It’s not always easy to distinguish between trash and valuable data. Remember that Welsh engineer who, in 2013, mistakenly disposed of a laptop hard drive with the private keys for his 8,000 Bitcoins, worth a jaw-dropping $751 million? The hard drive (and the money) was never recovered. That’s valuable data disguised as trash!
It works the other way, too. In many data lakes, databases and historians, there is plant data pretending to be valuable, but it’s really trash.
Why?
Because all data without context is trash.
Data context enriches raw data
What is data context? Data context (a.k.a. data contextualization) is the process of enriching raw data with additional information that enables a fuller, more complete understanding of its meaning and relevance. That additional information includes timestamps, identity, physical location, semantics and much more. Context transforms data from noise into actionable insight.
Without context, you’re like the guy sitting on top of a pile of horse poop thinking that there’s a pony in there somewhere. Instead of horse manure, you’re sitting on a pile of tag values and you know there’s insight in there somewhere.
Without context, you can't trust, interpret, compare or automate decision-making. Not only is there no AI, there is no SCADA, ERP or business intelligence.
The eight basic types of data context
1. Time context. Accurate time is the first and likely most important. When data cannot be put into the proper chronological sequence, nothing else matters. Without accurate, synchronized time, correlations, causality analysis, event reconstruction and predictive analytics all collapse. What came first? Did the motor overload or did the temperature spike?
2. Identity context. Identity context specifies exactly which asset, device, module or sensor generated the data. Identity includes device identity, sensor identity, network identity and version/firmware because all can affect data interpretation. Is this data from the Hydraulic pump? Is it from Press 14 or Press 15? The closer the data moves to the enterprise business systems, the more context and more detailed identity context is required.
3. Data lineage. Data lineage describes the entire chain of custody: where the data originated, how it was collected, which gateways or brokers forwarded it, what transformations were applied, and whether any quality flags changed. Strong lineage prevents “mystery values,” ensures traceability for audits and is crucial for validating AI/ML pipelines.
4. Location context. Location includes both Physical location and Logical location. Physical location can matter a great deal. Not just which plant, but where data is collected in a plant. Machines, for example, near the warehouse may be exposed to different temperature variations than other machines, impacting quality. Logical location identifies the stage of production, process step and ISA-95 level. Logical location provides process relevance.
5. Semantic context. Semantics refers to what the data means: its purpose, interpretation and domain meaning. It transforms raw numbers into meaningful concepts like “coolant temperature,” “axis position” or “cycle complete.” Semantic context includes enumerations, states and standardized information models (OPC UA, ISA-95, UNS). Without semantics, data is structurally correct but functionally useless.
6. Process context. Process Context references what was happening at the moment of measurement. It identifies operating mode, recipe, batch number, lot, job, phase, speed, load…etc. It correlates the data with the operation of the system or machine, enabling root-cause analysis, optimization and traceability.
7. Measurement context. Measurement context defines the quantitative characteristics of the data: units (°C, psi), scaling factors, precision, resolution, calibration, accuracy tolerance and quality codes (good, bad, uncertain, stale). It ensures values from different sources can be compared, normalized and trusted.
8. Organizational context. Organizational context describes how data is structured, grouped, and modeled across the enterprise. This includes adherence to ISA-95 hierarchy, templates, standard tag structures, naming conventions, OPC UA object models and UNS namespaces. It enables consistency, multi-site analytics and scalable system integration.
Why context is important
Our old Industry 3.0 systems required little context. It was enough to extract an oven temperature and send it to a SCADA system for translation, formatting and display. Little data moved from the factory floor to enterprise business systems.
It’s an entirely different game now with Industry 4.0, Smart Manufacturing systems and AI Chatbots. Data is fully integrated into enterprise business systems, not just at the plant, but through the entire corporation. Data context makes or breaks these systems. Insufficient context leads to false root-cause analysis, inconsistent reporting, slow troubleshooting and misleading dashboards that erode trust. With AI Chatbots, the problem is magnified. Insufficient context leads to bad predictions and hallucinations, destroying the AI value chain.
Data Context is a multiplier. It turns data into something you can actually use. Accurate and complete data context ensures:
- Faster troubleshooting
- More accurate KPIs
- Higher quality decisions
- Faster OEE improvements
- Efficient energy usage
- Trustworthy data pipelines
- Better cybersecurity posture
- Standardized integration across sites
- Reduced engineering effort
Setting data context
Here’s what we know. Most edge gateways on the market today treat plant-floor data the same way a teenager treats laundry: grab whatever’s lying around and throw it into a pile. Sure, the data moves. Sure, it lands somewhere. But none of it arrives ready for use. Tags are cryptic. Units are undefined. Time is a suggestion. And “Press_4_Run” is transferred as an ENUM, which becomes “mystery value number 7.”
“Edge gateways” fall into one of three categories:
- Basic protocol converters (90% of the market) – Think: Anybus, Red Lion DA gateways, traditional RTA gateways, ProSoft, Moxa, HMS Flexy, Weidmüller, Pepperl+Fuchs, Banner, ifm IoT gateways. These devices simply move data, convert protocols and publish payloads. They treat data like FedEx: “We don’t care what’s inside; we just deliver it.” These devices cannot associate rich context. At best, you can add a tag name and maybe a unit ID.
- Edge compute frameworks – Think: Ignition Edge, AWS Greengrass, Azure IoT Edge, Siemens Industrial Edge, Rockwell FactoryTalk Edge. These products allow you to script context associations or embed metadata via custom payloads or models. Using one of these? Assign a patient engineer.
- Industrial data hubs – Think: HighByte Intelligence Hub, Litmus Edge (higher-end SKUs), HiveMQ Edge + Sparkplug extensions. These high-end, complex products are some of the few that support the kind of context modeling described above.
An alternative to these high-end products is the simple to use, cell-level, lightweight, embedded PLC Historian from Real Time Automation. This product combines the critical features of a Historian with the additional features found in Edge Gateways, combined with the ability to add the kind of context described above.
The RTConnect A-B PLC Historian is easy to install and quickly configurable, offering the ideal set of features for most time-series data collection applications. Tags from multiple A-B PLCs are captured, normalized and saved. User-defined models are filled and published on demand, without subscriptions, licensing constraints or reliance on third-party middleware. With configurable storage of up to 1 TB, a comprehensive suite of publishing protocols (SQL, HTTP, FTP, WebSockets, USB, MQTT and email) and direct integration with InfluxDB for visualization and analytics, the Historian is an invaluable tool for plant floor operations, maintenance and process engineers.
