In marketing-analyst/assets/, create three Bruin SQL assets for the staging layer. All should materialize as views (cheap, always fresh) with schema staging.

staging/google_ads.sql → staging.google_ads

• Source: raw.google_ads_campaigns
• Cast date to DATE, trim/dedupe on (campaign_id, date) using QUALIFY ROW_NUMBER() OVER (PARTITION BY campaign_id, date ORDER BY date DESC) = 1
• Convert cost_micros → cost_usd = cost_micros / 1e6
• Derived metrics per row: ctr = clicks / NULLIF(impressions, 0), cvr = conversions / NULLIF(clicks, 0), cpc_usd = cost_usd / NULLIF(clicks, 0)
• Keep: campaign_id, campaign_name, date, impressions, clicks, conversions, conversion_value, cost_usd, ctr, cvr, cpc_usd

staging/klaviyo.sql → staging.klaviyo

• Source: raw.klaviyo_campaigns
• Cast send_time to TIMESTAMP and derive send_date = send_time::DATE
• Dedupe on campaign_id (keep the most recent snapshot)
• Derived metrics: open_rate = opens / NULLIF(recipients, 0), click_rate = clicks / NULLIF(recipients, 0), revenue_per_recipient = revenue / NULLIF(recipients, 0)
• Add a comment noting that opens are unreliable on iOS; prefer clicks or revenue for engagement signals
• Keep: campaign_id, send_time, send_date, recipients, opens, clicks, unsubscribes, revenue, open_rate, click_rate, revenue_per_recipient

staging/ga4.sql → staging.ga4

• Source: raw.ga4_traffic
• Cast date to DATE
• Dedupe on (date, sessionSource, sessionMedium)
• Classify channel from medium: cpc → 'Paid Search', organic → 'Organic', email → 'Email', (none) / direct → 'Direct', everything else → 'Other'
• Derived: cvr = conversions / NULLIF(sessions, 0)
• Keep: date, sessionSource, sessionMedium, channel, sessions, totalUsers, conversions, purchaseRevenue, cvr

For each asset, add a top-level description:, column descriptions, and at least one quality check (e.g. not_null on the grain columns, unique on the dedupe key).

Show me the SQL for each asset before writing the files. Don't run anything yet - I'll approve first.