Voice review

Review this plan by voice

The voice agent explains any task, answers questions grounded in this plan, and records a brain dump. Voice feedback is read-only and cannot approve or execute work. Allow the microphone when your browser asks.

The explicit Sonnet rollback keeps the existing capsule-only protocol. A lost connection retries the same Sonnet engine twice and reloads the complete persisted conversation. It never changes models during a conversation.

Sonnet 5 · no fallback
Using: browser default until voice starts
offline

Live transcript

not started

Voice answers are review evidence. Approval, sending, spending, and production changes stay in the originating chat. Voice trouble in this page? Open the standalone voice page.

0 of 9 decided

Review — Csupi click and call-tracking closeout

Plan: /Users/agency/Documents/Agty/csupi/action-plans/2026-08-10-csupi-client-call-tracking-next-steps.md

How to review this

T-001 — Build the client-ready Budapest click visualization

Undecided
Task

Create a polished Hungarian HTML presentation of the verified district ranking and the separate hour-of-day ranking.

Problem solved

The completed analysis is accurate but not yet packaged as a client-friendly visual story.

Cost of inaction

Raw tables can be misunderstood as proof of fraud or may fail to show the value of the analysis.

Benefit

Matt gets a clear presentation that separates geography, time, invalid clicks, and intent evidence.

Done

A responsive Hungarian page passes factual and client-safety review and remains staged for Matt. 80/20. One page shows the top districts, top hours, limits, and next decisions.

flowchart LR
  A[Verified analysis exists] --> B[Build client visual story] --> C[Matt can explain findings safely]
    
flowchart TD
  S1[Ground in report and validators] --> S2[Design separate district and hour views]
  S2 --> S3[Build responsive Hungarian HTML]
  S3 --> S4[Validate facts and wording]
  S4 --> D{Review evidence passes}
  D -- yes --> DONE[Review ready]
  D -- no --> S2
    
Decision

T-002 — Verify Google Ads exclusion capabilities and limits

Undecided
Task

Use current official sources and read-only account checks to verify which district and schedule exclusions are actually possible.

Problem solved

The click-report dimensions may not map directly to Google Ads targeting controls.

Cost of inaction

A live plan could depend on a control the platform does not support or whose location meaning differs.

Benefit

Matt gets a dated support matrix before approving any campaign change.

Done

Official documentation and read-only account evidence classify each requested control. 80/20. Answer District XI, 10:00 to 10:59, and combined-control feasibility.

flowchart LR
  A[Requested exclusions] --> B[Verify platform controls] --> C[Feasible control matrix]
    
flowchart TD
  S1[Identify campaign types] --> S2[Read current official docs]
  S2 --> S3[Inspect account read only]
  S3 --> S4[Classify requested controls]
  S4 --> D{Evidence is current and exact}
  D -- yes --> DONE[Feasibility documented]
  D -- no --> S2
    
Decision

T-003 — Design the exact reversible exclusion pilot

Undecided
Task

Convert the two requested exclusion axes into one exact approval packet with campaign scope, duration, metrics, and rollback.

Problem solved

The desired axes are known, but the operational safety fields are not.

Cost of inaction

The live task stays blocked or becomes too broad to interpret and reverse safely.

Benefit

Matt can approve one narrow, auditable plan rather than an ambiguous mutation.

Done

The packet contains exact campaigns, controls, sequence, duration, sample floor, success metric, and rollback. 80/20. Fully specify the smallest recommended pilot and queue the other axis second.

flowchart LR
  A[Intent without exact scope] --> B[Build approval packet] --> C[Safe reversible pilot]
    
flowchart TD
  S1[Validate feasible controls] --> S2[Identify exact campaigns]
  S2 --> S3[Compare sequential and simultaneous tests]
  S3 --> S4[Define metrics and rollback]
  S4 --> D{Every approval field exists}
  D -- yes --> DONE[Ready for Matt decision]
  D -- no --> S2
    
Open questions
  • Which campaigns, duration, sample floor, success metric, and rollback trigger should Matt approve?
Decision

T-004 — Apply the approved exclusion pilot

UndecidedNeeds input
Task

Apply only an exact, separately approved exclusion pilot and preserve provider and rollback receipts.

Problem solved

The requested exclusion becomes a real platform experiment instead of a planning note.

Cost of inaction

Possible waste continues, while acting too early could suppress genuine demand.

Benefit

A narrow approved change can be verified and reversed.

Not expanded — campaign scope, duration, success metric, rollback trigger, and live authority are still unknown.

Open questions
  • Approve the exact packet from T-003 before any live Google Ads write.
Decision

T-005 — Finish the Google forwarding-number blocker investigation

Undecided
Task

Give one strong Codex owner agent the unfinished GFN investigation and let it coordinate bounded technical sub-agents.

Problem solved

Prior route results do not yet give Matt one current, executable verdict.

Cost of inaction

Call attribution stays unproven and the investigation risks circling the same blocked paths.

Benefit

One reconciled owner, verdict, and next action replaces fragmented attempts.

Done

Every prior route is reconciled and the result is either a verified repair path or an exact external gate. 80/20. Identify the blocker that kept the reviewed fix out of production.

flowchart LR
  A[Fragmented route audit] --> B[One Codex owner reconciles evidence] --> C[Executable repair or exact gate]
    
flowchart TD
  S1[Recover audits and run state] --> S2[Reconcile unresolved routes]
  S2 --> S3[Trace Google and Cloudflare path]
  S3 --> S4[Independent technical checks]
  S4 --> D{Current verdict has evidence}
  D -- yes --> DONE[Repair path or human gate]
  D -- no --> S2
    
Open questions
  • Which exact external owner or authority controls any remaining production deployment step?
Decision

T-006 — Evaluate third-party call-tracking alternatives

Undecided
Task

Compare current low-cost call-tracking providers for Hungarian numbers and Google Ads attribution without signing up or purchasing.

Problem solved

The project needs a credible fallback if Google's forwarding path cannot be repaired quickly.

Cost of inaction

More time may be lost on the same route or on a provider that lacks Hungarian support.

Benefit

Matt gets an evidence-backed shortlist, recommendation, and spend gate.

Done

A dated comparison verifies Hungarian number support, integration, privacy, setup, and total cost. 80/20. Find one qualifying provider and one fallback.

flowchart LR
  A[Google path uncertain] --> B[Compare qualified providers] --> C[Approval ready fallback]
    
flowchart TD
  S1[Define must have criteria] --> S2[Research official provider sources]
  S2 --> S3[Verify numbers integration privacy and cost]
  S3 --> S4[Rank options and gaps]
  S4 --> D{At least one provider qualifies}
  D -- yes --> DONE[Recommend provider and fallback]
  D -- no --> S2
    
Open questions
  • Any purchase above the current human-gate threshold needs Matt's separate provider and spend approval.
Decision

T-007 — Implement and terminally prove the chosen call-tracking path

UndecidedNeeds input
Task

Implement the chosen tracking route and prove one real paid-call conversion end to end.

Problem solved

Configuration alone does not prove that paid calls are attributed correctly.

Cost of inaction

Optimization and outcome reporting remain untrustworthy.

Benefit

One terminal proof chain creates a reliable measurement foundation.

Not expanded — provider choice, spend, production authority, qualifying-call rule, and consent decisions are still unknown.

Open questions
  • Choose the route and approve spend, production scope, qualification, and consent before implementation.
Decision

T-008 — Finalize the client presentation after the technical work

Undecided
Task

Combine the verified exclusion and call-tracking outcomes into one final Hungarian presentation and talk track.

Problem solved

The client needs one coherent status story after the parallel workstreams.

Cost of inaction

Matt may present stale status or overstate what tracking and exclusions have proven.

Benefit

One polished presentation explains completed work, gates, risks, and next actions.

Done

The final HTML and talk track reconcile every claim to evidence, pass QA, and remain unsent pending approval. 80/20. Refresh the presentation with final status and a five-minute talk track.

flowchart LR
  A[Parallel technical outcomes] --> B[Synthesize truthful client story] --> C[Review ready final presentation]
    
flowchart TD
  S1[Collect receipts and gates] --> S2[Update Hungarian presentation]
  S2 --> S3[Separate complete configured blocked and unproven]
  S3 --> S4[Validate and run responsive QA]
  S4 --> D{Every claim has dated evidence}
  D -- yes --> DONE[Matt review ready]
  D -- no --> S2
    
Open questions
  • After Matt reviews the final version, which client delivery channel is authorized?
Decision

T-009 — Set the qualified phone-enquiry definition and value

UndecidedNeeds input
Task

Agree what counts as a qualified phone enquiry and what business value it carries.

Problem solved

ROAS has no reliable meaning without a qualification rule and value.

Cost of inaction

The account may optimize toward raw calls rather than commercially useful enquiries.

Benefit

Matt and the client get a shared basis for value reporting.

Not expanded — the qualification rule and commercial value require a business decision.

Open questions
  • What minimum duration, disqualifiers, evidence source, and value should define a qualified call?
Decision