Manufacturing environments suffer from inconsistent data models, fragmented systems, and unreliable pipelines. KŌJŌ Stack establishes consistent data structure, deterministic pipelines, and reusable data models-at the edge, before data reaches any downstream system. Manufacturing systems become composable when data is structured at the first mile.
Standardized at the Source
Shop floors combine machines from multiple decades and vendors, each producing data in different formats with different addressing. No canonical model exists-every downstream consumer builds its own interpretation of what the data means.
Production data is scattered across PLCs, SCADA systems, historians, and manual records. Each system holds a partial view. Reconstructing a complete picture of production state requires manual effort and domain-specific knowledge.
Data delivery between OT and enterprise systems is best-effort. Records are lost during network outages. Latency is variable and unpredictable. Downstream systems cannot distinguish between missing data and a lack of events.
Without structured, prepared data at the first mile, downstream systems inherit every inconsistency, gap, and limitation of the raw source data.
Without a canonical data model at the source, every analytics platform, historian, or enterprise application that consumes production data must build its own translation logic. Adding a new consumer means a new integration project. The complexity scales linearly with every system added.
When each machine and line produces data in vendor-specific formats with different addressing, constructing a plant-wide view requires manual effort and domain-specific knowledge. No single system holds a complete, consistent picture of production state.
Without reusable data models and pipeline configurations, onboarding new equipment or expanding to a new facility requires starting from scratch-custom adapters, custom schemas, custom integrations. Growth amplifies architectural debt instead of leveraging existing infrastructure.
Every machine-regardless of age, vendor, or protocol-publishes to the same ISA-95 compliant namespace. Enterprise → Site → Area → Line → Cell addressing provides a single canonical data model that all downstream systems consume from.
Data is normalized, structured, and quality-annotated at the edge. Downstream systems-analytics platforms, data lakes, enterprise applications-receive clean, structured data. No per-consumer translation logic. No protocol-specific adapters in analytics code.
Event-driven pipelines execute with bounded latency and predictable ordering. CEL expressions compute derived metrics at the edge. RBE filtering reduces data volume by 90%+ while preserving every meaningful state transition. Behavior is consistent across lines and shifts.
New lines and equipment adopt existing namespace models and pipeline configurations. Adding a machine or an entire line follows the same pattern-no custom development, no re-architecture. The data plane scales with the operation, not against it.
Discrete manufacturing shop floors combine machines from multiple decades and vendors: CNC equipment communicating over Siemens S7, packaging lines on Modbus RTU, robotic cells over EtherNet/IP, and inspection stations on OPC UA. Each system uses different register addressing, different data types, and different timing models. Without normalization at the point of ingestion, downstream systems must maintain per-machine translation logic-creating a fragile web of custom adapters that breaks with every firmware update or equipment change. The ISA-95 Unified Namespace resolves this by establishing a single addressing model that all machines publish to, regardless of protocol. New equipment adopts the existing namespace model immediately. This structural consistency is only achievable when data is standardized at the first mile, at the edge.
One canonical data model across all machines and lines
Clean, structured data eliminates per-consumer translation
Existing namespace models and pipelines are reusable
“The unified namespace gave us a single data model across 40 machines from 6 vendors. New equipment adopts the existing structure immediately-no custom development required.”
Owning the first mile ensures discrete manufacturing data is consistent, contextualized, and usable across the enterprise.
