Skip to content

System Invariants & Design Principles

Core rules and philosophy governing CostEngine (MfgIQ).

System Invariants

These are non-negotiable rules that the system MUST enforce at all times.

1. No Black-Box Calculations

Rule: Every cost value must be traceable to its source data and formula.

Why: MSMEs need to defend quotes to customers and auditors. "The system calculated it" is not acceptable.

Enforcement: - All formulas are visible - Source data is linked - Calculation steps are shown - No proprietary algorithms that hide logic

Example: 

Material Cost: ₹263.95
└─ Formula: (Gross Weight × RM Rate) / Yield
   ├─ Gross Weight: 2.304 kg (from Excel cell D13)
   ├─ RM Rate: ₹99.5/kg (from Material Master, updated 2026-01-20)
   └─ Yield: 86.8% (calculated: Net Weight / Gross Weight)

2. No Deletion of Historical Versions

Rule: All costing versions are immutable. Updates create new versions; old versions remain accessible forever.

Why: - Customer disputes require historical quote references - Trend analysis needs version history - Audit compliance demands immutability - Learning from past quotes

Enforcement: - Database constraints prevent deletion - "Delete" UI buttons disabled for versions - Only "Archive" (soft delete) allowed for drafts - Version numbers are monotonically increasing

Version Lifecycle: 

Draft → v1.0 (Published) → v1.1 (Updated) → v2.0 (Major Change)

All versions remain queryable and comparable

3. Human Judgement is Preserved

Rule: The system captures and honors manual overrides while making them visible and traceable.

Why: Experienced estimators have knowledge the system doesn't (shop floor conditions, vendor relationships, customer expectations).

Enforcement: - Every override requires a reason - Original calculated value is preserved - Override is highlighted in UI - Impact on total cost is shown - Override author and timestamp tracked

Override Example: 

Cycle Time: 1.25 hrs (OVERRIDDEN)
├─ System Calculated: 1.50 hrs
├─ Override Value: 1.25 hrs
├─ Reason: "Process improvement after trial run - verified with shop floor"
├─ Overridden by: John Smith
├─ Date: 2026-01-24 10:15 AM
└─ Cost Impact: -₹30 per piece (-5.2%)

Design Principles

1. Excel-First Onboarding

Principle: Absorb existing Excel logic, don't force rebuilding from scratch.

Implementation: - Excel template import - Field mapping configuration - Gradual formula migration - Preserve cell references in traceability

User Impact: Onboarding in days, not months

2. Operations-First Model

Principle: Manufacturing is about operations, not just materials.

Implementation: - Setup vs cycle time at core - Batch economics built-in - Tooling costs tracked - MHR (Machine Hour Rate) as first-class concept

Competitive Edge: Others focus on material BOM; we nail the gritty operations reality

3. Explainability Over Automation

Principle: Defend every number before optimizing speed.

Trade-off: - Slower than pure AI/ML pricing - More manual than fully automated systems - But: Trustworthy and defensible

User Impact: Estimators trust the system because they understand it

4. Cost-Price Separation

Principle: Margin is a commercial decision, not a costing input.

Implementation: - Cost calculation doesn't know about margins - Pricing layer is separate - Multiple margin scenarios without re-costing - Customer-specific pricing rules

Benefit: Change pricing strategy without touching cost data

Founder Note

"This is not about replacing estimators with algorithms."

The Real Problem

MSMEs don't lack costing data - they have too much: - Excel sheets per customer - Tribal knowledge per estimator - Inconsistent formats - No version control - Lost quotes due to errors

Our Solution Philosophy

Respect MSME mental models while adding structure:

  1. Capture expertise (don't replace it)
  2. Estimators know their shop floor

  3. System makes their knowledge transferable

  4. Reduce repetitive work (don't eliminate judgment)

  5. Automate calculations

  6. Preserve decision-making

  7. Make knowledge institutional (not just personal)

  8. New estimators learn from system

  9. Consistency across team

  10. Provide defensibility (not just speed)

  11. "Why this price?" has an answer
  12. Customers trust explained quotes

The Goal

Turn tribal knowledge into institutional knowledge while preserving the human element that makes manufacturing costing an art as much as a science.

Metrics of Success

Not "how fast?" but: - Can the estimator explain every line? - Would a customer accept the breakdown? - Can a new hire use this in week 1? - Does the owner sleep better knowing quotes are defensible?

Anti-Patterns to Avoid

What we deliberately DON'T build:

Anti-Pattern Why We Avoid It
AI Auto-Pricing Removes human judgment, creates black box
ERP Integration First Forces ERP adoption before value delivery
CAD Dependency Excludes process manufacturers without CAD
Fancy Dashboards Distracts from core workflow (quote creation)
Multi-Plant Day 1 Adds complexity before proving single-plant value

Evolution Strategy

MVP (First 6 Months)

Focus on single costing workflow done right: - Excel import - Core calculations - Explainability - Quote output - Versioning

Measure: Can we replace Excel for 80% of quotes?

Growth (6-18 Months)

Add customer-specific intelligence: - Margin rules per customer - Payment terms impact - Historical pricing trends - Win/loss tracking

Measure: Are quotes more likely to win?

Scale (18+ Months)

Build platform capabilities: - API for integrations - Multi-factory support - Advanced analytics - Procurement integration

Measure: Does it become the "costing layer" for manufacturing?