Manufacturing & Supply Chain

End-to-end walkthroughs showing how a precision parts manufacturer uses the MATIH Platform to unify IoT sensor data, predict equipment failures, automate quality inspection, and optimize plant operations with AI-driven analytics.

Industry Context

Manufacturing generates more data per facility than almost any other industry -- yet most of it goes unused. A single CNC machine produces thousands of sensor readings per minute: vibration, temperature, spindle speed, coolant pressure, power draw. Multiply that by hundreds of machines across multiple plants, and you have a firehose of time-series data that rarely makes it past the historian database.

Meanwhile, production planning lives in SAP, quality records sit in a separate MES, supplier data arrives as CSV attachments, and energy consumption is tracked by yet another system. The result: maintenance teams react to breakdowns instead of preventing them, quality engineers catch defects after they reach customers, and plant managers piece together OEE numbers from spreadsheets.

The MATIH Platform bridges these silos. IoT sensor streams, ERP transactions, quality records, and supply chain data flow into a single governed environment. Data scientists build predictive maintenance models. ML engineers deploy real-time inspection systems. BI leads create plant floor dashboards. And executives ask natural language questions about cross-plant performance.

Challenge	Impact	Platform Response
IoT data volume	100M+ sensor readings/day across 200 machines, mostly unused	Streaming ingestion via Kafka, time-series aggregation, S3/DuckDB for efficient querying
Predictive maintenance	Unplanned downtime costs $10K-50K per hour per production line	ML Workbench for survival models, Ray Serve for real-time health scoring
Quality control	Manual inspection catches only 85% of defects, costs $2M/year in scrap	Visual inspection models, automated defect classification, quality gate pipelines
Supply chain visibility	Late supplier deliveries cause production schedule disruptions 3x/month	Federated queries across ERP, supplier portal, and logistics data
Energy optimization	Energy is 15-25% of production cost, with significant waste during idle periods	Real-time energy dashboards, ML-driven consumption optimization
Cross-plant standardization	Each plant uses different codes, naming conventions, and processes	Ontology Service for taxonomy standardization, semantic layer for unified metrics

Company Profile: Apex Manufacturing

All walkthroughs in this section follow employees at Apex Manufacturing, a fictional precision parts manufacturer with the following profile:

Attribute	Value
Annual Revenue	$320M
Products	Precision-machined aerospace and automotive components
Plants	4 facilities across 2 states
Equipment	200 CNC machines (lathes, mills, grinders, EDM)
Workforce	1,200 production employees, 45 engineers
Shifts	2 shifts (6am-2pm, 2pm-10pm), 5 days/week
Data Team	3 data scientists, 2 ML engineers, 2 BI analysts, 1 COO
Key Customers	Tier 1 aerospace suppliers, automotive OEMs
Certifications	AS9100D (aerospace), IATF 16949 (automotive), ISO 14001 (environmental)

Sample Datasets

These are the core datasets used across all four walkthroughs. In a production deployment, these tables live in their respective source systems and are ingested into the platform via Airbyte connectors, Kafka streaming, or file imports.

Sensor and Machine Data

Dataset	Source	Rows	Key Columns
`sensor_readings`	IoT Gateway (Kafka)	100M	`machine_id`, `sensor_type`, `value`, `unit`, `timestamp`, `quality_flag`
`equipment_registry`	SAP PostgreSQL	5K	`machine_id`, `machine_type`, `manufacturer`, `install_date`, `plant_id`, `line_id`
`maintenance_logs`	CMMS PostgreSQL	200K	`log_id`, `machine_id`, `maintenance_type`, `description`, `technician_id`, `duration_hours`, `parts_replaced`, `timestamp`

Production and Quality Data

Dataset	Source	Rows	Key Columns
`production_orders`	SAP PostgreSQL	500K	`order_id`, `product_id`, `machine_id`, `planned_qty`, `actual_qty`, `start_time`, `end_time`, `status`
`quality_inspections`	QMS PostgreSQL	1M	`inspection_id`, `order_id`, `machine_id`, `dimension_measured`, `spec_min`, `spec_max`, `actual_value`, `pass_fail`, `inspector_id`
`defect_images`	Image Metadata DB	250K	`image_id`, `inspection_id`, `defect_type`, `confidence_score`, `bounding_box`, `camera_id`, `timestamp`

Supply Chain and Energy Data

Dataset	Source	Rows	Key Columns
`supplier_deliveries`	Supplier Portal CSV	300K	`delivery_id`, `supplier_id`, `material_id`, `ordered_qty`, `delivered_qty`, `promised_date`, `actual_date`, `quality_grade`
`energy_consumption`	Smart Meters (Kafka)	50M	`meter_id`, `machine_id`, `plant_id`, `kwh`, `power_factor`, `demand_kw`, `timestamp`
`bill_of_materials`	SAP PostgreSQL	85K	`bom_id`, `parent_part`, `child_part`, `quantity_per`, `unit`, `lead_time_days`

Data Sources

Apex Manufacturing's data lives in six systems. The platform connects to all of them through the Ingestion Service (Airbyte connectors), Kafka streaming, and file imports.

┌──────────────────────────────────────────────────────────────────────┐
│                     MATIH Ingestion Layer                            │
│              (Airbyte Connectors + Kafka + File Import)              │
└──────┬────────┬───────────┬──────────┬──────────┬──────────┬────────┘
       │        │           │          │          │          │
 ┌─────▼────┐ ┌─▼────────┐ ┌▼────────┐ ┌▼───────┐ ┌▼──────┐ ┌▼───────┐
 │ SCADA /  │ │ SAP ERP  │ │  IoT    │ │ QMS    │ │ CSV   │ │ Smart  │
 │Historian │ │PostgreSQL│ │ Gateway │ │Postgres│ │Supplier│ │ Meters │
 │(OPC-UA)  │ │          │ │(MQTT→   │ │        │ │ Portal│ │(Kafka) │
 │          │ │equipment │ │ Kafka)  │ │quality │ │       │ │        │
 │ machine  │ │production│ │ sensor  │ │inspect │ │deliver│ │ energy │
 │ status   │ │orders    │ │readings │ │defects │ │ies    │ │ usage  │
 │ alarms   │ │materials │ │         │ │        │ │       │ │        │
 └──────────┘ └──────────┘ └─────────┘ └────────┘ └───────┘ └────────┘

Source	Connector Type	Sync Mode	Frequency
SCADA / Historian (OPC-UA)	Airbyte custom connector	Incremental (timestamp)	Every 5 minutes
SAP ERP PostgreSQL	Airbyte PostgreSQL connector	CDC (incremental)	Every 15 minutes
IoT Gateway (MQTT to Kafka)	Kafka streaming ingestion	Streaming (real-time)	Continuous
Quality Management System	Airbyte PostgreSQL connector	CDC (incremental)	Every 15 minutes
Supplier Portal exports	File Import (Data Workbench)	Full refresh	Weekly
Smart Energy Meters	Kafka streaming ingestion	Streaming (real-time)	Continuous (1-min intervals)

Data Flow Architecture

                        Apex Manufacturing Data Flow

  ┌─────────────────────────────────────────────────────────────────────┐
  │                       STREAMING LAYER                               │
  │                                                                     │
  │   IoT Sensors ──▶ MQTT ──▶ Kafka ──▶ 5-min Aggregation ──▶ S3     │
  │   Energy Meters ──▶ Kafka ──▶ 1-min Aggregation ──▶ S3            │
  └──────────────────────────────┬──────────────────────────────────────┘
                                 │
  ┌──────────────────────────────▼──────────────────────────────────────┐
  │                      PLATFORM DATA LAYER                            │
  │                                                                     │
  │  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────────────┐    │
  │  │ Catalog  │  │  Query   │  │ Quality  │  │   Governance     │    │
  │  │ Service  │  │  Engine  │  │ Service  │  │   Service        │    │
  │  │          │  │ (Trino + │  │ (GX)     │  │ (access control, │    │
  │  │ 47 sensor│  │  DuckDB) │  │          │  │  audit trails)   │    │
  │  │ types    │  │          │  │ sensor   │  │                  │    │
  │  │ profiled │  │ time-    │  │ quality  │  │                  │    │
  │  │          │  │ series   │  │ gates    │  │                  │    │
  │  └──────────┘  └──────────┘  └──────────┘  └──────────────────┘    │
  └──────────────────────────────┬──────────────────────────────────────┘
                                 │
          ┌──────────────────────┼──────────────────────┐
          ▼                      ▼                      ▼
  ┌──────────────┐    ┌──────────────┐     ┌───────────────────┐
  │ ML Workbench │    │ BI Workbench │     │ Agentic Workbench │
  │              │    │              │     │                   │
  │ Predictive   │    │ OEE          │     │ "Why did Plant 2  │
  │ maintenance, │    │ dashboards,  │     │  OEE drop last    │
  │ quality      │    │ energy       │     │  week?"           │
  │ models       │    │ analytics    │     │                   │
  └──────────────┘    └──────────────┘     └───────────────────┘

Business KPIs

Apex Manufacturing tracks these key performance indicators across all workbenches and dashboards. Each walkthrough shows how the platform computes, monitors, and acts on these metrics.

Equipment Performance

KPI	Definition	Current Value	Target
OEE (Overall Equipment Effectiveness)	Availability x Performance x Quality	72.4%	> 85% (world-class)
Availability Rate	Run time / planned production time	88.1%	> 95%
Performance Rate	Actual output / theoretical max output	91.2%	> 95%
Quality Rate	Good parts / total parts produced	90.1%	> 99%
MTBF (Mean Time Between Failures)	Average operating hours between unplanned stops	340 hours	> 500 hours
MTTR (Mean Time To Repair)	Average hours to restore machine to production	4.2 hours	< 2 hours

Production and Quality

KPI	Definition	Current Value	Target
First-Pass Yield	% of parts meeting spec on first inspection	94.3%	> 98%
Scrap Rate	Scrap cost / total production cost	3.8%	< 2%
On-Time Delivery	% of orders delivered by promised date	91.7%	> 98%
Cycle Time Variance	Actual cycle time vs standard (std dev)	+/- 12%	+/- 5%

Supply Chain and Cost

KPI	Definition	Current Value	Target
Supplier OTD	% of supplier deliveries on time	87.3%	> 95%
Energy per Unit	kWh consumed per part produced	2.8 kWh	< 2.2 kWh
Maintenance Cost per Unit	Total maintenance spend / parts produced	$1.42	< $1.00
Inventory Days of Supply	Raw material inventory / daily consumption	18 days	10-12 days

Persona Walkthroughs

Each walkthrough follows one persona through all eight lifecycle stages, using real Apex Manufacturing data and scenarios. Start with the role closest to yours, or read all four to see how the platform enables cross-functional collaboration.

Walkthrough	Persona	Scenario	Primary Workbenches
Data Scientist Journey	Lin Wei, Senior Data Scientist	Building a predictive maintenance model for 200 CNC machines to reduce unplanned downtime	ML Workbench, Data Workbench
ML Engineer Journey	Tomas Rivera, ML Engineer	Deploying an automated visual inspection system for real-time defect detection on the production line	ML Workbench, Pipeline Service
BI Lead Journey	Carlos Mendez, BI Lead	Creating the plant performance and OEE analytics platform for 4 facilities	BI Workbench, Semantic Layer
Executive Leadership Journey	Karen Singh, COO	Using AI-assisted analytics for cross-plant operational strategy and capital planning	Agentic Workbench, BI Dashboards

How the Walkthroughs Connect

These four personas work on the same data at Apex Manufacturing. Their work products feed into each other:

  Lin Wei (Data Scientist)              Tomas Rivera (ML Engineer)
  ┌──────────────────────┐              ┌──────────────────────┐
  │ Predictive maint.    │              │ Visual quality        │
  │ model (C-index 0.84) │─────────────▶│ inspection pipeline  │
  │                      │   model      │                      │
  │ Sensor feature       │   registry   │ Edge deployment      │
  │ engineering          │              │ P99 < 200ms          │
  └──────────┬───────────┘              └──────────┬───────────┘
             │ health scores                       │ defect rates
             ▼                                     ▼
  ┌──────────────────────┐              ┌──────────────────────┐
  │ Carlos (BI Lead)     │              │ Karen (COO)          │
  │                      │◀─────────────│                      │
  │ OEE dashboards,      │  dashboard   │ Cross-plant strategy,│
  │ maintenance cost      │  access      │ capital planning,    │
  │ tracking              │              │ scenario analysis    │
  └──────────────────────┘              └──────────────────────┘

Lin Wei's predictive maintenance scores feed into Carlos's equipment health dashboards. Tomas's visual inspection system provides real-time quality data that drives the Quality Rate component of OEE. Carlos's dashboards give Karen the operational visibility she needs for capital planning decisions. The semantic layer ensures all four personas use the same OEE, MTBF, and quality metric definitions.

Prerequisites

Before following these walkthroughs, ensure you have:

A running MATIH Platform instance (see Installation)
The Apex Manufacturing sample dataset loaded (available in the platform's sample data catalog)
Completed the Quickstart Tutorials for the workbenches you plan to use

Related Chapters

Data Ingestion -- Configuring Airbyte connectors, Kafka streaming, and file imports
Query Engine -- SQL federation and DuckDB for time-series analytics
Data Catalog -- Metadata management, profiling, and lineage
Pipelines -- Temporal-based orchestration for streaming and batch
ML Service -- Model training, registry, and Ray Serve deployment
AI Service -- Text-to-SQL and multi-agent chat

Executive Leadership Journey Data Scientist Journey