Why context is everything: Avoiding data swamps in the cloud

In industrial operations, getting data to the cloud is critical to gain insights for increased productivity. Getting the data is simpler; the challenge today is making that data useful once it gets there.

Too often, companies race to implement cloud connectivity, collect massive volumes of OT data, and then find themselves stuck. The dashboards don’t make sense. The data is incomplete. Engineers don’t trust it. And worst of all, nobody uses it.

What happened?

They created a data swamp.

“If you don’t organize your data… you’re going to create a data swamp.”
– Arlen Nipper, Co-inventor of MQTT and CTO of Cirrus Link

What is a data swamp?

A data swamp is a collection of raw, unstructured data that lacks context. Think of it as a data lake gone bad — where instead of useful insights, you have floating variables with no names, no units, and no clear origin. You may have thousands of tags flowing into your cloud platform, but without metadata, structure, or source context, it’s impossible to make sense of any of it.

Arlen Nipper describes it this way:

“If you don’t organize your data, if you don’t have that machine definition and you’re just sending process variables into a data lake, you’re ultimately going to create a data swamp. Invariably, nobody knows where the data came from. You can’t contextualize it.”

Worse yet, organizations often realize this after the data has been ingested, after they’ve already invested in storage, connectivity, and cloud infrastructure. Without context, the cloud becomes just another expensive repository. 

Why context must start at the edge

Context can’t be added after the fact. It needs to begin at the source, on the plant floor, where raw signals are first captured.

That means:

  • Naming your assets in a consistent, logical way
  • Modeling your machines (pumps, valves, motors, etc.) so the data is understandable downstream
  • Tagging and scaling data properly, with units, limits, and descriptions
  • Publishing structured, contextualized payloads — not just raw values

Technologies like MQTT and Sparkplug make this possible. They enable data to be modeled and published from edge devices in a way that’s immediately usable — by humans, analytics platforms, and AI systems alike.

“We’ve gone from data falling on the floor in legacy SCADA systems, to publishing structured, usable data from the edge into a Unified Namespace,” said Nipper.

The cost of ignoring context

Many cloud initiatives fall into a pattern:

  1. Data is collected and sent to the cloud.
  2. AI and ML solutions require custom code, lookup tables and constant management i — because the data lacks structure.
  3. The data doesn’t deliver the targeted insight.
  4. The project stalls or dies.

Meanwhile, companies continue to pay for cloud storage, even though their teams aren’t using the data.

“We’ll ask customers, ‘What are you doing with the data in your cloud platform?’ And they say, ‘Nothing yet… we haven’t figured that out.’” – Arlen Nipper

That’s not just a missed opportunity, it’s a drain on resources. And it’s avoidable.

How a Unified Namespace solves the problem

The Unified Namespace (UNS) is the architectural pattern that brings structure to industrial data. It provides a single, real-time, logically organized source of truth for your enterprise, where every tag, asset, and event is published in a structured hierarchy.

With a UNS, your data is:

  • Contextualized at the source
  • Standardized across sites and systems
  • Immediately usable by downstream consumers (analytics, AI, BI, etc.)
  • Scalable for new use cases without rework

And most importantly — it avoids the data swamp trap.

Final thoughts

Context isn’t optional. If your industrial data strategy skips this step, you’ll end up with a system nobody trusts and a cloud bill nobody wants.

As Arlen said during the IIoT World Manufacturing Day panel:

“Try to figure out what problem you’re trying to solve, before you take any data to the cloud. Without that, many projects are just exercises in connectivity, not transformation.”

Want to learn more? Watch the full IIoT World panel featuring experts from Cirrus Link, Snowflake, and HighByte as they explore the real strategies behind cloud infrastructure, UNS architecture, and AI-ready OT data.