Harbor Repair: choose the next action

# Harbor Repair strategy and proof audit — harbor/1

**Decision:** The September funnel supports testing clearer proof and booking steps before seeking more leads. The numeric target fits stated capacity, but the funnel has no channel attribution and margin is unknown, so expected gains and affordability cannot be established.

## Scope and source ledger

This is a synthetic, private exercise using only `harbor-v1.json` and its approved method snapshot (`AUDIT-METHOD-2026.10.02`). The maintained method URL was not opened; no claim is made that it was read. No live customer tools, browsing, or external actions were used. The fixture is the sole business context.

| ID | Supplied source / date / scope | Supports | Status |
|---|---|---|---|
| S1 | Synthetic owner brief, 2026-10-06 | Business, offer, buyer, goal, baseline, target, capacity, authority | OBSERVED as supplied; not independently verified |
| S2 | Synthetic CRM September export, dated 2026-10-01; 2026-09-01–09-30; one pipeline, distinct contacts; no channel attribution | Leads and funnel counts | OBSERVED as supplied; account/read not independently verified |
| S3 | Synthetic invoiced revenue export, dated 2026-10-01; same September completed cohort | Revenue, completed jobs, gross margin field | OBSERVED as supplied; account/read not independently verified |
| S4 | Synthetic content inventory, 2026-10-06 | Three stories; permission and offer mapping fields | OBSERVED as supplied |
| S5 | Synthetic website observation, 2026-10-06 | CTA and story-link count; booking steps unknown | OBSERVED as supplied; independent site check not performed |
| S6 | Synthetic identity record, 2026-10-06 | Graph identity field and unknown panel/claim | OBSERVED as supplied; not an independent search observation |

## Business map

```mermaid
flowchart LR
  G[Goal: 18 completed jobs/month within 90 days] --> B[Local homeowners]
  B --> O[Booked home repair visits]
  O --> P[3 customer stories; permission unknown; offer mapping missing]
  P --> N[Website CTA: Contact us; booking steps unknown]
  N --> M[Completed jobs/month: 12 supplied September baseline]
  M --> T[Target 18; stated capacity 20]
```

The three stories are the strongest supplied proof inventory, but publication permission is unknown and none is mapped to the offer. The generic “Contact us” CTA and unknown booking steps leave an unmeasured gap between interest and a booked visit. These are evidence-based hypotheses to investigate, not proof that changing the page will increase jobs.

## Findings and measures

| Measure | Value | Basis |
|---|---:|---|
| September leads | 72 | S2 supplied count |
| Qualified / leads | 36 / 72 = 50% | S2 supplied counts; exact ratio |
| Booked / qualified | 18 / 36 = 50% | S2 supplied counts; exact ratio |
| Completed / booked | 12 / 18 = 66.7% | S2 supplied counts; rounded for display |
| Completed / leads | 12 / 72 = 16.7% | S2 supplied counts; rounded for display |
| Revenue per completed job | $18,000 / 12 = $1,500 | S3 supplied cohort; revenue average, not profit or margin |
| Target change from baseline | +6 jobs/month; +50% | S1 supplied baseline and target |
| Target use of stated capacity | 18 / 20 = 90% | S1 supplied target and capacity; leaves 2 jobs/month stated headroom |

The stated target is below capacity. This does not establish that capacity is operationally available or that demand exists. Revenue per job is not a margin measure. No ROI, channel performance, or causal conclusion can be calculated from these inputs.

## Panel and exam

- Graph identity: **OBSERVED in the supplied identity record** (S6); independent identity check not performed.
- Visible search Knowledge Panel: **UNKNOWN** (S6).
- Owner claim: **UNKNOWN** (S6).
- Exam: **unscored** (`null`). No exam, version, weights, coverage, or scoring template was supplied. No score is assigned.

## Prioritized actions for Jordan

These are proposed for approval; the fixture authorizes report files only, not implementing, publishing, scheduling, or contacting anyone. Deadlines are relative to this report date, 2026-10-06.

| Rank | Action / first step | Owner | Deadline | Dependency | Observable completion check |
|---|---|---|---|---|---|
| 1 | Review the three stories; confirm publication permission and map each permitted story to the booked home-repair offer and buyer concern. | Jordan | Within 7 days | Permission/source confirmation | Each story has a permission status and an offer/buyer mapping; no unapproved story is used. |
| 2 | Walk through the current contact-to-booking path and define what counts as a booking and completed job; record where each handoff is counted. | Jordan | Within 14 days | Access to the booking process and CRM | Written steps and event definitions; a test inquiry can be traced to a booking record. |
| 3 | Decide whether to test a specific booking CTA and offer-matched proof after actions 1–2; define a baseline and comparison window first. | Jordan | Within 21 days | Story permission and booking-path findings | A written test decision, named outcome metric, baseline, window, and owner; no claimed result before measurement. |
| 4 | Supply gross margin, channel attribution, and a budget/capacity view before deciding on paid acquisition or spend. | Jordan | Before any spend decision | Source exports and owner judgment | Margin and channel fields are documented or explicitly remain unavailable; spend decision recorded. |

Expected usefulness is strongest for validating permission and the booking path because those gaps are directly evidenced. Any conversion lift from a CTA or proof change remains a hypothesis. No agent action is authorized by this exercise.

## Open inputs and GCT questions

- What gross margin applies to the September jobs, and what labor/material capacity remains at 18 jobs?
- Which channel produced the leads, qualified contacts, bookings, and completions? Attribution is absent.
- What budget, if any, is available? No spend recommendation can be grounded without margin and budget.
- Do the three stories have written publication permission, and which buyer concern/result does each support?
- What are the booking steps and booking-step tracking definitions?
- Is a Knowledge Panel visibly present in a specified search context, and has Jordan claimed it? The supplied record leaves both unknown.
- GCT: What exact result (completed jobs and timeframe) matters most, what material (story/page/offer proof) may be used, and which audience segment of local homeowners should it address?

## Weekly Metrics → Analysis → Action

Proposed first weekly review: **2026-10-13**, owned by Jordan. This is a recommendation recorded in the report, not a scheduled event.

| Metric / source | Interpretation to make | Next owned action |
|---|---|---|
| Leads, qualified, booked, completed jobs; S2 CRM, with same dates and one pipeline | Check counts and stage ratios against the September baseline; do not infer channel cause. | Jordan records the next step with the largest evidenced gap and checks definitions. |
| Revenue and gross margin; S3/invoice and cost source | Separate revenue from contribution/profit; if margin remains missing, keep spend decisions open. | Jordan supplies margin or records it unavailable. |
| Story permissions and offer mapping; S4 | Establish which proof is usable and relevant to the actual offer. | Jordan completes the story review. |
| Booking path and CTA; S5 plus booking records | Determine whether an observable, traceable path exists. | Jordan documents steps and measurement before any test. |

## QA and execution receipt

| Check | Result | Evidence / limit |
|---|---|---|
| Identity | UNKNOWN | Synthetic record says graph identity OBSERVED; no independent identity verification. Panel and claim remain UNKNOWN. |
| Sources | PASS for fixture scope | All supplied IDs S1–S6 are represented; no live source was accessed. |
| Arithmetic | PASS | Ratios and target/capacity calculations recomputed from fixture values; display rounding only. |
| Exam/version | PASS | Unscored default honored; exam is null. |
| Unknowns | PASS | Missing margin, attribution, budget, permissions, booking tracking, panel and claim remain explicit. |
| Correction propagation | NOT APPLICABLE in v1 | No correction input was included in this revision. |
| Ownership | PASS | Proposed actions each name Jordan, deadline, dependency and completion check; none is represented as executed. |
| Visual rendering | UNKNOWN | Mermaid source included; no renderer/export was available or checked. PDF/other exports: NOT CHECKED. |

**Execution:** mode `synthetic_exercise`; runtime `Codex`; actual model `UNKNOWN`; surfaced instruction family `GPT-6`; requested model `gpt-6-luna`; actual routing and effort `UNKNOWN`. Method `AUDIT-METHOD-2026.10.02` supplied snapshot only; prompt `BUSINESS STRATEGY AUDIT STARTER 1.0.0`; input `harbor/1`. Tools used: local file read/write only. No browsing, connectors, or external actions. Elapsed time, allowance/cost, retries, and rework: UNKNOWN. No business result or delivery outside these private report files is claimed.