DEV Community

Fidwell Engineering
Fidwell Engineering

Posted on

Fixed-Scope vs. Time-and-Materials for Custom Business Apps: Engineering C01-C08 Written Boundaries Without Scope Creep

Short Answer: Pure Time-and-Materials contracts often absorb avoidable rework when UX flows and legacy ERP touchpoints stay undefined, whereas poorly bounded fixed-bid contracts break when teams hardcode assumptions. Splitting delivery into an explicit C01-C08 written scope contract, AI-assisted typed scaffolding + senior security review, and 5-Gate cloud staging lets SME buyers run custom web apps, LINE mini-programs, and lightweight business tools on predictable budget tiers (indicative reference bands USD 980–5,680 or JPY 180k–980k, conditional on signed C01-C08 written scope and complexity assessment) with full Git source handover.


Why Custom App Estimates Blow Up Under Both Classic Models

In real B2B software delivery — wholesale order portals, shop-floor barcode reporting, approval back-offices, or messaging mini-apps — cost overruns usually come from two architectural gaps rather than raw coding difficulty:

Failure Mode Why Time-and-Materials Suffers Why Unbounded Fixed-Bid Fails Engineering Remedy
Unspecified UI & workflow states Client tweaks screens after backend starts; hours compound Vendor trims edge-case validation to protect margin Interactive cloud sandbox check early via C02 screen inventory
Legacy ERP table coupling Engineers spend weeks decoding undocumented ledger schemas Direct database writes risk corrupting client accounting Read-Only DB View + Idempotent Outbox isolation boundary
Hand-coded repetitive CRUD Senior hourly rates burned on boilerplate tables & exports Corners cut on RBAC checks and migration scripts Schema-first codegen + AI boilerplate + senior lock/auth audit
Unowned accounts & source code Re-engagement locked to external agency credentials Closed binaries or proprietary builders block future edits Client-owned cloud/channel credentials + Git/DDL handover

Encoding C01-C08 Written Scope Before Writing Feature Code

Before sprint one begins, both engineering and buyer teams freeze a one-page JSON/Markdown scope ledger (C01 through C08) where explicit non-goals (C06) carry as much weight as included screens:

{
  "contractSpec": "C01-C08",
  "C01_roles": ["dealer_buyer", "warehouse_clerk", "finance_approver"],
  "C02_primaryViews": ["token_auth_entry", "catalog_order", "status_list", "csv_export"],
  "C03_legacyBoundary": {
    "inboundSync": "read_only_sql_view_inventory_and_price_tiers",
    "outboundWrite": "postgresql_idempotent_outbox_queue",
    "directLedgerMutation": false
  },
  "C04_externalIntegrations": ["messaging_id_token_verify", "csv_sftp_batch"],
  "C05_acceptanceGates": ["rbac_negative_tests", "duplicate_submit_idempotency", "migration_up_down"],
  "C06_explicitOutOfScopePhase1": [
    "multi_plant_aps_scheduling",
    "general_ledger_accounting_rewrite",
    "offline_native_hardware_drivers"
  ],
  "C07_deliverables": ["git_repository", "sql_ddl_migrations", "container_compose", "runbook"],
  "C08_accountSovereignty": "all_cloud_domains_and_channel_ids_under_buyer_company_entity"
}
Enter fullscreen mode Exit fullscreen mode

By locking C06_explicitOutOfScopePhase1, phase-one projects fit realistic budget envelopes instead of forcing full enterprise replacement on day one. Conditional on signed C01-C08 scope, typical starter tiers can scale incrementally from single-flow forms/prototypes (~USD 980–1,480 / JPY 180k–380k indicative range) up through multi-role ordering portals and Read-Only ERP Outbox adapters (~USD 1,780–5,680 / JPY 480k–980k indicative range), while preserving clean extension seams for Phase 2.


Read-Only View + Idempotent Outbox: Protecting Legacy ERPs Under Retries

Whether delivered on fixed scope or staged iterations, connecting a modern web/mobile interface to an existing ERP requires an anti-corruption queue in PostgreSQL so shaky factory Wi-Fi or webhook timeouts never double-post production orders or shipments:

CREATE TABLE erp_dispatch_outbox (
  id                BIGSERIAL PRIMARY KEY,
  idempotency_key   TEXT NOT NULL UNIQUE,
  step_state        TEXT NOT NULL CHECK (
    step_state IN ('PENDING', 'LEASED', 'ACKED', 'RECONCILE_REQUIRED')
  ),
  fencing_token     BIGINT NOT NULL DEFAULT 1,
  payload           JSONB NOT NULL,
  created_at        TIMESTAMPTZ NOT NULL DEFAULT now()
);

-- Conditional fencing update prevents stale worker commits after lease takeover
UPDATE erp_dispatch_outbox
SET step_state = 'ACKED'
WHERE id = $1
  AND step_state = 'LEASED'
  AND fencing_token = $2;
Enter fullscreen mode Exit fullscreen mode

In our self-hosted orchestration kernel ZeroOS, non-idempotent legacy steps transition into RECONCILE_REQUIRED rather than blindly re-firing writes when an upstream response goes ambiguous. When technical teams also need deterministic specification/manual retrieval alongside custom portals, LongMemOS (~3.1 MB binary, embedded_models=false, internal LongMemEval benchmark score 463/500 [92.6%]) attaches verbatim source file + line-range evidence replay without external vector DB subscriptions.


The 5-Gate Staging Pipeline Before Final Handover

Instead of treating acceptance as an end-of-project surprise, every milestone passes five deterministic gates:

  1. Gate 1 — Early Clickable Cloud Sandbox: Typically delivered around 2–4 business days after deposit and C01-C08 sign-off (adjusted by screen complexity) so stakeholders verify actual navigation before deep integration code is written.
  2. Gate 2 — Contract & Negative RBAC Suite: Verify role isolation and signed server-side token validation against client-supplied payload tampering.
  3. Gate 3 — Clean-Slate Schema Migration Check: Rebuild PostgreSQL tables, indexes, and seed fixtures from scratch in CI.
  4. Gate 4 — Retry & Idempotency Drill: Replay identical Idempotency-Key requests across simulated network drops and verify single-row ledger integrity.
  5. Gate 5 — Sovereign Asset Handover: Transfer the complete Git history, DDL scripts, environment template, and deployment runbooks into buyer-owned repositories and cloud accounts.

Scoping Consultation

Engineering teams or technical buyers looking to evaluate C01-C08 written scope boundaries for a custom business app, ERP read-only bridge, or offshore Phase 1 build can reach the architecture team via the on-site inquiry form:

Top comments (0)