wiki/clients/current/quarra/2026-03-04-inventory-performance-marketing.md · 1116 words · 2026-04-05

Weekly Call — Inventory & Performance Marketing (2026-03-04)

Overview

Weekly review covering two primary topics: (1) Amazon FBA inventory stockouts suppressing ROAS, and (2) a full Google Ads account overhaul for Quarrowstone, including a live demonstration of [1] diagnosing and fixing the account in real time via the Google Ads API.

Attendees: Gilbert Barrongo, Mark Hope, Sebastian Gant, Karly Oykhman
Source: Fathom Recording


Key Decisions


Action Items


Inventory & Performance Impact

Current Stockouts

Product FBA Stock In Transit 30-Day Sales Velocity
Old World White Popcorn 0 units 2,400 units 425 units/mo
Kidney Beans (5 lb) 0 units 288 units 72 units/mo
Various 25 lb SKUs 0 units In transit Lower velocity

Shipments scheduled for March 9 are delayed due to unresolved pickup issues Carly is managing.

Performance Metrics (Last 30 Days)

Inventory stockouts are the primary suppressor of ROAS. Mark's position: resolve the stockout problem permanently and stop discussing it week-over-week.

AI Inventory Tool (In Development)

Gilbert is building an AI-driven shipment quantity calculator using:
- 30-day sales velocity per SKU
- Lead time (typically 2–3 weeks)
- Current FBA stock levels

Demo planned for next week's call.


Root Cause Analysis

The Quarrowstone account had zero conversions across all campaigns. Claude Code's API-based diagnosis identified the following:

Issue Detail
Wrong bidding strategy "Maximize Conversions" with no conversion history — algorithm has nothing to optimize
Budget too low $7/day at $4 CPC = ~2 clicks/day; insufficient signal
16 duplicate conversion goals Auto-created goals confused the bidding algorithm
Keyword bloat 200 keywords; only 10 generating any traffic
Irrelevant broad match traffic Queries like "bronze statues," "marble floor cleaner," "wax statue maker"
No negative keywords Two largest campaigns had zero negatives
Low Quality Scores Scores of 2–4; root cause is poor landing page relevance
Brand terms in non-brand campaigns Brand CPC ~$0.30 vs. non-brand ~$5.00; wasted budget
Retargeting display: zero impressions Likely missing audience list
Optimization score below 80 Google penalizes accounts below this threshold

Fixes Applied (This Session)

  1. Bidding strategy: Changed all search campaigns to Maximize Clicks, $4 max CPC
  2. Budgets: Search campaigns → $16/day; retargeting → $10/day
  3. Conversion goals: Reduced from 16 → 2 biddable actions (contact form, Google lead form)
  4. Negative keywords: Added 839 negatives via Claude Code, including:
    - Brand terms (to be isolated in dedicated brand campaign)
    - Irrelevant materials: bronze, fiberglass, granite, marble
    - Competitor terms (~95 exact match)
    - Fine art / foundry queries (~71 exact match)
    - Restoration-specific terms (~180)
    - Residential / transactional modifiers: "for sale," "buy," "shop"
  5. Keyword purge: Removed irrelevant keywords from ad groups (127 → 104 remaining)
  6. Optimization score: Applied Responsive Search Ads recommendation; score raised 78 → 83

Remaining Items (Not Yet Done)


Claude Code Live Demo

Mark demonstrated [1] connecting to the Google Ads API via ~100 pre-built MCP server tools. Key observations:

Next step for Gilbert: Set up a Google Ads-only Claude Code workspace (isolated from Mark's multi-tool environment) and configure cron jobs for automated weekly audits.

"It's like a hacker. An AI hacker." — Gilbert Barrongo


ClientBrain Demo

Mark briefly demonstrated [2] (clientbrain.asymmetric.pro), an internal tool aggregating all client communications (email, Slack, Fathom transcripts, ClickUp tasks, Google Drive) into a searchable, AI-queryable interface. Features shown:

Gilbert's login is broken; Mark to fix auth issue before next session.


Sources

  1. Claude Code
  2. Clientbrain
  3. Index
  4. Index
  5. Google Ads Bidding Strategy