How DataOngoing Maps Enterprise Architecture

From Operational Friction to Deterministic Logic in Two-Week Sprints

The Logic Was Never the Bottleneck

Most mid-market and enterprise businesses running NetSuite, Salesforce or a multi-SaaS stack believe their software is broken or too slow.

When an order takes 15 seconds to save, when warehouse operators run lot allocation out of an offline spreadsheet, or when a CFO discovers that the gross margin in the monthly financials is off by six points because landed cost is sitting in a clearing account, leadership blames the platform.

Then the traditional consultancies step in. They propose a six-month "diagnostic and discovery phase" priced somewhere between $250,000 and $400,000. They send a team of junior analysts to interview department heads, compile a 120-page deck of high-level swimlanes, and conclude with a recommendation to spend another $500,000 on a complete reimplementation.

We reject that model. The software was never the bottleneck. The database engine was never the bottleneck. The thinking was the bottleneck.

Every enterprise failure mode in the more than 100 production system maps we have drawn traces back to one root cause: ungoverned architectural fragmentation. Business logic was layered on over five years, script on top of script, workflow on top of workflow, middleware webhook on top of middleware webhook, until the database lock queues choked and the operational truth detached from the general ledger.

We do not write decks that gather dust. We map the physical, economic reality of the business, model the deterministic database logic, and deploy working, tested software in two-week production sprints. Here is how we map architecture.

1. The Three Layers of Enterprise System Reality

To map an architecture that survives production scale, an architect has to separate the enterprise into three layers and work through them in order.

1. Layer 1. The economic reality

   Physical trucks, pallets, cold storage, vendor invoices, customer agreements, labor hours and margin velocity. Molecules and cash.

2. Layer 2. The governed record of truth (NetSuite or the ERP core)

   The ACID-compliant ledger, lot-numbered inventory, subrecords, revenue schedules, permissions and tax nexus. Deterministic logic and APIs.

3. Layer 3. The intelligent surface

   Mobile warehouse portals, OCR ingestion, CRM synchronization, custom Suitelets and governed AI worker toolchains. The part people see.

Traditional consulting starts at Layer 3, with what the screen looks like, and builds a new interface over corrupted database logic.

DataOngoing starts at Layer 1. We trace the physical movement of inventory, the contractual obligations in the customer agreement and the dollar-for-dollar flow of cash. Only when the economic physics are modeled do we architect Layer 2, the governed ledger rules, and deploy Layer 3, the surface.

2. The Four Structural Anti-Patterns We Remove

When we ingest and audit an enterprise architecture, the same four structural diseases show up again and again.

Anti-pattern 1: the cascading UI lock chain (the 15-second save)

• The symptom. Warehouse managers and customer service representatives wait 12 to 18 seconds every time they save a sales order or an item fulfillment. Under peak volume they receive RCRD_HAS_BEEN_CHANGED collision errors.
• The structural flaw. Eight user-event scripts attached to the same record type by four development teams over six years. Each one executes its own record.load() and search.create() calls synchronously on the user thread, exhausting governance points and holding table locks.
• The fix. The Single Entry Point router. Every trigger is consolidated into one dispatcher. Read-only validation runs in a sub-second beforeSubmit; GL allocations, heavy lifting and third-party notifications are queued to Map/Reduce pipelines. The save drops to under 800 milliseconds.

Anti-pattern 2: the lying gross margin (landed-cost clearing decay)

• The symptom. The controller reviews month-end and finds $250,000 sitting in an unreconciled freight and customs clearing account. Sales celebrated a 42% gross margin on paper; the true margin after landed cost was 31%.
• The structural flaw. Ocean freight, customs, drayage and cold storage are billed weeks after the inventory was received and sold. The month-end journal entry divorces the expense from the lot it belonged to.
• The fix. Automated landed-cost allocation: an estimated accrual captured on receipt, the actual vendor bill matched against the specific lot, and the COGS adjustment written to the transaction line. The margin on a sale is true the day it posts.

Anti-pattern 3: the shadow spreadsheet subledger (a $50M plant run on Excel)

• The symptom. A manufacturing facility or food distributor tracks production batch scheduling, design change requests or QA lot releases in a spreadsheet on a shared drive because "NetSuite couldn't handle the calculations."
• The structural flaw. Rigid standard forms push operational teams outside the system of record, and every re-keying introduces typographical errors, orphan records and broken audit trails.
• The fix. Native Suitelet operations portals: purpose-built interfaces inside NetSuite's own infrastructure. Operators get the calculation speed of a spreadsheet, and every click writes to a governed custom record with role-based permissions and a complete audit history.

Anti-pattern 4: the brittle middleware web

• The symptom. HubSpot, Stripe, Shopify, a 3PL and NetSuite are connected by a drag-and-drop iPaaS. When an API format changes or a webhook times out, orders drop silently, shipments go unbilled and someone reconciles by hand.
• The structural flaw. Middleware treats ERP tables like key-value stores and ignores transaction sublists, inventory location subrecords and currency revaluation.
• The fix. Direct RESTlet integration with idempotent queuing. External systems post to secured NetSuite endpoints that validate mandatory subrecords, respect concurrency limits and log payload hashes so nothing posts twice. The native-versus-middleware comparison has the cost side.

3. The Process Mapping Standard

Whether we are documenting an existing environment or architecting a ground-up modernization, a correctly mapped transaction moves through five stations. The person at the start never waits on the work at the end.

Five standards are non-negotiable on every map:

1. Deterministic state transitions. A transaction cannot exist in an ambiguous state. Every workflow has explicit, stage-gated statuses: Pending Allocation, Allocated, Staged, Fulfilled, Billed.
2. Subrecord and location governance. A line item is never an isolated scalar. Every inventory line binds to an inventory subrecord, a specific bin or lot, and a declared subsidiary and tax nexus.
3. Idempotency and replayability. Every automated transaction stores an external reference id. If an integration retries five times, it updates the existing record and never creates a phantom duplicate.
4. Asynchronous separation of concerns. Never make a human being wait on a computer for a calculation that does not change their next action.
5. Dollar-for-dollar reconcilability. Every automation that touches inventory or revenue leaves a trail a controller can audit, before and after, in under three minutes.

4. From Map to Production in Two Weeks

We do not believe in multi-year ERP initiatives that never show incremental value. Mapping happens first, in the 48-72 hour diligence read; the build that follows is a fixed-price, fourteen-day sprint with one named leak in scope.

If the code does not run in production, the milestone is not billed. The Two-Week Basis-Point Sprint page has the terms.

The Next Step

If your business is living with multi-minute save delays, unallocated landed-cost leakage or integrations held together by spreadsheets, you do not need another $300,000 slide deck. You need an architect who understands the balance-sheet impact of every line of code.

The four retrospectives that apply this map are published alongside this paper: perishable lot traceability with a live cutover, invoicing fired from the carrier scan, a HubSpot, Stripe and NetSuite revenue spine, and governed AI workers inside NetSuite.

Start with the 48-72 hour diligence read.

Figures in this paper are modeled composites unless a source class says otherwise. Source text: markdown version  •  Service: Technology Due Diligence for Private Equity