7.1.1 UNS_LANDING Model

Purpose

The UNS_LANDING schema represents the raw, historical landing zone for contextualised OT data published by HighByte into Snowflake.

Its role is to serve as the immutable, replayable source of truth for all downstream analytics, feature engineering, and ML workflows, while preserving maximum fidelity from the source systems.

This schema intentionally prioritises traceability, flexibility, and governance over query convenience.

Design Intent

The UNS_LANDING model is designed around the following principles:

• Minimal transformation at ingest

  Data is landed in Snowflake with only essential routing and governance metadata applied.

  No feature engineering or aggregation occurs at this stage.

• Payload-centric storage

  Each row represents a single UNS payload/event, which may contain multiple tags or signals.

  This avoids exploding high-frequency data into tag-level rows prematurely.

• Schema evolution by design

  Raw payloads are stored as semi-structured data to accommodate:

  • Changes in machine configuration

  • New signals or attributes

  • Payload versioning over time

Data Grain & Structure

• Grain:

  1 row per UNS payload/event

• Key characteristics:

  • A payload may represent:

    • A batch of high-frequency data

    • An event-based update

    • A contextualised asset snapshot

  • The payload structure is defined and governed upstream (HighByte), not by the Snowflake table schema

Routing & Context Columns

To ensure enterprise usability and lineage, each record includes fixed metadata columns that define where the data came from and what it represents:

create or replace TABLE TMA_APPOMAX_DB.UNS_LANDING.UNS_LANDING (
	EVENT_TS TIMESTAMP_NTZ(9),
	SITE_ID VARCHAR(16777216),
	AREA_ID VARCHAR(16777216),
	LINE_ID VARCHAR(16777216),
	CELL_ID VARCHAR(16777216),
	ASSET_PATH VARCHAR(16777216),
	SOURCE_SYSTEM VARCHAR(16777216),
	MACHINE_TYPE VARCHAR(16777216),
	QUALITY VARCHAR(16777216),
	PAYLOAD VARIANT
)COMMENT='HighByte UNS landing table (multi-tag payload). 1 row per payload/event.'
;

• ISA-95 routing context

  • SITE_ID

  • AREA_ID

  • LINE_ID

  • CELL_ID

  • ASSET_PATH

• Source & contract metadata

  • SOURCE_SYSTEM (e.g. HighByte, Ignition)

  • MACHINE_TYPE (payload schema/contract identifier)

These fields enable:

• Cross-site filtering

• Asset-level lineage

• Deterministic downstream joins without inspecting the payload body

Example Data: UNS_LANDING

Time Semantics

• EVENT_TS

  The original event or measurement timestamp from the OT/source system

  Used for:

  • Analytics alignment

  • ML feature windowing

  • Cross-system correlation

This distinction is critical for auditability and replay scenarios.

Data Quality Handling

• QUALITY represents payload-level quality at publish time

  • Examples: GOOD, DEGRADED, BAD, PARTIAL

• Detailed tag-level quality and status are preserved inside the PAYLOAD itself

This approach allows:

• Fast filtering on payload usability

• Full fidelity inspection when deeper analysis is required

Usage Pattern

The UNS_LANDING schema is not intended for direct consumption by:

• BI tools

• Dashboards

Instead, it serves as:

• The authoritative source of contextualised OT event data

• The primary dataset for data science exploration and experimentation

• The replayable foundation for downstream transformations

It acts as the input layer for:

• Curated analytics views (where required)

• Inference-aligned feature datasets (ML_LANDING)

• Model retraining and feature redefinition workflows

It also provides a replay source for:

• Reprocessing

• Feature definition updates

• Model retraining and validation cycles

Retention & Governance

• Data retention is long-term and configurable based on enterprise policy

• Records are treated as append-only

• No in-place updates or business logic mutations are applied

This ensures:

• Full lineage

• Reproducibility

• Compliance with audit and governance requirements

Relationship to Other Models

• Upstream:

  Receives contextualised UNS payloads from HighByte

• Downstream:

  • 7.1.2 ML_LANDING Model – inference-aligned feature event records

  • 7.1.3 PRED_LANDING Model –  via inference-aligned feature event records (ML_LANDING)

UNS_LANDING is the foundation layer upon which all higher-value data products are built.

Summary

The UNS_LANDING model provides a stable, replayable, and schema-flexible foundation for enterprise analytics and AI, preserving raw contextualized OT data without coupling it to downstream use cases.