The voice coach helps you explain the weak period without blame, defend a staged budget, choose the campaigns to propose, select the strongest client-facing assets, and rehearse difficult questions. It is grounded in the refreshed Google Ads evidence, the rebuilt client deck, the strategy council, the persona work, and the competitor research.
Ask it to run a seven-minute rehearsal, role-play the client, challenge the proposed 40/35/15/10 allocation, or help decide what belongs in the main presentation versus the appendix. Voice feedback is read-only and cannot approve or execute work.
The private room uses one Sonnet engine for the whole conversation. If the connection drops, it retries the same engine twice and reloads the persisted conversation. It never changes models during a conversation.
Sonnet 5 · no fallbackOpen the fresh private Sonnet meeting-coach link sent to your Telegram, then allow microphone access. The reusable grant is deliberately absent from this public review page.
Voice answers are review evidence. Approval, sending, spending, and production changes stay in the originating chat. Do not forward the private Telegram link.
Plan: /Users/agency/Documents/Agty/Analit/action-plans/2026-08-09-analit-priority-funnel-orchestration.md
J or K, to open it.1 delegate, 2 drop, 3 idea, 4 ask first, 0 clear, J and K move between tasks.! to highlight important feedback. Comments do not authorize work.The private voice room is grounded in a fixed capsule containing 19 selected source files. It can explain and challenge the material below, but it cannot read live ad accounts or make changes during the call.
meeting-preparation-playbook.md, source-inventory.md, the reviewed action plan, the morning ideas, and the strategy council synthesis.Start with T-018. Its commercial decision controls the measurement and campaign work that follows.
Use Google Slides for historical performance and the HTML pages for strategy choices. The competitor pages below are the preferred comment surfaces.
Design a proposal-first before-and-after measurement story from campaign source through paid booking, showing both the Google Sheet and QlickCRM variants plus a future remarketing-list design.
The team needs one visible measurement story before it can decide how the Google Sheet and QlickCRM should share outcome feedback.
Attribution stays ambiguous, implementation can follow the wrong architecture, and Matt loses the overview he needs.
A client-safe before-and-after proposal makes the tradeoff, identifiers, and remarketing use visible before any system is changed.
A client-safe proposal shows current and proposed measurement, both source-of-truth variants, persona identifiers, remarketing design, admin effort, privacy, failure recovery, and a recommendation. No live wiring. 80/20. One before-and-after Mermaid pair and one decision table.
flowchart LR ADS[Campaign source] --> LEAD[Enquiry record] LEAD --> SHEET[Google Sheet feedback] SHEET --> OUTCOME[Quote and paid booking] OUTCOME --> VIEW[Human readable funnel]
flowchart TD
S1[Inspect both systems] --> S2[Map fields and owners]
S2 --> S3[Draw both architectures]
S3 --> S4[Compare risks and effort]
S4 --> D{Decision evidence complete}
D -- yes --> READY[Ready for human choice]
D -- no --> S1
Turn every material weakness from the 30-day report into a verified repair proposal, evidence-backed access gate, or historical limitation while keeping this round read-only.
Missing campaign fields, proxy engagement, unclear events, incomplete funnel coverage, and access gaps can make the funnel misleading.
The agency may optimize against incomplete or wrongly interpreted evidence and cannot prove where enquiries came from.
Every measurement stage receives a business definition, source, test, owner, and freshness boundary.
A dated vulnerability matrix shows a repair proposal, isolated test, or clear block for each weakness. 80/20. Campaign-source capture, exact 30-second engagement, and guide-download conversion are ranked without a live write.
flowchart LR GAPS[Report weaknesses] --> TEST[Current evidence checks] TEST --> FIX[Approved measurement repairs] FIX --> PROOF[Verified funnel evidence]
flowchart TD
S1[Reproduce every limitation] --> S2[Define event meaning]
S2 --> S3[Back up and repair]
S3 --> S4[Run positive and negative controls]
S4 --> D{Read back matches intent}
D -- yes --> READY[Document visual proof]
D -- no --> S3
Compare GFN with suitable third-party call-tracking approaches and prove the recommended path on staging.
A phone-link click is not a connected or qualified call, and a forwarding-number visual alone does not prove attribution.
Calls remain unattributed or overcounted, while an unsuitable provider can introduce routing, privacy, or cost risk.
A reversible proof connects campaign, session, call duration, quote, and paid booking without risking the live number.
A current-source matrix covers Hungarian availability, dynamic insertion, number ownership, duration, consent, integrations, reliability, rollback, and cost, followed by a staged proof. 80/20. Two call-tracking architectures, one provider matrix, and one verified staged call-event chain.
flowchart LR VISIT[Campaign visit] --> NUMBER[Dynamic phone number] NUMBER --> CALL[Connected call] CALL --> QUALIFY[Qualified duration] QUALIFY --> OUTCOME[Quote and paid booking]
flowchart TD
S1[Inspect current phone path] --> S2[Research GFN and providers]
S2 --> S3[Compare privacy cost and reliability]
S3 --> S4[Build staging proof]
S4 --> D{Qualified call chain proven}
D -- yes --> READY[Recommend architecture]
D -- no --> S2
Stage the approved measurement, attribution, and call-tracking paths through quote, deposit, and paid booking. No production wiring in this round.
The business currently lacks a verified chain showing which campaigns create quotes and paid customers.
Budget decisions remain based on upstream activity, and the agency cannot demonstrate commercial outcomes.
A live feedback loop and professional before-and-after page make both performance and agency value visible.
An isolated consent-safe test journey proves the proposed source capture through paid-booking transitions with stable persona identifiers and responsive before-and-after evidence. Production wiring remains a later approval. 80/20. One deterministic test record crosses the isolated chain.
flowchart LR ADS[Campaign] --> ENGAGE[Page engagement] ENGAGE --> ENQUIRY[Form or qualified call] ENQUIRY --> QUOTE[Quote] QUOTE --> PAID[Paid booking]
flowchart TD
S1[Confirm approved architecture] --> S2[Back up and implement joins]
S2 --> S3[Prove enquiry and quote]
S3 --> S4[Extend to paid booking]
S4 --> D{Read back and retries pass}
D -- yes --> READY[Publish client visual]
D -- no --> S2
Create one complete persona-led funnel example with a stable identifier, plus six clearly labelled candidate personas for the ongoing pipeline. Reuse the existing read-only Sonnet voice reviewer.
Campaigns cannot be designed coherently by choosing creatives without connecting persona, message, page, capture, and follow-up.
One generic funnel continues to serve different motivations, and the team risks overbuilding the first idea it sees.
Matt can inspect one complete example, see the repeatable process, and discuss it by voice before deeper work.
One complete persona hypothesis and six labelled candidates include stable IDs, a funnel, CTA, up to three safe lead-magnet directions, form, email, and outcome join. The page has comments and the existing Sonnet voice review. 80/20. One end-to-end example and a future-work roadmap.
flowchart LR EVIDENCE[Calls and enquiries] --> IDEAS[Seven persona candidates] IDEAS --> REVIEW[Comments and voice review] REVIEW --> SELECT[Selected funnels]
flowchart TD
S1[Ground in source evidence] --> S2[Create seven candidate hypotheses]
S2 --> S3[Map one complete funnel]
S3 --> S4[Publish commentable HTML and voice]
S4 --> D{Human review complete}
D -- yes --> READY[Pass feedback to next stage]
D -- no --> S4
Transform submitted comments, selections, and the voice transcript into the next iteration of the chosen funnels.
A broad idea review only creates value when feedback becomes traceable design decisions.
The team can lose why ideas were accepted or rejected and may overbuild concepts Matt did not choose.
Selected funnels become coherent with an explicit receipt for every binding choice, suggestion, rejection, and uncertainty.
A feedback receipt maps comments and voice decisions to revised end-to-end funnels, and the republished HTML supports one more review round if needed. 80/20. One selected persona refined end to end with a complete feedback trace.
flowchart LR INPUT[Comments and voice] --> RECEIPT[Decision receipt] RECEIPT --> REFINE[Selected funnel refinement] REFINE --> REVIEW[Next human review]
flowchart TD
S1[Save comments and transcript] --> S2[Classify decisions]
S2 --> S3[Refine selected funnels]
S3 --> S4[Regenerate review HTML]
S4 --> D{Feedback trace complete}
D -- yes --> READY[Open next review]
D -- no --> S2
Turn the measurement proposal, one persona example, one Hungarian competitor example, one US competitor example, and the 30-day findings into one professional HTML presentation that separates evidenced work from proposed approval-gated work.
Analit needs a connected business story, not a collection of internal tasks and isolated creative ideas.
Good work remains invisible, while campaign components are approved without understanding the whole path to paid booking.
The client can see what was observed, what will change, how the funnels work, how success will be measured, and which values are mock examples.
A responsive client HTML connects dated evidence to the one persona and two competitor examples, with creative, targeting, pages, lead magnets, forms, emails, measurement, mock dashboard, timeline, next-meeting agenda, and approval gates. 80/20. One complete persona funnel plus the overall strategy proposal.
flowchart LR FINDINGS[Dated findings] --> STRATEGY[Selected funnel strategy] STRATEGY --> CLIENT[Client presentation] CLIENT --> APPROVAL[Approved execution scope]
flowchart TD
S1[Combine approved evidence] --> S2[Draft complete campaign paths]
S2 --> S3[Build before and after visual]
S3 --> S4[Verify claims and breakpoints]
S4 --> D{Client presentation ready}
D -- yes --> READY[Request approvals]
D -- no --> S2
Build Meta drafts, pause for human review, and leave publication as a later separately approved phase. No live Meta mutation in this review round.
The strategy needs a safe route to execution without autonomous creative, targeting, budget, or launch decisions.
The plan remains theoretical while weak or disconnected campaigns continue unchanged.
Approved funnels become live through controlled checkpoints, evidence receipts, rollback information, and post-launch verification.
Draft previews and an exact change manifest are shown before approval. 80/20. One persona campaign exists as a complete draft and is not published in this round.
flowchart LR PLAN[Approved strategy] --> DRAFT[Meta drafts] DRAFT --> HUMAN[Human approval] HUMAN --> LIVE[Limited live release] LIVE --> PROOF[Post launch proof]
flowchart TD
S1[Create exact change manifest] --> S2[Build drafts and previews]
S2 --> S3[Request human approval]
S3 --> D{Exact items approved}
D -- yes --> S4[Publish and read back]
D -- no --> S2
S4 --> READY[Record proof and rollback]
Create a general questionnaire-style lead path for SEO and referral visitors who do not enter through a persona campaign.
Organic and referral visitors lack a structured way to express their needs, which weakens segmentation and follow-up.
Existing traffic remains harder to qualify and easier to lose before a useful conversation begins.
Analit gets a low-friction enquiry experience that improves context while preserving clear call and quote paths.
A consent-safe staging flow maps minimum useful answers to segments, verifies positive and negative submissions, and includes a responsive before-and-after handoff. 80/20. A clickable staged questionnaire with one verified segment output.
flowchart LR VISITOR[SEO or referral visitor] --> QUESTIONS[Short questionnaire] QUESTIONS --> SEGMENT[Useful segment] SEGMENT --> ACTION[Call or quote request]
flowchart TD
S1[Inspect current homepage] --> S2[Map minimum questions]
S2 --> S3[Build staging flow]
S3 --> S4[Test positive and negative paths]
S4 --> D{Segment output verified}
D -- yes --> READY[Publish visual handoff]
D -- no --> S3
Present the existing collapsible-biography prototype, capture the interaction decision, and release only after approval.
The client request has a prototype but no personal review, interaction decision, approval receipt, or live implementation.
Long biographies remain difficult to scan while completed prototype work stays disconnected from the request.
Matt gets a quick visual decision and a narrow reversible route to the live site.
The prototype review records single-open or multiple-open behavior, and only an approved release proceeds through backup, staging, responsive QA, live verification, and rollback notes. 80/20. A working prototype link with both interaction options and no production write.
flowchart LR PROTOTYPE[Existing prototype] --> REVIEW[Matt review] REVIEW --> CHOICE[Interaction choice] CHOICE --> RELEASE[Approval gated release]
flowchart TD
S1[Locate and verify prototype] --> S2[Present both behaviors]
S2 --> S3[Record decision and approval]
S3 --> D{Live release approved}
D -- yes --> S4[Back up stage and verify]
D -- no --> READY[Keep prototype only]
S4 --> READY
Find the earlier rebuild, verify what exists on staging, and prepare an evidence-bound adoption proposal.
A potentially valuable rebuild sits outside the current evidence set, so its scope and commercial value cannot be judged.
Work may remain unused, or a proposal may be made from memory for something that cannot be reproduced or supported.
Matt gets provenance, responsive evidence, reusable scope, migration risks, and a professional proposal without deploying anything.
The historical session, files, repository, staging URL, ownership, and reproducibility are proven, followed by a visual comparison and proposal or an honest bounded not-found receipt. 80/20. Locate the staging URL and produce a screenshot-based inventory of what exists.
flowchart LR HISTORY[Claude history] --> ARTIFACT[Proven rebuild] ARTIFACT --> QA[Staging review] QA --> PROPOSAL[Adoption proposal]
flowchart TD
S1[Search history and artifacts] --> S2[Verify ownership and staging]
S2 --> S3[Inspect three responsive widths]
S3 --> S4[Compare current and rebuilt states]
S4 --> D{Evidence supports proposal}
D -- yes --> READY[Produce client proposal]
D -- no --> STOPPED[Return not found receipt]
Park a future comparison and commercial concept for a reusable custom CRM and automation product.
The agency may eventually need more control, but building now risks substantial unpaid work before demand and results are proven.
Without an idea record, the product opportunity will be repeatedly rediscovered and distract current priorities.
The commercial idea is preserved with dependencies and a clear trigger for revisiting it later.
Not expanded: this is a deliberately parked commercial idea, not current delivery.
Create one complete Hungarian and one complete US or international Bali-travel competitor teardown with visible ad, destination page, sitemap cues, SEO evidence where available, lead-capture surfaces, and clearly labelled inferred funnel steps.
Creative and persona choices need grounded market examples without a shallow or risky competitor dump.
The strategy may repeat weak patterns or rely on unsupported assumptions about credible competitors.
Two high-quality examples prove a repeatable, paced research method for the client presentation.
Two evidence bundles include ad creative and copy where available, destination URL, page and lead-capture screenshots, sitemap or page notes, source dates, and an inference boundary. No form is submitted and no personal inbox is used. 80/20. One Hungarian and one US example with a short comparison.
flowchart LR H[Hungarian example] --> EVIDENCE[Visible evidence] U[US example] --> EVIDENCE EVIDENCE --> FUNNEL[Inferred funnel] FUNNEL --> STRATEGY[Strategy input]
flowchart TD S1[Select two bounded sources] --> S2[Capture ads and landing pages] S2 --> S3[Record sitemap and search evidence] S3 --> S4[Label inference and unknowns] S4 --> READY[Publish comparison]
Create a static dashboard prototype for campaign, SEO, engagement, enquiry, quote, deposit, and paid-booking views without implying that mock values are real.
The client needs to see what improved measurement will look like before the full data path exists.
The proposed measurement system remains abstract and the next wiring step is harder to discuss.
Matt can show a professional visual target, explain knowns and unknowns, and set the next-meeting agenda.
A responsive static HTML contains the future funnel, campaign and SEO views, mock-data labels on illustrative values, evidence dates, unknowns, notes, and a wiring agenda. It passes 1280, 768, and 390 pixel QA. 80/20. One clean before-and-after dashboard view.
flowchart LR FACTS[Dated evidence] --> MOCK[Mock dashboard] MOCK --> REVIEW[Client review] REVIEW --> WIRING[Future real data wiring]
flowchart TD S1[Define stages and meanings] --> S2[Build static views] S2 --> S3[Label mock values] S3 --> S4[Add notes and agenda] S4 --> READY[Responsive QA pass]
Create a small creative pack for the one fully mapped persona using grounded psychology, competitor evidence, and the agreed funnel.
The strategy needs a concrete example of how persona psychology becomes an ad hook, copy, creative, page, and lead-capture offer.
The presentation stays abstract and client approval cannot focus on a real example.
Matt gets a strong example to review and a repeatable template for later creative batches.
One persona pack includes a grounded psychology map, hooks, ad copy, a small set of creative concepts, CTAs, up to three safe lead-magnet directions, and a stable persona identifier. Hypotheses are labelled. 80/20. One complete persona-to-creative example.
flowchart LR PERSONA[Persona evidence] --> PSYCH[Psychology map] PSYCH --> COPY[Hooks and copy] COPY --> CREATIVE[Creative concepts] CREATIVE --> FUNNEL[Funnel example]
flowchart TD S1[Read grounded sources] --> S2[Assign stable persona ID] S2 --> S3[Draft hooks and concepts] S3 --> S4[Connect CTAs and lead magnets] S4 --> READY[Package for presentation]
Define the exact meaning, evidence, owner, timestamp, and duplicate rule for enquiry, quote, deposit, paid booking, travellers, and revenue.
Advertising activity is not joined to clearly owned commercial outcomes.
Campaign decisions remain based on proxies and late or ambiguous updates.
One accepted outcome contract controls every later measurement and experiment.
A client-safe ownership map defines every stage and its accepted evidence. 80/20. One accepted quote definition and one accepted paid-booking definition with accountable owners.
flowchart LR A[Platform action] --> E[Recorded enquiry] --> Q[Quote sent] --> D[Deposit] --> B[Paid booking] --> R[Revenue]
flowchart TD S[Stage] --> DEF[Definition] --> EV[Evidence] --> OWN[Owner] --> SLA[Update timing] --> DUP[Duplicate rule]
Represent the same test journeys in a Google Sheet oversight ledger and QlickCRM, then choose one future canonical authority using explicit winner rules.
Matt needs a transparent overview without creating two competing sources of truth.
Records can drift while architecture is chosen from preference instead of evidence.
One system wins operational authority and the other becomes a one-way view.
A visual comparison tests completeness, latency, duplicates, owner adherence, exportability, and effort without changing real records. 80/20. One identical test journey in both views and a five-criterion winner rule.
flowchart LR J[Test journey] --> S[Google Sheet view] J --> C[QlickCRM view] S --> X[Compare evidence] C --> X
flowchart TD
X[Shadow period] --> C{Winner rule passed}
C -- Sheet --> S[Sheet authority]
C -- QlickCRM --> Q[QlickCRM authority]
S --> V[Other system is one-way view]
Q --> V
Select one departure, package, or service using capacity, fulfilment, price, margin, and cancellation evidence, then validate its offer, proof, CTA, and response owner.
The creative concepts exist, but the underlying sellable unit, economics, promises, and proof are not yet accepted.
Advertising could overpromise or create demand that Analit cannot serve.
The first sellable unit and offer become credible, supportable, and safe to present.
The unit, capacity, fulfilment, economics, every promise, proof item, CTA, and response owner are marked verified, approval needed, or unavailable. 80/20. Select one sellable unit, one promise, one CTA, and one response owner.
flowchart LR
U[Sellable unit] --> C[Capacity and economics] --> P[Persona promise] --> R[Rights check] --> A[Admissible offer]
flowchart TD
I[Claim or asset] --> E{Evidence exists}
E -- yes --> V[Verified]
E -- unclear --> H[Human approval]
E -- no --> X[Unavailable]
Define one hypothesis, one changed variable, the eligible denominator, a deduplicated business numerator, lag, guardrails, information target, and final decision rules before launch.
A fixed duration alone does not create a valid experiment.
Small samples and proxy metrics can manufacture a false winner.
The team knows whether the test is eligible before spending.
The pre-registration returns ELIGIBLE, INELIGIBLE, or NEEDS MORE BASELINE. 80/20. One variable, one denominator, and one deduplicated quote-qualified or paid-booking outcome.
flowchart LR H[Hypothesis] --> V[One variable] --> D[Denominator] --> N[Business numerator] --> R[Decision rule]
flowchart TD
P[Preflight] --> E{Enough information}
E -- yes --> GO[Eligible]
E -- no --> BASE[Needs more baseline]
E -- unsafe --> STOP[Ineligible]
Use non-customer test data and a stable journey identifier to prove source capture, enquiry, quote, deposit, paid booking, retries, duplicates, and missing states.
The proposed measurement chain is not credible until one complete journey survives tests.
Production work risks silent breaks at the most valuable stages.
Analit gets proof before any live rollout or campaign decision.
One isolated test journey reaches every agreed stage with visible negative controls. 80/20. One deterministic test record reaches quote and paid booking with a stable journey ID.
flowchart LR A[Test source] --> E[Enquiry] --> Q[Quote] --> D[Deposit] --> B[Paid booking]
flowchart TD T[Test record] --> P[Positive path] T --> R[Retry] T --> U[Duplicate] T --> M[Missing state] P --> V[Evidence receipt] R --> V U --> V M --> V
Keep dated performance in Google Slides and current strategy choices in a focused HTML decision room, with completed evidence and future ideas moved to labelled appendices.
The current review mixes facts, completed milestones, future ideas, and current gates.
The client can confuse historical evidence with live proposals.
Each review surface has one clear job and one next decision.
Historical and strategy roles are explicit, completed materials are linked, and comments plus voice context work at three viewports. 80/20. One decision room for outcomes, system authority, offer, and experiment eligibility.
flowchart LR S[Google Slides] --> F[Dated facts] H[Strategy HTML] --> D[Current decisions] A[Appendices] --> E[Completed evidence and future ideas]
flowchart TD R[Morning review] --> S[Slides] S --> H[Strategy HTML] H --> C[Comments and voice] C --> O[Next orchestrator plan]