Can I vibe code Northbeam?

Build a full-stack E-Commerce Multi-Touch Attribution (MTA) platform using Next.js (App Router), ClickHouse, PostgreSQL, and Node.js workers.

CORE DATA ARCHITECTURE & SCHEMA:
1. PostgreSQL (Metadata & Sync Configuration):
   - Stores user accounts, store domains, ad network OAuth tokens (Meta, Google Ads, TikTok), campaign mapping, and model rules.
   - Schema: `stores`, `ad_accounts`, `campaigns`, `ad_creatives`, `conversions`, `attribution_models`.

2. ClickHouse (High-Speed Analytics Engine):
   - Table `tracking_events`: `event_id` (UUID), `store_id` (UUID), `visitor_id` (UUID), `session_id` (UUID), `timestamp` (DateTime64), `event_type` (Enum: page_view, click, add_to_cart, purchase), `utm_source`, `utm_medium`, `utm_campaign`, `utm_content`, `utm_term`, `click_id` (fbclid, gclid, ttclid), `referrer`, `ip_hash`, `user_agent`, `order_id` (Nullable), `revenue` (Nullable Decimal).
   - Table `ad_spend_daily`: `store_id`, `date`, `platform`, `campaign_id`, `ad_id`, `impressions`, `clicks`, `spend`.
   - Table `attributions`: `store_id`, `order_id`, `model_name` (Enum: first_touch, last_touch, linear, time_decay, position_based, shapley), `ad_id`, `campaign_id`, `attributed_revenue`, `attributed_weight`, `calculated_at`.

FIRST-PARTY TRACKING PIXEL:
- Create a lightweight JavaScript web snippet (`tracker.js`) that captures web visits, URL parameters (`utm_*`, `gclid`, `fbclid`, `ttclid`), and assigns persistent first-party cookies (`_mta_id`, 1-year expiry with `SameSite=Lax`).
- Create an API route POST `/api/v1/track` in Node.js that ingests payload, resolves geo/device headers, writes to ClickHouse async via a stream batcher, and returns HTTP 204.
- Implement server-side CAPI endpoint POST `/api/v1/webhooks/shopify/orders-create` to receive Shopify order webhooks, link the order's customer IP/email/phone hash back to the `visitor_id` click path in ClickHouse, and insert the conversion record.

ATTRIBUTION ENGINE WORKER:
- Create a background Node.js/TypeScript worker process that computes attribution for every new purchase:
  1. Fetch all touchpoints for `visitor_id` leading up to `purchase_timestamp`.
  2. Implement Linear Model: equal weight assigned across all touchpoints.
  3. Implement Time-Decay Model: exponential decay with a 7-day half-life favoring touchpoints closer to conversion.
  4. Implement Position-Based (U-Shaped) Model: 40% First Touch, 40% Last Touch, 20% distributed evenly across middle touches.
  5. Store attribution weights in the ClickHouse `attributions` table.

ANALYTICS DASHBOARD UI:
- Build Next.js dashboard using Tailwind CSS and Recharts:
  1. Executive Overview: Compare Blended ROAS, CAC, Total Revenue, and Net Profit across custom date ranges.
  2. MTA Model Comparison Matrix: Side-by-side data table comparing Facebook/Google reported ROAS vs. First Touch, Last Touch, Linear, and Time-Decay ROAS.
  3. Creative Performance Tab: Aggregate ROAS, CTR, Spend, and Attributed Revenue grouped by `utm_content` / ad creative ID.
  4. Data Health Indicator: Show status of web pixel incoming stream, CAPI payload latency, and sync state for Meta/Google APIs.

OUT OF SCOPE:
- Bayesian Media Mix Modeling (MMM) statistical pipelines.
- Live automated bid/budget adjustments back to Meta/Google ad managers.
- Multi-currency conversion processing.

$ each button prefixes agent-specific run instructions · build your own product, never copy proprietary code, trademarks or designs