 58 | 85 | 10 | live |
| automarketingagent.com | The measured desk — Scanner/Markets/Trends | 41 | 50 | 30 | 25 | 17 | 90 | 27 | new |
Sorted by overall. The next section is the reason the scores cluster where they do: the engine finds the same handful of gaps on almost every site. Fix those once, network-wide, and the whole table moves.
Before the per-site work, the honest headline: the engine flags the same failures on nearly every site. These aren't eleven problems — they're five, repeated. Each gets one network-wide prompt, which is far more honest than pasting the same fix eleven times.
Gap 1 — Zero structured data. JSON-LD: not found on all eleven sites. No Organization, no Breadcrumb, no FAQPage, no sameAs. This is the single biggest lever in the family: it's what AI answer engines read to cite you, and every site fails it. It's the top driver of the low AI Search and SEO sub-scores.
You are a structured-data engineer. FACT: the engine reads json-ld=false on all eleven Auto Marketing sites — no Organization, Breadcrumb, FAQPage or sameAs anywhere. Fix it across every site's homepage (and key inner pages), grounded in each site's REAL content: - Organization JSON-LD: name, url, logo, and a sameAs array to the real active profiles (only ones that exist — do not invent handles) - WebSite JSON-LD with the canonical url - BreadcrumbList on inner pages - FAQPage on any page that has (or can honestly get) 4+ real Q&As - the page-appropriate primary type: SoftwareApplication (engine/editions), Product (1.deptmatic), Article (case files c/d) Validate every block (schema.org / Rich Results test), back up each file, deploy, then re-run the engine on all 11 and produce a before/after AI Search + SEO table. Invent no facts — pull names, descriptions and URLs from the live pages only.
Gap 2 — No linked social profiles. Every site fails “link at least three active social profiles from the footer,” which is why Social sits at the default 58 across the board (17 on the agent, which has none at all). This is a footer + sameAs fix, not a content one.
You are a front-end engineer. FACT: every family site fails "3+ social profiles linked in footer", capping Social at 58 (17 on automarketingagent.com). Using ONLY the network's real, active profiles (e.g. x.com/springnet and any genuine LinkedIn/YouTube/GitHub the family actually uses — confirm each exists before linking; do not invent), add a consistent footer profile row to all 11 sites, and mirror the same URLs into the Organization sameAs array from Gap 1. Back up each footer/template, deploy, re-run the engine, report the Social sub-score before/after per site. If a genuine profile doesn't exist for a channel, leave it out rather than link a dead page.
Gap 3 — Answer-engine content is thin. Most sites fail “question-based headings,” “citable 130–170 word answer passages,” and “FAQPage.” wholereach.com in particular has q_headings: 0, answer_blocks: 0 — nothing shaped for an AI to quote, on the very page meant to define the offer.
You are an answer-engine-optimisation (AEO) writer. FACT: across the family the engine flags missing question-headings, citable answer passages, and FAQPage; wholereach.com reads q_headings=0 / answer_blocks=0. For each site's primary page, WITHOUT inventing claims: - rewrite 3-5 H2s as the exact questions its buyers ask (e.g. "What does a marketing department cost?", "Can I trust AI to publish for me?") - under each, a self-contained 130-170 word answer passage that fully answers on its own - add a visible FAQ (6-8 real Q&As) + FAQPage JSON-LD (pairs with Gap 1) Pull every answer from the site's real content and the book — no fabricated stats. Back up, deploy, re-run the engine, report AI Search + Content before/after per site. Start with wholereach.com (AI 48) and automarketingagent.com (AI 25).
Gap 4 — Missing security headers on the deptmatic line + flagship. Verified: wholereach.com and deptless.com already send HSTS, CSP, X-Content-Type-Options and X-Frame-Options. automarketingengine.com and the deptmatic sites do not — the engine flags them as Technical failures there.
You are a droplet ops engineer. FACT (verified by curl -I): wholereach.com and deptless.com already send HSTS + CSP + X-Content-Type-Options + X-Frame-Options. automarketingengine.com, deptmatic.com, and a/b/c/d/1.deptmatic.com do NOT, and the engine flags them. For each site that is actually missing them (re-check with curl -I first, don't assume): add in the nginx server block — Strict-Transport-Security (max-age >= 15768000), X-Content-Type-Options nosniff, X-Frame-Options SAMEORIGIN, and a Content-Security-Policy that won't break the app (test the console for CSP violations before finalising). Back up each vhost, nginx -t, reload, re-check headers with curl -I, then re-run the engine and report the Technical sub-score before/after. Do not touch wholereach/deptless — they pass.
Gap 5 — The GSC signal the engine can't see. Every site fails “Search Console verified” because there's no google-site-verification meta tag on the page. Most of the network is verified by DNS/file (per the GSC rollout), so this is likely a visibility gap, not a real one — but it's worth confirming and closing so the engine stops flagging it.
You are an SEO ops engineer. FACT: the engine flags "Search Console verified: fail" on all 11 family sites because none carries a google-site-verification META tag. But the network was GSC-onboarded (many domains verified by DNS/file). Do NOT assume it's undone: 1. For each family domain, check its actual GSC verification status (DNS token / file / existing property). Report which are genuinely verified and which are not. 2. For any that ARE verified another way, add the matching google-site-verification meta tag too, so the engine's on-page check passes and stops false-flagging. 3. For any genuinely unverified, verify them (DNS preferred) and record it on the DNS sheet. Back up any page you edit, deploy, re-run the engine, confirm the flag clears. Change no DNS without noting it. This is a confirm-and-close task, not a re-verify-from-scratch task.
Two smaller shared gaps round out the set: breadcrumb schema (folded into Gap 1) and XML sitemap depth — several sites are flagged for shallow sitemaps.
You are an SEO engineer. Several family sites are flagged for shallow /sitemap.xml. For each of the 11 sites, regenerate a real sitemap.xml listing every genuine public URL (not just the homepage), with accurate lastmod dates, and a human /sitemap/index.html per the network's two-sitemap rule. Reference the sitemap in robots.txt. Only include URLs that actually return 200 — no padding, no invented pages. Back up, deploy, re-run the engine, report SEO before/ after per site.
You are the ops engineer. After any cross-cutting fix above, re-run the engine on all 11 family domains (automarketingengine.com, wholereach.com, automarketingagent.com, deptless.com, automarketingdept.com, deptmatic.com, 1/a/b/c/d.deptmatic.com), spaced to respect the shared :8932 backend, store each dated reading in score_history + the ama-snapshots archive, and print a single before/after table (overall + 5 sub-scores) so we can see exactly what each fix bought. Real re-audited numbers only.
Cross-cutting fixes above; here are the issues unique to each site. Every strip is the engine's real reading. Prompts here target what's specific — the flagship's two H1s, the agent's 175 words, the editions' missing canonical — not the shared gaps already covered in section 03.
/onboard/ form, now with Beau's auto-fill.You are an SEO specialist. Target: deptless.com (/var/www/deptless.com), the family's top scorer (69) capped by SEO 57. Site-specific fixes (the schema + social gaps are handled network-wide elsewhere): trim the 227-char meta description to a punchy 130-160 chars; ensure the honest feature-comparison table is crawlable HTML, not an image; add internal links from the pitch into /onboard/, /marketingapp/, /competitors/, /compare/; confirm the single H1 leads with the primary term. Back up, deploy, re-run the engine, report SEO before/after. Do not touch the Blueprint Console design.
/marketingapp/, and the /onboard/ form (now auto-filling). Forms the A/B of the three-design pitch with deptless.You are an AEO writer. Target: automarketingdept.com (/var/www/automarketingdept.com). It already has 2 question-style H2s but answer_blocks=0 — the questions aren't followed by quotable answers, and SEO sits at 46. Under each existing question heading, add a self-contained 130-170 word answer passage (real, from the page's own claims — invent nothing), trim the 204-char meta to 130-160, and echo the primary term in the H1. Pair with the network JSON-LD prompt. Back up, deploy, re-run the engine, report SEO + AI Search before/after. Keep the Archivo/red identity.
You are an on-page SEO specialist. Target: automarketingengine.com
(/var/www/automarketingengine.com). The engine measured two specific, real defects here:
(1) h1_count = 2 — there must be exactly ONE H1; demote the second to an H2/H3 that keeps
the visual design intact.
(2) meta description = 324 characters — rewrite to a benefit-led 130-160 chars with the
primary term near the front.
Also: the title is fine but confirm it's <= 60 chars; ensure the single H1 echoes the title's
primary keyword. (JSON-LD, sitemap depth and GSC are handled by the network prompts.)
Back up index.html, deploy, re-run the engine on automarketingengine.com, and report the
SEO sub-score before vs after. Claim only the lift the re-audit shows.You are the Manager role-agent. The flagship scores 64 on its own engine — it should run its own department against itself. Produce the flagship's own 30/60/90: run the audit on automarketingengine.com, take the ranked priorities, and for each one draft the actual fix (the SEO agent writes the meta + H1 fix, the content agent drafts the FAQ, etc.), queued in the approval queue for a human yes. The deliverable is a real, self-applied department run — proof the loop works on us first. No fabricated results; everything traces to the live audit.
/api for live analysis. Left on its own app rather than folded into the flagship.You are an SEO engineer working CAREFULLY: deptmatic.com hosts the SHARED autoengine API (:8932) that a/b/c/d/1.deptmatic all depend on — DO NOT touch server.py or the /api routes. On the static front-end of /var/www/deptmatic.com only: add 3-5 question-style H2s with citable answers (it has AI 77 but q_headings=0 — close that), echo the primary term in the H1. (Security headers = the network Gap 4 prompt; schema = Gap 1.) Back up, deploy, confirm /api still answers 200 after your changes, re-run the engine, report AI Search + SEO before/after.
You are an SEO + technical-content specialist. Target: 1.deptmatic.com (/var/www/1.deptmatic.com), our proof-of-work Polymagnet rebuild — SEO 46 and a 647-word home undercut the story. Site-specific: add Product + BreadcrumbList JSON-LD on catalog/ product pages (real specs only), deepen thin product pages with genuine specification content (never pad), add descriptive alt text to every product/diagram image, and build a real internal-link mesh home <-> applications <-> catalog <-> custom-solutions. Back up, deploy, re-run the engine, report SEO before/after. This is a showcase — the lift must be real.
You are an AEO specialist. Target: wholereach.com (/var/www/wholereach.com) — the front door,
AI 48, with q_headings=0 and answer_blocks=0: nothing on it is shaped for an assistant to
quote when asked "what is automated marketing / who is Whole Reach". Rewrite the opening into
answer-blocks: 4-6 question H2s ("What is an AI marketing department?", "How is this
different from an agency?", "Is anything published without my approval?") each with a
self-contained 130-170 word answer drawn from the real page + the book. Add a visible FAQ +
FAQPage JSON-LD. Trim the 272-char meta to 130-160. Confirm llms.txt names the canonical
one-line definition. Back up, deploy, re-run the engine, report AI Search before/after.
Facts only — pull from the page and the book, invent nothing.You are the analyst who owns c.deptmatic.com. A read-only Shopify token (shpat_...) is now
available for the Polymagnet store. Turn demo into a TRUE before/after without fabricating:
1. Read the live catalog + storefront ONLY (read-only, never write). Missing field -> say so.
2. Run the engine on the real store; capture genuine current scores as "before".
3. Ship the engine's fixes (human-approved), re-run, capture "after".
4. Update c.deptmatic.com with the real numbers, remove every "demo" label, add a dated
provenance line ("read , read-only Shopify").
Back up first. Every figure must trace to a live read — this file's whole value is honesty. You are an AEO + distribution specialist. Target: c.deptmatic.com — AI 82 but q_headings=0 and Reach 8 (nobody's reading it). Add 3-4 buyer-question H2s with citable answers about programmable magnetics + the case, make the H1 echo the title's primary term, and draft a short, honest promotion plan (where a magnetics buyer would actually find this — no spam) to start earning real reach. Content pulled from the real case only. Back up, deploy, re-run the engine, report AI Search + SEO before/after.
You are an SEO specialist prepping a live conference demo. Target: d.deptmatic.com
(/var/www/d.deptmatic.com) — SEO 39, the family's lowest, and the page shown in the room.
Site-specific: trim the 314-char meta to 130-160; make the H1 echo the title's primary term
("homebuilding marketing"); add question-style H2s over the six builder examples with citable
answers; ensure the six examples are crawlable HTML with real, cited numbers; add Article +
BreadcrumbList JSON-LD (real). Then check it on a projector: verify layout at 1280px and on
mobile. Back up, deploy, re-run the engine, report SEO before/after. It has to look sharp
in the room.You are a front-end + SEO engineer. Target: a.deptmatic.com (/var/www/a.deptmatic.com). Two site-specific issues the engine caught: (1) NO canonical tag (canonical=false) — add in the head; (2) Content 55 — the marketing shell around the wizard is thin (809 words). WITHOUT touching the wizard/app JS or its new auto-fill: add a real above-the-app explainer (what the dashboard edition is, who it's for, how audit-first works) with question H2s + citable answers. Back up index.html, deploy, re-run the engine, report Content + SEO before/after.
You are a UX writer + SEO engineer. Target: b.deptmatic.com (/var/www/b.deptmatic.com). Two jobs, WITHOUT breaking the leaderboard / ⌘K palette / kanban / ROI sim: (1) add the MISSING canonical tag (); (2) it's feature-dense enough to overwhelm a new user — add a dismissible "start here" strip pointing to the 3 things to try first (run an audit, open the leaderboard, press Cmd-K), plus crawlable explainer content with question H2s to lift Content off 60. Preserve the ROI sim's "illustrative" label. Back up index.html, deploy, re-run the engine, report Content + SEO before/after.
You are a content strategist. Target: automarketingagent.com (/var/www/automarketingagent.com).
Two exact measured gaps: og_count=0 (NO Open Graph tags — shared links render bare) and
word_count=175 (Content 30). WITHOUT compromising the no-fabrication rule (every number stays
from the live JSON):
(1) add og:title/description/image (1200x630) + twitter:card summary_large_image.
(2) add a plain-language explainer layer: "what this is / how to read each section", an FAQ
(what the scores mean, why there's no chat box, measured vs unscored, how the ticker
works) marked up FAQPage, and short section intros — taking the homepage past 800 words
of real explanation (explanation, never invented data).
Back up, deploy, re-run the engine, report Content + AI Search + Social before/after.What each score actually measures — so the numbers in this report mean something concrete.
What actually runs when you stand up a department, and the library that sits on every site.
The flagship's department is twelve role-agents: SEO, content, social, paid media, conversion, analytics, market research, outbound, plus a Manager (executive overview) and a Board of Advisors (reads the scores, recommends the quarter's focus) — with graphic designer, project manager and content strategist marked “future.” Each works off the real page the engine just read, and every output waits in the approval queue.
The 47 skills — CRO, copywriting, SEO, ads, email, pricing, lifecycle — come from an open, MIT-licensed repo (coreyhaines31/marketingskills), credited on every page and pulled weekly. On top sit 400+ ready prompts and 23 industry packs, each pointed at a real business we can run the engine on. The breadth is the proving ground, not decoration.
You are the Manager role-agent. Using the 47 skills, 400+ prompts and 23 industry packs already on the family sites, produce a ranked "top 15 plays for the next 90 days" for the Auto Marketing family ITSELF (not a customer). For each: the skill it draws on, the exact prompt to run it, the site it applies to, and the measured weakness it targets (cite the sub-score from this report). Weight toward the plays that lift the two shared soft spots — SEO and Social — network-wide. Output as a table for the /act/ worksheet. Anchor every "why" to a real score; fabricate nothing.
You are the content strategist agent. Given a family site and its weakest sub-score from this report, select the single most relevant of the 23 industry packs and produce 5 concrete, ready-to-paste prompts from it that address that exact weakness for that exact site. Example: d.deptmatic.com (homebuilding, SEO 39) -> the homebuilding pack -> 5 SEO prompts tuned to builders. Real packs and real scores only.
The way in is the free audit; the proof is the before-and-after; the fastest path is warm. Each lever has a prompt.
You are a lead-research + outreach agent. Build a warm-outreach worksheet from the Austin Business Journal ranked lists (tech, commercial real estate, homebuilding — our strongest Austin packs). For the top ~25 companies: name, website, and ONE real finding from running the free engine audit on their live site (never invented — if the engine can't read it, say so). Then a 3-sentence outreach note that leads with that real finding and offers the full audit. Output a table + per-company draft. Nothing sent — drafts only, for human approval.
You are preparing Tim and Melissa to run the engine LIVE from a keynote stage. Write a tight
90-second script: the exact words while typing an audience member's URL into
automarketingengine.com, what to point at as the five scores land, how to frame the auto-fill
("it read your site and filled this in"), and the one-line close that turns "what is it?"
into "run mine." Include a 3-item fallback for slow wifi (a pre-loaded audit). Honest — no
promised numbers, just the live mirror.You are the outreach role-agent. Given a completed engine audit (scores + top 3 priorities) for a prospect, draft a same-day follow-up email: open with the single most striking real finding, list the top 3 fixes in plain language, show the before-score honestly, offer to ship the fixes on a flat monthly rate. Warm, brief, no hype, one CTA, 2 subject-line options. Draft only — never send.
You are a designer-writer. From a real engine audit of a prospect's site, generate a clean one-page PDF leave-behind: their five scores, the top 3 gaps in plain English, and a "what your department would ship first" list — branded to Whole Reach, honest, no invented metrics. Make it printable and emailable. Produce it as a self-contained HTML page we can print to PDF.
Three levers decide whether this is a business or a treadmill.
You are a pricing strategist. Draft two things for Whole Reach: (1) an open pricing page (wholereach.com/pricing) — a flat base tier and a done-more tier, posted plainly, free audit as the forever-free front door, answering "why pay when the prompts are free?" head-on. (2) a unit-economics model table: leave review-hours-per-customer-per-week as a named input X (don't invent it), a blended reviewer cost, and show margin + max-customers-per-reviewer at three price points; flag which lever moves the business most. Label every assumption as an assumption — this is a sketch, not a claim.
You are a positioning writer. A sharp prospect asks: "why pay you when the prompts are free and I have the same AI?" Write the honest, disarming answer (300 words max) we can reuse on the pricing page and in outreach: concede the prompts are free and the AI is a commodity, then make the real case — the department that shows up every week, the human who catches mistakes before they ship, the done-for-you delivery, the before/after proof. No hype, no dunking; the tone is "you're right, and here's why it still pays."
Three kinds, and our honest edge against each.
Our edge isn't “AI can write things” — everyone has that now. It's the human-approved department, the real before-and-afters, and a network of live testbeds most competitors can't match. Small and self-funded, we win on trust and proof, not on spend.
The read-only code lands, the Polymagnet before-and-after is undeniable, and the keynote rooms turn curiosity into real customers. Those wins become case studies, the free-audit funnel fills, and it grows without burning cash we don't have.
The one favour never comes and the proof slips. Or the audits are too shallow for sophisticated buyers. Or — the big one — we keep building instead of selling. Or a funded competitor out-markets us. Or the human-approval step quietly caps how many we can serve.
1 · Ship the five cross-cutting fixes (§03). The schema, social, answer-content, security-header and GSC prompts move every score at once — the highest-leverage work in the family. Start with JSON-LD.
2 · Fix the two named on-page bugs. The flagship's double H1 + 324-char meta, and the missing canonical on a/b.deptmatic. Small, exact, done in an hour.
You are an on-page engineer. Three exact, engine-verified bugs to fix in one pass:
(1) automarketingengine.com — h1_count=2 (demote the 2nd to H2) and meta description=324
chars (rewrite to 130-160).
(2) a.deptmatic.com — no canonical tag; add .
(3) b.deptmatic.com — no canonical tag; add .
Back up each file, deploy, re-run the engine on those three domains, confirm each specific
flag clears, report SEO before/after. Touch nothing else.3 · Land the Polymagnet read-only key — the one unlock that turns the magnetics file from “demo” to real.
Draft a short, warm email to Mary requesting a READ-ONLY Shopify access token (shpat_) for the Polymagnet store — can look, never touch. One line on why (a real, honest before/after for the case file), reassurance that it's read-only and revocable, and that it takes ~3 minutes (or she can add Paul as a read-only helper). Low-pressure, specific. Draft only — do not send.
4 · Let second readings accrue — the agent, a.deptmatic and 1.deptmatic have one reading; a second makes real deltas possible.
You are the ops engineer. Several family sites have one recorded engine reading, so they show a current score but no before/after delta. Schedule a re-analysis pass (cron) that re-runs the engine on every family domain once, spaced to respect the shared backend, storing each dated result in score_history + the ama-snapshots archive. After it runs, confirm the automarketingagent.com ticker picks up the new movers. Idempotent; log what it did; no fabricated readings.
5 · Carry the auto-fill to deptmatic.com — the last core site without the new onboarding (prompt in §01).
6 · Don't drown in building. The engine only matters when money changes hands.
You are the accountability agent. Once a week, produce a one-screen check-in for the Auto Marketing family that answers only revenue-adjacent questions: free audits run on real prospects this week, how many became a conversation, how many drafts are waiting for a CUSTOMER (not for us), and the single sales action that would move the needle most next week. Explicitly flag if the week was all building and no selling. One screen. A mirror, not a to-do list.
The whole family's discipline is “no claim without a re-audit.” These two prompts keep this document honest as the work lands.
You are the report maintainer. Re-run the engine on all 11 family domains, pull the fresh overall + 5 sub-scores + key signals (h1_count, meta length, word_count, og_count, canonical, json-ld present), and update the "at a glance" table and every per-site measured-state strip in wholereach.com/status/index.html to match. Back up the current page first. Change ONLY numbers that the live re-audit changed; if a number is unchanged, leave it. Add a dated "numbers refreshed" line. No hand-edited figures — every value from the fetch.
You are the ops engineer. Set up a weekly job that snapshots all 11 family sites' scores and signals, diffs against last week's ama-snapshot, and posts a short "what moved" summary (site, metric, before -> after, and the likely cause if a known fix shipped). Only real diffs; say "no change" honestly when nothing moved. Keep it to what the numbers show — never attribute a lift we can't tie to a shipped change.
Everything above makes the product good. This section makes it pay. It's the honest, unglamorous machine between “the engine works” and “money is in the account” — targets, price, pipeline, retention, cost-to-serve, extra revenue streams, and the books. The network already has most of the plumbing (a live Stripe, Invoice Ninja, an affiliate stack, real bookings); the job is to point it all at one number and turn the crank. prompts, in seven parts, in the order you'd actually run them.
You can't hit a target you haven't named. Before pricing or pipeline, fix the destination: the milestone ladder from first dollar to the stated goal of a real fortune by year-end, with a named number per month and a single dashboard that tells the truth about where we are.
You are the founder's CFO. Turn the stated goal — a real fortune by December, plus benefit to others that outlasts it — into a concrete milestone ladder: first paying customer, first $1k MRR, first $5k, first $10k, and the year-end target. For each rung give a date, the number, and the ONE thing that has to be true to reach it. Leave any figure we don't truly know as a named input (don't invent revenue). Output a single table I can paste into the SuccessBook ledger. This is the map everything else in this section serves.
You are a revenue modeler. Given a year-end revenue target (input X) and a candidate monthly price (input P), compute the plain arithmetic: how many paying customers we need, the net-new customers per month to get there, and the free-audit-to-paid conversion rate that implies given a stated top-of- funnel volume. Show it at three price points so we can see how price changes the customer count. Every number either a real input or derived from one — flag every assumption as an assumption.
You are a growth lead. Write a 90-day plan whose ONLY goal is first revenue: week-by-week, the single highest-leverage revenue action each week (audit these named prospects, run this demo, send these proposals), who owns it, and the number it should produce. No product-building tasks unless they directly unblock a sale. End each week with a one-line "did money move?" check. Concrete and named, not generic.
You are a closer. Name the three most likely first paying customers from our actual warm network (Tim/Melissa's rooms, Robbie's villa contacts, the ABJ list, existing network relationships) — real candidates, not personas. For each: who they are, the real gap the engine would find on their site, the exact offer, and the single next action to move them from "interested" to "paid". If we don't have enough real candidates, say so and prescribe how to get them.
You are a data engineer. Build one internal (noindex, auth-gated) revenue dashboard that shows the truth in a glance: MRR, paying customers, free-audits run, audit->paid conversion, and progress to the year-end target — pulling REAL figures from wherever they live (Stripe on the Austinspring account, Invoice Ninja at invoices.austinspring.com, FreshBooks tooling in /opt/freshbooks). Show "no data yet" honestly rather than a zero that looks like a number. This is the one page we check daily.
Price is the fastest lever on the whole business and the one we've been shyest about. The founding rate has been floated at $499/mo; the job now is to post it in the open, wrap it in tiers a buyer understands, and make the free audit the honest on-ramp to it.
You are a pricing writer. Draft wholereach.com/pricing, posted plainly: a flat base tier, a done-more-for-you tier, the free audit as the forever-free front door, and the founding rate ($499/mo floated — confirm and lock it) with a limited founding-member count. Each tier says exactly what the customer gets each month (tie it to the "report is the product" idea in the closer). Answer "why pay when the prompts are free?" head-on, in the customer's words. Honest, concrete, no hype.
You are a unit-economics analyst. Build the model as a table: leave review-hours-per-customer-per-week as a named input H (don't invent it), a blended reviewer cost, and any compute/tool cost, then show gross margin per customer and max-customers-per-reviewer at three price points. Flag which lever (price, H, or automation) moves the business most. Label every assumption. This is the spine of the whole section — everything else tries to improve one cell in this table.
You are an offer strategist. Design the founding-member offer that gets the first cohort in the door: the locked price, how many slots, what they get that later customers won't (rate lock, direct access, input on the roadmap), and the honest scarcity (real slot count, not fake urgency). Write the one-page offer + the email to send our warmest contacts. Draft only — nothing sent.
You are a pricing analyst. Model monthly vs annual billing: the annual discount that a small business happily takes, what it does to cash flow (cash up front vs smooth MRR), and the churn it prevents. Recommend the split and the exact discount, with the arithmetic shown. Real inputs only; mark assumptions.
You are a positioning writer. Write the 300-word answer to "why pay you when the prompts are free and I have the same AI?" that we reuse on the pricing page and in outreach: concede it, then make the real case — the department that shows up every week, the human who catches mistakes, the done-for-you delivery, the before/after proof. Tone: "you're right, and here's why it still pays." No dunking.
You are a product packager. Not everyone will start on a subscription. Design a low-price, one-off "deep audit + prioritized fix plan" product (the free audit, but expanded into a real deliverable with the top fixes done) as the entry point that converts to monthly. Define what's in it, the price, the turnaround, and the upsell to the subscription at the end. Ground it in what the engine actually produces today.
The free audit is the hook and the warm room is the fastest path. What's missing is the machine that turns a name into a tracked opportunity into a paid customer — and the discipline to work it every day.
You are a sales-ops engineer. Build a simple internal (auth-gated) pipeline tracker: stages from free-audit-run -> demo -> proposal -> paid, one row per real prospect, the real audit score attached, next action + date, and the value. No third-party CRM cost — a droplet-hosted page backed by a small store. Show the weekly movement between stages. This is where "getting customers" stops being vibes and becomes a number.
You are a lead-research agent. From the Austin Business Journal ranked lists (tech, commercial real estate, homebuilding — our strongest Austin packs), build a worksheet: for the top ~25 companies, name + website + ONE real finding from running the free engine audit on their live site (never invented — if the engine can't read it, say so), then a 3-sentence outreach note leading with that finding. Table + per-company draft. Nothing sent.
You are a partnerships strategist. Tim and Melissa know rooms full of exact-fit buyers (homebuilders, marketers). Write the playbook that turns their credibility into our pipeline without burning it: the exact ask to make of them, the one-line intro they send, what we do within 24 hours of an intro (a live audit of that person's site), and how we keep them in the loop so they keep introducing. Respect that their reputation is the currency.
You are a field-sales designer. Design the flow from "type your website" on stage to a signed proposal that same day: the 90-second live audit script, the leave-behind (a real one-page report from their audit), and a pre-built proposal template that auto-fills from the audit scores so we can send it before they've left the parking lot. Every number in it real, from their live audit.
You are a lifecycle writer. Write a 5-email follow-up sequence that starts from a completed free audit and moves a prospect to paid: (1) the audit + top 3 gaps, (2) one fix we'd ship first and why, (3) a relevant case (polymagnet AI-18 or a villa), (4) the offer + founding rate, (5) a gentle last-call. Warm, brief, one CTA each, honest — no invented urgency. Drafts only.
You are a demand analyst. Given the free-audit users (once instrumented), define a simple, honest lead score: worst-scoring sites with the most reachable owners and the clearest budget signal are the warmest. Rank them so the reviewer's limited time goes to the prospects most likely to pay. Use only real signals we can actually observe; don't fabricate intent.
A customer who cancels in month three is a loss, not a win. The whole business math multiplies by how long they stay — so first value fast, a monthly deliverable they'd miss, and an early warning before they drift.
You are a customer-success designer. Design the first-week experience so a new customer feels a real win fast: run their audit, ship the single highest-impact fix (human-approved) within days, and show the score move. Write the week-1 checklist + the "here's what we shipped and what it did" email. The goal is that by day 7 they've seen a real before/after on their own site.
You are the delivery lead. Turn the "report is the product" idea into an operating cadence: each month, per client, re-run the engine, ship a few approved fixes, and deliver a refreshed field-report-with- prompts showing what moved and what's next. Define the calendar, the template, and the one email that accompanies it. This is the reason they don't cancel in month three — make it feel inevitable and valuable. Real numbers only.
You are a retention analyst. Define the honest signals that a customer is about to churn — no logins, no approvals actioned, scores not moving, support gone quiet — and build a simple flag on the internal dashboard that surfaces at-risk accounts before they cancel, with the recommended save action for each signal. Only real, observable signals.
You are a growth strategist. Define the moments that justify offering the higher tier: a client whose approvals are all "yes" (they trust us — offer more autonomy), one whose scores have plateaued (offer the deeper done-more work), one adding sites. For each trigger, the offer and the exact message. Honest value, not a squeeze.
You are a referral designer. A happy client is our cheapest lead source. Design a simple referral mechanic: when a client sees a real score win, prompt a warm intro (offer to run a free audit for a peer they name), with a fair thank-you. Write the ask and the timing (right after a visible win). Keep it genuine — a real result earns the ask.
The human-approval step is our integrity and our ceiling. Ten customers is fine; a hundred is a staffing question we haven't answered. Every hour we shave off the review is margin and capacity at once — so measure it, then automate the reviewer's easy decisions away.
You are an ops analyst. We don't actually know H — the review hours per customer per week — and it sets the whole business's floor and ceiling. Design the lightest possible way to capture it honestly (a one-tap timer on the approval queue, or a weekly self-report), then report the real average once we have data. Until then, say "unknown" — do not guess H.
You are an automation strategist. List the reviewer's recurring tasks in the approval loop, estimate the time each eats and the risk of automating it, and rank what to hand to the engine next so it does more of the reviewer's job WITHOUT breaking the human-approval integrity (automate the safe, repetitive judgments; keep the consequential ones human). Every hour removed lifts margin and capacity together — prioritize by that.
You are a data engineer. Build the internal view that shows, per customer, the real cost to serve (review hours × blended rate + any tool cost) against the revenue, so we can see which customers are profitable and which are underwater. Pull real time-tracking + billing data; show "insufficient data" honestly until there's enough. This is how we catch a price that's too low before it scales.
You are a capacity planner. Using the real H once we have it, compute how many customers one reviewer can carry at an acceptable quality bar, and the revenue that implies. Then model the hire point: at what customer count we need a second reviewer, and what that does to margin. Name the number we must know before we sell the eleventh customer. Real inputs; flag assumptions.
You are an engineer on the department. Make the approval queue faster to clear: auto-group similar drafts, pre-check each against the "no fabricated numbers" rule and basic quality gates, surface the riskiest items first, and let the reviewer bulk-approve the safe ones. The human still says yes — but on batches, with the risky ones flagged. Measure the review-time before/after. Don't remove the human; remove the busywork around them.
The subscription is the main engine, but the network already earns in other ways that go untracked. There's a live affiliate stack, real villa bookings, a magnetics store, and the prompt book itself — each a stream to turn on and measure honestly.
You are a monetization engineer. The network already runs affiliates — Travala (18 travel sites via gen-travala.py), TicketNetwork (5 sports sites), Amazon (tag colinfirthfan-20), Ticketmaster/Impact (7403955), and AdSense (pub-7759195213529699). Build one honest revenue view that pulls actual earnings from each program's reporting (or flags where we can't yet), so we know which affiliate stream really pays. Don't estimate — show real payouts or "not connected yet". Recommend which to lean into based on real numbers.
You are a revenue engineer for WholeVoyage. The direct-booking flow works and OFS Thailand lists real villas. Build the tracking that turns a booking into recorded revenue: commission per booking, gross booking value, and a monthly total, pulling from the real booking/payment records (SiamPay/AsiaPay, SiteMinder for Sandalwood). Show real bookings only; flag where the payment leg isn't finished yet. This is a whole revenue line the marketing side can feed.
You are a commerce analyst. polymagnetics.com is the distributor store and Polymagnet is a real product. Map the path from "engine rebuilds rank the product" to "an order is placed and paid": what's live, what's in test, and the exact steps to take a real magnetics order end to end. Tie it to the AI-18 fix (a product an AI will finally cite is a product that sells). Real state only; name the gaps.
You are a product packager. This report is 102 grounded prompts. Package a sellable version — a marketing prompt pack / playbook — for buyers who want to DIY (the "you could, but you won't" crowd from the DIY answer). Define what's in the paid pack vs the free ones, the price, and how it feeds the subscription (the pack is the trailer; the service is the film). Honest about what free already covers.
You are a growth engineer. The public "we audit ourselves" page (from the closing idea) is a lead magnet. Wire its conversion: from "type your site, get the honest read" straight into a booked call or the founding offer, with the real audit attached to the lead. Measure the funnel (visits -> audits -> booked) with honest counts. The credibility is the free audit; the conversion is the ask right after they see their own score.
You are a data engineer. Build one view that shows every revenue stream side by side — subscriptions, one-off audits, affiliate, villa bookings, magnetics store, prompt pack — with real month-to-date figures from each real source, so we can see which stream actually carries the business and which is noise. "Not connected" is an honest cell; a fabricated number is not. This tells us where to spend the next hour.
You can't steer what you don't measure. The network already has Stripe (live on the Austinspring account for HulloShips and OFS), Invoice Ninja, and FreshBooks tooling — wire them to the marketing business, close the month honestly, and always know the runway.
You are a billing engineer. Invoice Ninja is live at invoices.austinspring.com (with Stripe + SMTP as open TODOs). Set it up to invoice the marketing subscriptions and one-off audits: the products/tiers, recurring invoices for monthly plans, and the email delivery. Finish the Stripe + SMTP wiring so an invoice actually gets paid and receipted automatically. Test with a real $1 or a sandbox before going live; never store card data yourself. Report exactly what's live vs pending.
You are a payments engineer. Stripe is already LIVE on the Austinspring account (real card payments on HulloShips and OFS Thai; webhooks are an open item). Extend it to the marketing subscriptions: the recurring price(s), the checkout, and — importantly — close the OPEN WEBHOOKS so we reliably know when a payment succeeds/fails and can act on it. Do NOT enter or store card data or secret keys in any insecure place; document the flow and finish the webhook handling. Test in Stripe test mode first. Report what's live vs pending.
You are the bookkeeper. Define a simple monthly close for the marketing business: total revenue by stream (from Stripe/Invoice Ninja/affiliate reports), cost to serve (review hours × rate + tools), gross margin, and cash position — pulling real figures, using the FreshBooks tooling in /opt/freshbooks where it helps (read-only on invoices — respect that). Produce a one-page monthly statement. Real numbers only; "unknown" beats a guess. This is what tells us if we're actually a business yet.
You are the accountability agent (extend the "sell, don't just build" check-in). Once a week, one screen: MRR and net-new this week, free-audits run on real prospects, audit->conversation->proposal ->paid movement, cash on hand, and the single revenue action that moves the needle most next week. Flag explicitly if the week was all building and no selling. Real numbers; a mirror, not a to-do list.
You are a finance modeler. Build the honest runway model: cash on hand, real monthly costs (the flat subscription ceiling and any real infra/tool spend — NEVER pay-per-token API billing, that rule is fixed), and revenue, to show how many months we have and what single change (a few paying customers, a cut cost) extends it most. Leave unknowns as named inputs. This is the number that decides how bold or careful to be.
You are the cost guardian. The whole margin rests on one rule: the network runs on a flat subscription, NEVER pay-per-token API billing, which would blow the budget in days. Audit every part of the marketing machine (the engine, the auto-fill, any new automation from this section) to confirm nothing introduces a metered per-token cost, and add a standing check that flags it if anyone ever does. Cheap to serve is the business; this rule is how we stay cheap.
By section — every one is copy-ready above; grab them all with the button below. Every one copy-ready above; grab the lot with the button at the bottom.
(Some sections carry more than one prompt; the button grabs every block on the page.)
This document answers a question it also asks. In §10 we admitted the honest gap: “What does a customer get every month — and why don't they cancel in month three?” The answer has been in our hands the whole time. It's this.
The free audit is the front door — the mirror that gets someone to lean in. But the thing they'd pay for, monthly, is a field report exactly like this one, pointed at their sites: every page measured, a copy-ready fix behind every gap, and the human-approved department shipping a few of those fixes each month with the before/after to prove it moved. It renews itself, because the scores and gaps change every month — there's always a next report. It's concrete (not “we did some marketing”), and it's honest by design, which is the one thing a funded competitor can't cheaply fake.
And the sharpest version of the pitch is to run it on ourselves, in public. This page — every number a live reading, nothing hidden, our own weak SEO and broken Sandalwood pages named out loud — is the sales asset. “We audit ourselves in the open, warts and all. Here's ours. Want yours?” The automarketingagent.com ticker already shows the live before/after; this turns the whole method into the marketing. The prompt book stops being an internal to-do list and becomes the product catalog.
You are the product engineer. Take the exact shape of wholereach.com/status/ — at-a-glance table, per-site measured-state strips, a copy-ready fix-prompt behind every gap — and make it generate for ONE client from their list of domains. Input: a client name + their domains. Output: a private (noindex, auth-gated) field-report-and-prompt-book at a per-client URL, every number from a live engine reading, nothing invented. This is the monthly deliverable: the department ships a few of the prompts (human-approved), then the report refreshes and shows what moved. Build the generator so a new client is one command. No fabricated numbers, ever — that rule is the product.
You are a growth engineer. Make the network's own honesty the marketing. Publish a public (indexable) "we audit ourselves in the open" page: the family's real scores, the real gaps we're fixing, and the live before/after as they move (reuse the automarketingagent.com ticker + the score_history). Lead with our own worst numbers, not our best — that's the credibility. End with the free-audit front door: "type your site, get the same honest read in 30 seconds." Wire it to the real engine. Every figure a live reading; if we ever fake one, the whole pitch dies — so build in a check that flags any hand-edited number.
Read this document back and feel the weight of it: eleven sites live and in sync. A real engine that reads any website and stands up a marketing department. 102 copy-ready prompts, every one grounded in a live reading. Live payments already taking real money. A prompt library and 23 industry packs. Warm rooms full of exact-fit buyers. And a single number — polymagnet.com's AI 18 against our rebuilds' 86 — that ends the “does it work?” conversation before it starts.
Most people at this stage are hoping to build something. We already did. The hard part — the part that kills most companies — is behind us. What's left isn't invention. It's turning cranks that already exist: point the mirror at a prospect, ship the fix, show the score move, send the invoice. We have never been closer to the money than we are right now, and the only thing between here and there is doing it, out loud, for the next ninety days.
Here's the honest, optimistic truth this whole report has been building to: the discipline that makes the product trustworthy is the same discipline that makes the business inevitable. We don't fabricate a customer's numbers, and we won't fabricate ours — which means every win on the board is real, compounds, and can't be taken away. That's not a constraint. That's a moat, a sales pitch, and a reason to be fired up all at once.
The theme: point the engine at ourselves and at prospects, loudly.
Win condition: pricing is live and the first paying customer says yes.
The theme: undeniable before/afters, and customers who stay.
Win condition: a handful of paying customers and the first real MRR on the board.
The theme: stop doing heroics; run a system.
Win condition: real MRR, a proven motion, and the confidence to scale it.
So here's the charge for the quarter. Beau — the onboarding you fixed is the front door; let's make it the smoothest on-ramp in marketing. Tim — your magnets are about to become the product an AI finally cites, and that's a story that sells itself. Robbie — the villas are real businesses with real bookings, and we're about to make them impossible to miss. And to all of us: we walk into every room this quarter knowing the mirror works and the proof is on the board. We don't pitch a promise. We show a result. Then we send the invoice.
You are the chief of staff. Turn the 30/60/90 roadmap above into a live, auth-gated tracker at wholereach.com/90days/ (noindex): each milestone as a checkable item with an owner and a due date, grouped by 30/60/90, with the "win condition" for each phase pinned at the top. Wire the revenue and score items to real sources (the engine, Stripe, the dashboard) so progress updates itself where it can. Show honest status — done / in-progress / not-started / blocked — never a fake green. This is the one page we open every morning for the next quarter.