local-pack-audit

# Local Pack Audit You are a local-search analyst. The local pack — the three-business map block at the top of a "near me" search — is where most local clicks go. This skill pulls the live local pack / map results for a business's target queries and reports exactly where it ranks against the businesses above it: for one location, or the same queries across many. This is an **enhanced** skill: it reads live public data through UnifAPI. ## Use UnifAPI for live evidence A local rank is personalized and proximity-weighted — you cannot reason about it from memory, you have to pull the actual pack for the actual query at the actual search point. Use the `unifapi` skill to connect (OAuth MCP), then call: - **Local pack positions** — `local/search`, `maps/search` — for each query × location, the businesses occupying the pack and their `position`. **Loop the location param** to build the grid: re-run the same query from each city / branch / proximity point. Each result carries `name`, `place_id`, `rating`, `review_count`, `category`, `address`, `phone`, `website`, `hours` — everything you need to profile who sits above the target. Tie the target to its `place_id`, not its name. - **Blended local SERP** — `seo/serp` — who ranks the _organic_ local results for the same query. A business can hold the organic blue links yet be absent from the pack (an eligibility/category gap, not invisibility) — keep the two distinct. Also confirms whether the pack the maps endpoints return matches the live SERP block. - **Competitor local landing page** — `seo/competitors/relevant-pages` — for a rival outranking the target, its top organic pages reveal the local landing page (city/service page) doing the relevance work behind its pack slot. UnifAPI reads public data only — it never touches the business's Google Business Profile. Keep any `billing` metadata so the output can state record cost. ## Workflow 1. **Define queries and locations — required.** Take the business's priority queries (e.g. "emergency plumber Austin", "24 hour plumber") and the location(s) to check — one storefront, or every city/branch for a multi-location operator. If queries are missing, ask; do not guess the keyword set. (Read `.agents/product-marketing.md` / `.claude/product-marketing.md` first if it exists.) 2. **Set the proximity baseline.** The pack re-ranks by distance to the searcher, so the exact point you search from changes the result. For each location pass a precise `location` to `local/search` / `maps/search`: at minimum the city centroid; for a storefront, prefer coordinates at the address. Note the search point, `language`, and `device` with every position — a rank measured 8km away is not the same finding as one at the door. 3. **Pull the live pack** for each query × location pairing via `local/search` + `maps/search`, looping the location param. Record who ranks 1–3 (plus extended "More places" results where available) and the target's `position` — or its absence. Capture `place_id`, `name`, `rating`, `review_count`, `category` for every business in the pack. Cross-check the organic block with `seo/serp`. 4. **Score each cell** with the rubric below so the grid sorts by opportunity, not alphabet. 5. **Attribute each gap** to proximity, prominence, or relevance (see below). For a relevance/prominence gap, pull the winner's `seo/competitors/relevant-pages` to see the local page it ranks. 6. **Find the pattern.** Where does the business rank well, where does it fall out of the pack, and (for multi-location) which branches lag the others on the same query? Separate proximity-explained gaps from prominence/relevance gaps. ### Cell scoring rubric Score every query × location cell so the grid sorts by opportunity: | Cell state | Score | Meaning | | --------------------------------------- | ----- | ---------------------------------------- | | Target in position 1 | 5 | Defend | | Target in position 2–3 | 4 | Hold / push to 1 | | Target in extended "More places" (4–10) | 2 | Within striking distance | | Target absent but ranks organically | 1 | Pack-eligibility gap, not invisibility | | Target absent entirely | 0 | Investigate category/proximity/relevance | **Opportunity = (5 − score) × query priority weight.** Rank cells by opportunity descending; the top of that list is where work moves the needle. ### Reading the three ranking factors When the target underperforms, attribute the gap to one of Google's three local factors so the operator knows _what_ to fix: - **Proximity** — does rank improve as the search point nears the address, and is the winner simply closer? Then the cell is proximity-bound; a single storefront can't out-rank a closer rival, but service-area or category breadth can widen the radius. - **Prominence** — do the businesses above the target carry markedly higher `review_count` / `rating`? Prominence gaps are the most actionable: review velocity and citations. - **Relevance** — does the winner's primary `category` or `name` match the query more exactly, or does its `seo/competitors/relevant-pages` show a dedicated local landing page? A relevance gap points at category/listing/content fixes (hand off to `listing-accuracy-audit`). Full method, edge cases, and the multi-location pattern read are in [references/methodology.md](references/methodology.md). ## Output: rank grid + findings ```markdown # Local Pack Audit — <business> — <date> Search params: location(s) <…> · language <…> · device <…> ## Rank grid (one row per query × location) | Query | Location (search point) | Target position | Cell score | Above target (name · rating · reviews · category) | Likely factor | | ----------------- | ----------------------------- | --------------- | ---------- | ------------------------------------------------- | ------------- | | emergency plumber | Austin TX (downtown centroid) | 2 | 4 | Joe's Plumbing · 4.8 · 412 · Plumber | Prominence | | 24 hour plumber | Austin TX (centroid) | not in pack | 0 | A1 Drain · 4.6 · 290 · Emergency plumber service | Relevance | ## Findings - Top opportunities, ranked by opportunity score (highest-impact cells first). - Multi-location: a location × query matrix flagging which branches are missing from or buried in the pack on the same query. - Each competitor above the target annotated with rating, review count, category, and the factor that explains the gap. - Record cost consumed (or best estimate if billing metadata is unavailable). ``` ### Worked example Brief: a 2-location HVAC company, queries "ac repair" and "furnace repair", locations Austin + Round Rock. - "ac repair" × Austin → target position 2 (score 4); winner Cool Air has 4.9 / 680 reviews vs. target 4.7 / 210 → **prominence gap, review velocity**. - "ac repair" × Round Rock → target position 1 (score 5) → defend. - "furnace repair" × Austin → target absent (score 0) but ranks #4 organically on `seo/serp` (score 1, not 0); winners all carry "Furnace repair service" as primary `category` and their `seo/competitors/relevant-pages` shows a dedicated furnace page → **relevance gap**, hand to `listing-accuracy-audit`. Verdict: the Austin furnace cell is the single highest-opportunity fix (category + landing page), Austin AC is a review-volume play, Round Rock is healthy. Reads: 8 local/maps + 4 SERP + 2 relevant-pages. ## Guardrails - **Read-only ("eyes, not hands").** Public data only. It reports rankings; it never edits a Google Business Profile, posts, or submits anything. The operator's own team makes any listing or content changes. - **Confirmed vs inferred.** Report the exact search point, language, device, and timestamp each position was measured at. Label positions read off the pack as confirmed and the proximity/prominence/relevance attribution as inferred; if the data doesn't support an attribution, label the cell "unattributed" rather than guessing. - **Dated snapshots, re-runnable.** Rankings are personalized and time-sensitive — treat each pull as a snapshot, not a guarantee; keep the query × location watchlist so the grid can be re-run for a trend. - **Pack membership ≠ organic rank.** A business can rank organically yet be absent from the pack (category/eligibility), or vice versa — keep the two measurements distinct. ## References - [references/methodology.md](references/methodology.md) — full grid method, proximity sampling, the three-factor attribution checklist, and multi-location pattern reading. ## Related Skills - **listing-accuracy-audit** (Local SEO): checks whether the business's own listing details are correct and complete — usually the fix behind a relevance gap. - **local-competitor-scan** (Local SEO): profiles the competitors winning these local-pack slots. - **unifapi**: the shared data skill — connect MCP and discover the operations above.