Process Evidence · 02
Establish domain language, boundaries, and state models
I defined boundaries and state models so terms such as payment, billing, course access, subscription, refund, and points would have consistent meanings across code and documents.
I divided the system by domain objects and state transitions—Order, PaymentIntent, Payment, Entitlement, Subscription, and Ledger—rather than by features.
How I worked
Actual Working Approach
- Input
I collected recurring nouns from requirement documents, the existing database, and payment and subscription code, state-change points with different lifecycles, and dimensions such as language and currency that change with global expansion.
- Questions to ask
I checked whether current state was mixed with historical facts, whether external side effects ran before internal intent was recorded, and whether access rights were treated as an incidental order field.
- Output
I established a domain model separating purchases, billing cycles, payment attempts, actual payments, course access, the ledger, and pending/final states.
Normalizing domain-modeling evidence through B2B audits, reflections, and project archives
For the domain-modeling card, I retained only material that explained lifecycles, ownership, and the separation of current state from historical facts, rather than lists of nouns.
- B2B models
- I grouped the actor, affiliation, and role-approval model audit with reflections on Company · Vendor, the user domain, Tax ↔ UI Category, VendorStaff, and Attachment as B2B modeling evidence.
- Production models
- I linked the Entitlement reflection separating purchase facts, viewing rights, subscription policy, and viewing history, language-specific catalog classification, and the Accrual · Ledger · Wallet reflection as representative verification evidence.
- Supporting examples
- The six eCRF responsibility boundaries, geofencing and AR commerce, content moving from templates to instances, and embedded Firestore value objects serve only as supporting examples of the same modeling practice.
- Stating limitations
- Relationship records with empty bodies and Study Room learning records are not used as direct evidence. Card copy reflects only material substantiated by code, documents, or reflections.
Separating authorization, affiliation, review, and categories into distinct domains
In a private B2B platform, I separated Company, Vendor, UserAffiliation, UserRole, CertificationRequest, TaxCategory, and UICategory. The model allowed one user to have affiliations and roles in multiple organizations, with separate approval lifecycles for company, vendor, and employee verification. A later audit separately recorded where this model intent diverged from actual code contracts.
Rendering the diagram.
Where This Led to Implementation
- Points Accounting & Rewards System
I separated expected payouts, the immutable ledger, and current balances into Accrual, Ledger, and Wallet, establishing boundaries between current state and historical facts.
View the Accrual · Ledger · Wallet model - Points Accounting & Rewards System · Global catalog rules
I separated product display language, category codes and translations, order currency, and notification fallback text so display language would not become entangled with accounting and payment rules.
View language and category models - Multi-Gateway Subscription Billing System
I divided video and subscription access policy into ProductVideo, SubscriptionProduct, and Access Policy, and payment state into Order, SubscriptionInvoice, PaymentIntent, and Payment.
View video, subscription, and payment models
- Input
- I extracted recurring nouns, states, and approving actors from actor-specific plans and verification, authorization, and category requirements.
- Boundaries
- I separated account identity, organizational affiliation, domain-specific roles, verification requests, tax categories, and display categories into models with different reasons to change.
- State model
- I treated company, vendor, and employee verification as separate pending facts with request, approval, and rejection histories, rather than as current affiliation information.
- Output
- I then organized APIs and the database around Identity, Affiliation, Certification, and Taxonomy boundaries rather than screen menus.
- Audit
- The employee-approval flow shows an intent to create affiliation and roles in one transaction. However, entity/service field mismatches and authorization-guard contract mismatches prevent me from presenting it as a completed RBAC case.
Documenting each model's invariants alongside its name
Instead of disclosing every table column, the portfolio describes the questions each model must answer and the conditions it must never violate at a level suitable for public disclosure.
- Payments and subscriptions
- I assigned purchase intent to Order, billing cycles to Invoice, gateway attempts to Intent, and actual authorization results to Payment.
- Points and ledger
- I separated Accrual as the payout commitment, Ledger as append-only evidence of changes, and Wallet as the current-balance snapshot.
- Global presentation
- I separated classification codes and locale-specific names, terms translations, payment currency, and notification-message selection so UI language and server-side canonical data would not be mixed in one model.
- Disclosure scope
- I explain design decisions through responsibilities, state transitions, and invariants, rather than disclosing complete internal table schemas.