LeadBridge

What we actually built (week-by-week)

• Weeks 1-2 — discovery and ISR landings. Mapped the 14 neighbourhoods the agency operates in and built per-area landing pages on Next.js with Incremental Static Regeneration: same template, unique copy, signals (parking, transport, schools, average €/m²) pulled from a small CMS table.
• Week 3 — lead form with weighted scoring. A short form (intent, budget, timing, financing) plus an internal score (0-100) computed server-side from the form payload, so the agency could prioritise without manual reading.
• Week 4 — agent panel and round-robin assignment. Lightweight panel with assigned leads, status (new / contacted / visit / offer / closed), notes and reassignment. No fancy permissions: 12 agents and a manager view.
• Week 5 — automation: 3-step email + Telegram alert. Sequence in Resend (5 min, 24 h, 72 h) with merge fields from the lead, plus a Telegram bot pinging the manager when a lead has score ≥ 70 and no contact in 4 hours.
• Week 6 — analytics and handover. Channel attribution panel (organic landing, portal, Instagram) with cost per lead and conversion to visit; recorded video session for agents and 30 days of monitoring + bugfix included.

Concrete technical choices and why

Next.js with ISR on Vercel was picked over WordPress so neighbourhood landings could be edited by the agency in a Notion-like table without redeploying, while keeping near-static performance for SEO. PostgreSQL on Supabase instead of a homemade DB to ship the agent panel faster with row-level security. Resend for transactional email because deliverability mattered for follow-up; Telegram for internal alerts because the manager already lived there. No Salesforce, no HubSpot: the cheapest tool that solved the actual bottleneck.

Cost ceiling for the MVP was set at ~€90/month of cloud spend; we exited the engagement under that figure with room for ~5× growth before re-architecting.

Lessons we apply to similar real-estate MVPs

The score field matters more than the form length: agents trusted a 0-100 number with traffic-light colours much more than a multi-page qualification flow. Email sequencing only paid off when paired with the Telegram alert; without the human escalation, 24 h follow-up still slipped.

Per-area landings outperformed a single "cities we cover" page by a wide margin: each landing ranked for long-tail queries (e.g. "pisos en venta + barrio + 2 hab") that the agency's website never targeted before. We replicate the same playbook in other real-estate engagements when the agency has at least 8-10 well-defined catchment areas.