# DOP Cold Audit — Proof Pack (Internal)
**Author**: DOP (Leonard) · **Date**: 2026-08-09 · **Status**: Internal — NOT client-facing
**Purpose**: Consolidate the 5 real Cold Audits as the *Verified* evidence base, contrast against Bizl's *Indicative* (Tier-3) claims, and nominate the public case-study wedge.
**Repo**: all source audits committed under `research/cold-audits/` (Gitea, `tonyjbala`).

---

## 1. Evidence-class framing (per Tony's ruling)
You established the DOP value model on two evidence classes. The Cold Audits operationalize exactly that split:

- **Verified Evidence** — Tier-1/2, observed directly by DOP (live site schema, live search results, direct inspection). *This is what our 5 audits produce.*
- **Indicative Evidence** — Tier-3, self-reported by a vendor (Bizl's 4 case studies). Useful as a *wedge* in outreach, never as *proof* of DOP's own capability.

The entire DOP credibility position rests on this distinction: **we show our work; Bizl shows a screenshot.** The audits below are the proof layer that makes that claim defensible.

---

## 2. The 5 audits (Verified Evidence)

| # | Vertical | Target | Headline finding (observed) | Commit |
|---|---|---|---|---|
| 1 | HVAC | Gilmore Heating/Air (Cameron Park) | 3 conflicting phone numbers + Antelope/95843 schema geo vs Cameron Park target | `ca87ac49` |
| 2 | Plumbers | D. Martel Plumbing (El Dorado Hills) | **Zero machine-readable NAP** — site publishes only AboutPage/ContactPage schema | `2ced37bc` |
| 3 | Dentists | Cameron Park Family Dentistry | Schema `name` = generic "Dentist in Cameron Park, CA" (not the brand); no `Service` entities | `320dea1c` |
| 4 | Law | Gordon Law Firm (El Dorado Hills) | Ranks #1 at home but invisible for Cameron Park; phone masked `+191****6261` in schema | `71d25516` |
| 5 | Multi-Location | Above the Gate Dentistry (2 sites) | 403 bot-block prevented full capture; structural NAP-split risk identified, **PENDING re-audit** | `16f04eb8` |

All 5 are Tier-1/2: every number is either observed (schema dump, search result position) or explicitly flagged *unverified*. No vendor self-report is presented as fact.

---

## 3. Verified vs Indicative — the contrast that matters

Bizl's 4 case studies claim outcomes like "tripled," "4x," and "67–90 days." All are:
- **Tier-3** (self-reported by one vendor, single-firm),
- internally **disjunct** (one says 67 days, another 90; one says "tripled," another "4x"),
- **unfalsifiable** by an outside reader.

Our 5 audits contain **observed, falsifiable facts**:
- A real business publishing **three different phone numbers** (Gilmore).
- A real business with **no machine-readable NAP at all** (Martel) — literally invisible to any AI-answer engine parsing structured data.
- A real multi-location practice whose **site hard-blocks crawlers (403)** — a structural silent-loss vector.

This is the wedge. When we say "most local businesses are losing customers silently and can't see it," we can *show* three of them doing exactly that, by name, with the evidence. Bizl can only *assert* it.

---

## 4. Wedge nomination (which become public case studies)

- **Primary wedge — Gilmore HVAC.** The 3-phone conflict + wrong-city schema geo is the most visceral "silent loss" story: a customer in Cameron Park gets sent to Antelope, or to a disconnected line. Highest emotional + factual punch.
- **Secondary wedge — D. Martel Plumbing.** Zero machine-readable NAP is the cleanest "invisible to AI" proof — a single, undeniable technical fact that generalizes to every SMB. Easiest to explain in one sentence.
- **Hold — Above the Gate (multi-location).** This is the one vertical where Bizl had *no* case study, so it's our potential differentiator. But the audit is **PENDING re-audit** (403 block). Do not publish until Tier-1 schema + both GBP listings are verified. A half-audited multi-location case would undercut the "we show our work" position.

---

## 5. Honest verification gaps (carried forward)
- **Foursquare**: unverified across all 5 (outage during audit window) — the single most important AI-data source per the DOP model; schedule a re-check.
- **GBP**: not directly inspectable (Google bot-blocks the VPS IP) — scored via Yelp Tier-2 proxy only.
- **Above the Gate**: site returned 403 to curl + browser; partial audit by design.
- **AI-answer presence**: web_search proxy, not ChatGPT/Perplexity screenshots (no LLM-answer tool available).

None of these invalidate the Verified findings; they bound how far we can currently push each claim. State them in any client-facing version.

---

## 6. Recommended action
1. **Build the public case study** from Gilmore (primary) + Martel (secondary) — both are fully Verified and need no re-audit.
2. **Re-audit Above the Gate** with a reachable IP (residential proxy or direct GBP access) to complete the multi-location proof, then promote it as the differentiator wedge.
3. **Re-check Foursquare** for all 5 once the source is back — close the biggest remaining gap in the Verified base.
4. **Retire Tier-3 reliance** in GTM: keep Bizl numbers only as Indicative wedge context, never as DOP-proof.

---
*All 5 source audits + this pack: `research/cold-audits/` on Gitea, committed as `tonyjbala`, byte-verified.*