Settle — Prerequisites & Decision Register

Settle has more infrastructure surface area than a typical SaaS application. Three capabilities force decisions that most apps defer indefinitely: column-level encryption for PII (SSNs, death certificates), physical mail dispatch for Tier 2 notifications, and a state-specific rules engine covering 50 US jurisdictions. This section strips each down to its minimal viable form before any vendor is added.

Questioning Each Potential Service

The following table forces a deferral decision for every service that is commonly added "just in case." The decision is recorded here so it does not have to be re-litigated during architecture review.

Each decision below includes the options considered, a comparison of trade-offs, and a recorded recommendation. Decisions marked Decided are final for MVP. Decisions marked Deferred are explicitly not being made yet. Revisit deferred decisions when their stated trigger is reached.

-- Column definition (never store plaintext SSN)
ALTER TABLE estate_contacts
  ADD COLUMN ssn_encrypted TEXT;

-- Write (application layer, key from env)
UPDATE estate_contacts
   SET ssn_encrypted = pgp_sym_encrypt($1, $2)
 WHERE id = $3;
-- $2 = process.env.DB_ENCRYPTION_KEY

-- Read (decrypt only in application, not in views)
SELECT pgp_sym_decrypt(ssn_encrypted::bytea, $1)
  FROM estate_contacts
 WHERE id = $2;

Every vendor in the stack must have a documented exit path before it is adopted. This table defines that path for each MVP service. "Lock-in level" measures how much application code changes if the vendor is replaced.

This is the complete list for local development and the first deploy. MVP requires five accounts. The total time to create accounts and run docker compose up locally should be under 30 minutes.

Minimum .env Template

# Database
DATABASE_URL=postgresql://settle:settle@localhost:5432/settle

# Encryption — generate with: node -e "console.log(require('crypto').randomBytes(32).toString('base64'))"
DB_ENCRYPTION_KEY=<32-byte base64 string>

# Session
SESSION_SECRET=<random 64-char string>

# Email
RESEND_API_KEY=re_<your key>
EMAIL_FROM="Settle <notifications@yourdomain.com>"

# Error tracking
SENTRY_DSN=<your dsn>
NODE_ENV=development

The maximum external service count is 5 at MVP (6 counting Lob when it ships). This is a firm constraint, not a guideline. Every service added increases operational surface area for a solo or two-person team. Adding a service requires removing or justifying an existing one.

Infrastructure costs at three stages. The goal is $0–25/month at MVP with generous free tiers carrying the load. The first significant cost step is Lob when Tier 2 ships — that cost must be priced into subscription tiers, not absorbed.

These questions are explicitly not answered yet. Each has a stated trigger — the event that forces the decision. Do not answer them early. Do not let them block MVP development.

• Q-01 — Trigger: User feedback
  When does print-and-mail friction become an integration requirement for Lob?
  Measure: support tickets citing print-and-mail difficulty, or funnel drop-off after PDF download. Threshold: >10% of estates abandon at the mail step, or >20 support tickets/month on the topic. Action: implement Lob behind IMailDispatcher. Price per-estate tier to cover $15/estate in mail costs.
• Q-02 — Trigger: Database size
  When do documents move out of Postgres into dedicated object storage?
  Threshold: total document storage exceeds 5 GB or total document count exceeds 5,000, whichever comes first. Action: implement Cloudflare R2 behind a storage repository interface. Migration script: read each bytea blob, write to R2, replace with R2 URL.
• Q-03 — Trigger: User feedback
  When does the manual benefit discovery checklist get automated?
  Threshold: users report the manual checklist as the most frustrating part of the 38-task plan, or benefit discovery rate is measurably lower than comparable estates completing it with professional help. Action: identify the 3–5 most valuable benefit APIs (SSA, PBGC, state unclaimed property) and integrate one at a time.
• Q-04 — Trigger: Year 2 roadmap
  Does Settle need a native mobile app?
  This affects the frontend framework choice. If yes: evaluate whether a React migration is warranted for React Native code sharing, or whether a separate Flutter app is preferable. If no: SvelteKit with a responsive PWA is sufficient. Decide before the Year 2 roadmap is written, not during it.
• Q-05 — Trigger: Enterprise or institutional sales
  When does SOC 2 certification become a sales requirement?
  SOC 2 Type II requires: formal access controls, encryption key management (env variable is insufficient — move to KMS), audit logging, vendor agreements (BAAs), and a formal incident response policy. Estimated lead time: 6 months. Do not begin the process without a closed deal that requires it, or a clear pipeline of deals that do.
• Q-06 — Trigger: Revenue validation
  What is the per-estate pricing model, and does it cover Lob costs at Year 2 scale?
  Required before Lob is integrated. The Lob cost at 5,000 estates/year is ~$75,000. A $200/estate subscription with 37.5% allocated to mail costs covers it. Validate actual mail volume per estate with the first 50 manually-processed estates before committing to an integration pricing model.