ERP MVP — From Dashboard to Company System

TL;DR

A dashboard shows data. An ERP creates operations. The difference: dashboard reads, ERP writes. In 4 sprints you go from “eye on the number” to “company system running”.

Why evolve, instead of building from scratch

Typical founder tries to buy a big ERP (TOTVS, Senior, Bling) and:

• Implementation takes 6-12 months
• Customization becomes hostage to the vendor
• 60% of features they paid for never get used
• Team resists because changing is expensive

Another strategy: start from the dashboard (which you already built in lesson m06-build-dashboard) and evolve module by module.

ERP MVP stack

Layer	Tool	Why
Frontend + agent	Lovable	You describe, it builds
Database + Auth	Lovable Cloud	Internal database, auth, realtime — no external setup
Automation	n8n	NF-e, billing, email, integrations
Security	Aikido	Continuous code + cloud scanning
Payments	Stripe (B2B) or Asaas/Cobre Fácil (BR boleto)	Direct API
NF-e	NFE.io or Plug Notas	API + SEFAZ

Total monthly cost (real SMB operation): ~US$ 80-150/month. Compare with TOTVS at R$ 8-15k/month.

5 modules · 4 sprints

Sprint 1 (2 weeks) — Sales

Reuses the existing dashboard + adds writes:

• Customer registration screen (with tax ID validation)
• Order creation screen (with real-time margin calculation)
• Pipeline management screen (draggable Kanban: lead → qualif → proposal → closed)
• Commission auto-calculated per rep
• Discount approvals above X% require manager OK

Tables in Lovable Cloud:

customers (id, legal_name, tax_id, segment, rep_id, credit_limit, status)
orders (id, customer_id, rep_id, date, status, total_value, total_cost)
order_items (id, order_id, sku, qty, unit_price, discount_pct)
opportunities (id, customer_id, rep_id, stage, estimated_value, prob, expected_date)

Critical RLS: rep only sees their customers/orders. Manager sees everything in the region.

Sprint 2 (2 weeks) — Inventory + Purchasing

• Product registration screen (SKU, average cost, supplier, lead time)
• Stock movement (entry by purchase invoice · exit by sale)
• Automatic ABC curve (top 20% SKUs = X% of revenue)
• Reorder point (balance + lead time + turnover)
• Purchase suggestion grouped by supplier

Tables:

products (id, sku, description, category, average_cost, supplier_id, reorder_point)
stock (id, sku, balance, last_cost, last_movement)
stock_movements (id, sku, type, qty, unit_cost, origin, date, user_id)
purchase_orders (id, supplier_id, status, total_value, expected_delivery)
po_items (id, po_id, sku, qty, negotiated_price)

n8n automation: when balance hits reorder point, auto-create draft PO and notify purchasing in Slack.

Sprint 3 (2 weeks) — Finance

Reuses the financial spreadsheet as a starting point. Adds:

• Accounts receivable with automatic aging
• Collection cadence (D-3 reminder · D+1 overdue · D+7 negative listing)
• Semi-automatic bank reconciliation (imports OFX/CSV)
• Projected cash flow 13 weeks (rolling)
• Cost center by expense category

Tables:

accounts_receivable (id, order_id, customer_id, value, due_date, paid_at, status)
accounts_payable (id, supplier_id, category, value, due_date, paid_at, status)
bank_transactions (id, account_id, date, value, type, reconciled_with)
bank_accounts (id, bank, branch, account, current_balance)

n8n automation:

• Boleto generated via Asaas/Cobre Fácil when order invoices
• Optional dynamic PIX
• D-3, D+1, D+7 personalized email

Sprint 4 (1 week) — NF-e + Collections + Permissions

• NF-e issuance via NFE.io or Plug Notas (CFOP/CST based on customer profile)
• XML stored in Lovable Cloud Storage
• Automatic collections via boleto + PIX (Asaas, Cobre Fácil, Mercado Pago)
• Granular per-user permissions (Admin, Manager, Rep, Purchasing, Finance)
• Audit log of who changed what

Tables:

invoices (id, order_id, access_key, sefaz_status, xml_url, total_value)
users (id, email, role, region, active)
permissions (role, resource, actions)
audit_log (id, user_id, action, entity, date, payload_old, payload_new)

The 5 ready modules

┌──────────────────────────────────────────────────────────┐
│   SALES     │  INVENTORY  │   FINANCE   │  NF-e  │ AUTH  │
├──────────────────────────────────────────────────────────┤
│  Customers  │  Products   │  AR / AP    │  Issue │ Roles │
│  Orders     │  Movements  │  Aging      │  XML   │ RLS   │
│  Pipeline   │  Purchases  │  Collection │  Cancel│ Audit │
│  Commission │  ABC Curve  │  Cash 13w   │        │       │
└──────────────────────────────────────────────────────────┘

The master prompt for Lovable

Paste this to generate the initial skeleton:

Build a lean ERP for Distribuidora Aurora (B2B electrical materials).

Stack: Lovable + Lovable Cloud (internal database) + n8n (automations).
Users: Admin, Manager, Rep, Purchasing, Finance — RLS by role + rep only sees their own customers.

Modules for Sprint 1:
1. Customers (CRUD with tax ID validation + credit limit + segment)
2. Orders (header + items · real-time margin calculation · approval for discount > 8%)
3. Pipeline (Kanban: Lead → Qualif → Proposal → Negotiation → Won/Lost)
4. Commissions (automatic calculation per rep, monthly closing)

Visual: dark mode, OKLCH palette blue + violet + mint, Inter Tight + Inter typography.
Google + email/password login.
Mobile-first with tables that become cards on mobile.

Import initial data via manual spreadsheet upload (will be provided by the user).

Return the running app + public URL + checklist of what's missing for Sprint 2.

Metrics to track evolution

Sprint	Success metric
Sprint 1	100% of new orders enter through the system (not Excel)
Sprint 2	System balance matches physical count with error < 2%
Sprint 3	Aging of overdue accounts drops 30% in 60 days
Sprint 4	Invoice issued in < 3 min after billing (vs 15 min manual)

Pegadinhas

• ! Migration is the real pain. Importing historical Excel is 40% of the work. Reserve a full week just for this.
• ! Human approval before automating 100%. Automatic collection that wrongly negativizes a customer becomes liability. Create a human checkpoint in the first 90 days.
• ! Backup BEFORE go-live. Lovable Cloud with backup active. Configure manual daily backup to S3 too.
• ! NF-e is regulatory. Use a provider (NFE.io, Plug Notas) — don't build from scratch. Penalty for wrong NF is high.
• ! Performance matters early. 50k orders in a non-indexed table chokes. Create indexes on customer_id, rep_id, date from the beginning.
• ! Don't migrate everything at once. Sprint 1 (sales) entirely · then Sprint 2. Don't try to open all modules together.

Production checklist (before releasing to the team)

• Lovable Cloud with backup enabled
• External daily backup to S3/Backblaze (don’t depend only on Lovable Cloud)
• Aikido scanning code + cloud + container
• Custom domain configured with SSL
• Human approval on irreversible actions (cancel order, negative-list customer)
• Audit log of who changed what
• Team training (minimum 2 sessions of 1h)
• Rollback plan if things go wrong
• Contingency support (you or someone trusted available for 2 weeks post-go-live)

Real total cost

Estimate for SMB with 10-30 active users:

Item	Monthly
Lovable Pro	US$ 25
Lovable Cloud (included in Pro)	US$ 0
n8n Cloud Starter	US$ 20
Aikido (free up to 10 repos)	US$ 0
NFE.io or Plug Notas (~500 NFs/month)	~$ 20
Asaas (collection, ~$ 0.30/boleto)	variable
Custom domain	~US$ 1
Total	~US$ 75-110/month + variables

About $ 120/month to run an SMB ERP, vs $ 1,500-3,000 for a traditional ERP. Difference of 20-25x with more flexibility.

When NOT to follow this path

• Highly regulated industry (pharma, food with traceability)
• Complex tax operation (interstate transfer with tax substitution, ICMS-ST)
• Massive volume (1M+ orders/month)
• Mandatory compliance audit (SOC 2, ISO 27001 from day 1)

For 80% of Brazilian SMBs, the stack here solves it.

Next step

You finished the full technical track. Paths:

• Apply: follow Sprint 1 with your business spreadsheet
• Go deeper: full prompt library with 51 templates
• Contribute: github.com/jvictordevai/openacademy