// How It Works — 05
From blank slate to self-running store.
The complete process: build the foundation right, engineer conversion, wire the ads infrastructure, then let AI agents run the routine — with human gates on everything that matters. Seven phases, no mystery.
// A — The map
Five layers, one operating system.
A store isn't a website — it's an operation with a UI. Every build follows the same shape, sized to the business: foundation first, conversion on top, traffic wired in, automation layered over the routine, operations keeping it alive.
FIG — THE ECOMMERCE OPERATING SYSTEM. EACH LAYER GETS THE FULL DISCIPLINE, WHATEVER THE SIZE.
// B — The process
Seven phases, zero mystery.
Timelines depend on scope, but the sequence never changes — each phase produces something you can inspect before the next one starts.
- PHASE 01Discover & auditEverything starts with the business, not the tools: what you sell, who buys, current stack, data flows, and — critically — which AI plugins are already touching your customer data. The output is a risk + opportunity map you can act on even if we never work together.Deliverable — Audit report — stack, risks, upside, priorities
- PHASE 02FoundationThe platform goes in properly: Shopify/Plus or a custom stack, the product-data spine designed before any theme work, checkout path, payments, shipping, tax, and the analytics skeleton. The boring layer done right is what every later phase stands on.Deliverable — An operable store with a clean data spine
- PHASE 03Conversion systemTheme architecture merchants can actually run, speed budgets per template, landing-page systems for campaigns, trust signals, and — the part most stores skip — designed failure states: out-of-stock, payment errors, empty carts. Conversion is infrastructure, not decoration.Deliverable — Conversion infrastructure with performance budgets
- PHASE 04Ads & traffic infrastructureThe traffic engine: Meta pixel + Conversions API, GA4, server-side tagging, product feeds for Merchant Center, campaign architecture (prospecting / retargeting / structured testing), UTM discipline, and a reporting dashboard that reads in one screen. Campaigns can be run by you, your team, or me — the infrastructure makes results measurable either way.⚡ AUTOMATION ENTERS HEREDeliverable — Tracking that survives iOS updates + a feed that never rots
- PHASE 05Automation overlayNow the agents earn their place: Hermes Agent self-hosted in your infrastructure for briefings, catalog QA, and scheduled ops; OpenClaw for chat-native approvals; Grok for real-time signals. Every action logged, cost-capped, and gated — the layer runs the routine so humans run the judgment.⚡ AUTOMATION ENTERS HEREDeliverable — Self-running ops layer with audit logs and human gates
- PHASE 06Launch & verifyEnd-to-end test orders through real payment flows, tracking verification (every event checked server-side), performance pass against the budgets, backup restore drill, and the runbook written down. Launch is boring on purpose.Deliverable — Launch checklist — all green, nothing assumed
- PHASE 07Operate & improveThe system stays alive: uptime watch, daily briefings, weekly optimization loop (speed, conversion, campaign learnings), and new automation recipes as the store grows. The stack improves every month; the discipline never changes.⚡ AUTOMATION ENTERS HEREDeliverable — A living system — monitored, backed up, improving
// C — The ads layer
Traffic you can actually read.
Ads are where most stores burn money blind. This layer is built as infrastructure — tracking, feeds, and campaign architecture that stay measurable no matter who holds the wheel.
01Tracking that doesn't lie
Meta Pixel + Conversions API with event deduplication, GA4, and server-side tagging. iOS privacy changes broke naive pixel setups years ago — the server-side layer is where attribution truth lives now.
02Product feeds that feed themselves
Merchant Center / catalog feeds generated from the product-data spine, checked nightly by the automation layer — disapprovals caught before they pause your campaigns.
03Campaign architecture
Structured account: prospecting, retargeting, and testing campaigns with clean naming and budgets — so every ad dollar has a job and a measurable return path.
04Landing & conversion systems
Campaign-specific landing systems built on the conversion infrastructure — fast, message-matched, instrumented. Traffic without a landing system is money with a leak.
05Reporting in one screen
A dashboard pulling spend, revenue, and automation activity into one view. You see what's working without opening five platforms.
06The automation hooks
The morning briefing includes ad performance and anomalies; the agent flags budget drift and creative fatigue — humans still move the money.
// D — Live ops
What a day looks like when it's running.
The end state of the whole process — a store whose routine runs itself and whose exceptions arrive at you, not the other way around.
06:30
Morning briefing lands
Overnight sales, stock risks, ad anomalies, webhook failures, catalog issues found by the nightly QA sweep — one message, zero dashboards opened.
09:00
Human gates cleared
The agent prepared refund drafts, catalog fixes, and campaign tweaks. You approve, adjust, or ignore — each decision logged.
15:00
Ops sweep
Order anomalies, fulfillment delays, and inventory drift detected and routed before customers notice.
03:00
Nightly sweep + backups
Catalog QA runs, backups dump and verify, reports regenerate. The morning briefing is already written before you wake up.
// E — Deep dives
Each layer has its own page.
// Engage
Phase one costs you nothing but a conversation.
The audit is where every engagement starts — tell me about the store and I'll come back with the risk + opportunity map.