Project plan

Sync the Lovable scaffold locally, set up the Supabase environment, and commit the planning docs alongside the code.

• Clone the repo locally and run bun install.
• Verify the Lovable scaffold loads with bun run dev.
• Confirm the Supabase project is provisioned and keys are in .env.
• .env.example and .dev.vars.example committed as templates.
• data/ added to .gitignore for the upcoming HubSpot CSV.
• Project plan, README, and HubSpot export checklist committed to the repo.

Validate the Supabase schema, build the HubSpot importer with a dry-run default, write to the database only after the dry-run report looks clean.

• Added last_engaged_at timestamptz column to subscribers in the Phase 0 closeout migration.
• Build scripts/import-hubspot.ts with a --dry-run default and explicit --commit to write.
• Stream-parse the CSV via papaparse, validate each row with Zod.
• Export contacts from HubSpot and save to data/hubspot-export-{date}.csv.
• Generate data/dry-run-report.json with totals, invalid emails, duplicates, opt-outs, bounces, engagement breakdown.
• Review the report, then proceed to Phase 2.

Run the CSV through email verification, segment by engagement, import only the verified warm segment to Supabase first.

• Choose a verification provider (ZeroBounce, NeverBounce, or Bouncer). Roughly $600 for 120k.
• Upload the CSV, download the verified result.
• Drop invalid, catch-all, disposable, and role-based addresses.
• Segment: warm is engaged in the last 6 months, cold is older or never engaged.
• Run the importer with --commit --segment=warm to write the warm segment.
• Tag everyone with migrated:hubspot-2026-05 for traceability.

Set up Resend, authenticate news.adway.ai, build the first React Email template, send a test to 5 to 10 colleagues.

• Create Resend account, start on the free tier.
• Authenticate news.adway.ai with MX, SPF, DKIM, and DMARC records.
• Install resend and @react-email/components.
• Build emails/templates/base-layout.tsx and test-send.tsx.
• Build scripts/send.ts and the supabase/functions/resend-webhook edge function.
• Send the test newsletter to a hardcoded internal list of 5 to 10 colleagues.

Build sender reputation on news.adway.ai by ramping the send volume gradually. Pause immediately if bounce or complaint rates spike.

• Day 1: send to the 500 most recently engaged warm contacts.
• Day 3: 1,000. Day 5: 2,500. Day 8: 5,000.
• Day 12: 10,000. Day 16: 20,000.
• Day 20 onward: full warm segment at normal cadence.
• Bounce rate target under 2%, spam complaint rate under 0.1%, warm open rate 20 to 40%.
• Pause sends if bounce passes 5% or complaints pass 0.3%.

Decide what to do with the older 40 to 60k cold contacts once the warm segment is sending cleanly. Plan recommends a single re-permission campaign.

• Option A: archive cold contacts and never email them. Safest for deliverability and GDPR.
• Option B (recommended): one re-permission campaign tied to a specific recent Adway shipment. Active equals those who opt back in.
• Option C: a slow re-engagement series of 3 pieces over 6 weeks. Higher reach recovery, higher reputation risk.
• Frame the re-permission email around something specific Adway has shipped, not a generic 'we miss you'.