Monolith to Multi-Tenant SaaS in 4 Sprints — No Rewrite Required

Can you turn a brittle legacy app into a multi-tenant SaaS without rewriting it from scratch? We just did. In four sprints, our team relaunched a seven-year-old e-commerce monolith as a subscription-based platform powered by Angular 19 SSR, Node 20 + Fastify and Terraform Cloud.This post distills everything that worked, what blew up in our faces, and a copy-paste migration checklist. Grab a coffee: 8-minute read.

1 — Why Even Migrate? (Hint: Money & Velocity) ☕

| Metric | Before | After | Delta |
|—-|—-|—-|—-|
| Monthly Release Cadence | 1 / month | 12 / month | × 12 |
| Infra Cost / Tenant | €165 | €97 | -41 % |
| LCP 75th p (field) | 4.1 s | 1.9 s | -54 % |
| Net Promoter Score | 34 | 65 | +31 |

ROI kicker: each 1-second LCP drop boosted funnel conversion by 6 %. Numbers made finance very, very happy.

2 — Audit the Monster in 3 Dimensions 🕵️

Before touching code, we ran a 3-D audit. Score every module 1 → 5:

Rule of thumb: anything scoring ≥ 4 goes into the “strangler fig” backlog—decouple after you stabilize the happy path.

3 — Architecture Choice: Modular Monolith + Feature Flags 🚀

Why not micro-services right away?

We chose Modular Monolith:

Single repo keeps onboarding trivial.
Feature flags let us ship dark features to one tenant at a time.
Move to services only when a module outgrows the monolith.

4 — Frontend Overhaul: Angular 19 with Native SSR 🖼️

# add server-side rendering in two commands
npx ng add @angular/ssr
npm run build:ssr && npm run serve:ssr

Two lessons learned

Lazy-hydrated Islands: heavy graphs & charts blew up renderApplication memory. We wrapped them with ngSkipHydration and hydrated on IntersectionObserver.
TC39 Temporal API: Angular 19’s new date pipes + Node 20 eliminated 30 kB of Moment.js dead weight.

Result: LCP .

5 — Backend & Tenancy: Fastify + Postgres RLS 🗄️

Fastify because 80 k req/s on a single M6g large with zero tuning.
Row-Level Security (policy USING (tenant_id = current_setting('app.tenant_id'))) keeps one DB until we hit 1 TB—then we partition.
Observability: OpenTelemetry → Grafana Cloud; one dashboard per tenant with UID templating.

6 — CI/CD: Green-Only Deploys in 45 Lines 📦

# .github/workflows/deploy.yml  (core)
on: [push]
jobs:
  test: …          # npm ci && npm test
  build_ssr: …     # npm run build:ssr
  deploy:
    needs: build_ssr
    runs-on: ubuntu-latest
    permissions: { id-token: write }
    steps:
      - uses: hashicorp/setup-terraform@v3
      - run: terraform init && terraform apply -auto-approve

Prod deploy in 11 min. If tests fail, prod is untouched.

7 — Security First (Really) 🔐

Fun fact: Week 1, 37 % of traffic was credential-stuffing bots—blocked automatically.

8 — Cost Lever Matrix 💸

9 — Five Lessons We Keep Re-Learning 🤹‍♂️

Feature flags > long-lived branches.
Measure field LCP, not just Lighthouse.
Docs or die. Every interface change = one ADR file.
Tenant-id on Day 0 – retro-fitting is hell.
Post-launch “broken windows” sprint saves morale.

10 — Pocket Checklist (Steal Me) ✅

☐ Build a coupling matrix
☐ Add tenant_id column everywhere now
☐ Ship risky slices behind flags
☐ Synthetic health check per tenant
☐ Schedule “Fix Broken Windows” sprint after go-live

Shipping a SaaS is never one-click magic. But with a modular plan, ruthless DevOps discipline and an obsession for user experience, you can turn a creaky monolith into a growth flywheel in under a month. Share your war stories below—let’s swap scars! 🚀