84% of Manufacturers Cannot Use Their Own Data

Manufacturers generate enormous volumes of data from machines, equipment, and enterprise systems, yet 84% of companies report that data access is slowing them down, according to a panel discussion at IIoT World featuring experts from HighByte, HiveMQ, CrateDB, and Deloitte. The session, moderated by Hamish Mackenzie, examined why data usability, governance, and architecture are the real bottleneck, and how a combination of MQTT messaging, unified namespaces, and flexible databases reduced one multi-plant UNS deployment to two months across 30 plants.

Data Silos Persist Because Tools Fail, Not People

The problem with manufacturing data silos is often framed as a cultural issue: teams refuse to share. The panel challenged that assumption. Different departments own different data sources, and the difficulty of sharing stems from a lack of tools and processes, not a lack of willingness.

Three practices consistently reduce fragmentation: data consolidation through a unified repository for diverse sources, schema flexibility where the database system supports multimodal data formats without requiring every source to conform to one rigid structure, and real-time accessibility where data can be queried as it arrives without extensive preprocessing.

One manufacturing client applied these principles using a real-time analytics database to detect anomalies and reduce downtime. The existing equipment stayed in place, with the analytics layer running on top of it.

MQTT Solves the OT-to-IT Transfer Problem

Moving data from operational technology systems to enterprise IT environments requires a protocol built for constrained environments. MQTT is a lightweight, publish-subscribe messaging protocol that uses small packet sizes, making it efficient for environments with limited bandwidth.

MQTT brokers ensure reliable transfer between machines and enterprise systems and scale to thousands or millions of connected devices simultaneously. During connection interruptions, no data is lost. The protocol maintains message delivery even when the link between OT and IT is temporarily broken.

A Unified Namespace Deployed Across 30 Plants in Two Months

Without a unified data layer, accessing manufacturing data means navigating cryptic naming conventions in legacy systems and time-consuming processes to retrieve individual data points. A unified namespace (UNS) organizes all of that data logically, providing a single source of truth for the entire enterprise.

When combined with DataOps, UNS allows manufacturers to standardize and contextualize data across systems, use MQTT brokers to disseminate data efficiently, and give teams ready access to the information they need, accelerating decision-making. One deployment achieved UNS rollout across 30 manufacturing plants in two months. The same company went on to execute 10 times more data projects with the same team size.

Legacy Equipment Does Not Require Replacement

Many manufacturers operate decades-old equipment that was never designed to produce telemetry. The panel outlined a practical approach: add modern sensor layers, including cameras and IoT sensors, to existing equipment rather than replacing it. Modular solutions that abstract the integration layer make it possible to collect data from outdated systems without disrupting production.

Starting small allows teams to identify which use cases scale and which do not. The priority is a roadmap for updates, focused on where the most value exists, rather than eliminating all technical debt at once.

A real-time database can act as a data companion for legacy systems, supporting dashboards and analytics on equipment that was never designed for such applications. The original hardware stays in place; the data layer wraps around it.

The Database as AI Infrastructure

AI applications in manufacturing depend on clean, high-quality data that is accessible in real time. A database serving AI models needs to handle high-speed data ingestion, support schema flexibility for diverse data types, and store embeddings for hybrid searches and AI queries.

Predictive maintenance systems, for example, analyze anomalies and trends across sensor data. The database stores that data in a usable format, so AI systems can generate actionable insights without additional preprocessing.

IT and OT Must Lead Together

When asked who should lead digital transformation, the panel agreed on a combined approach. IT brings enterprise system expertise and cybersecurity knowledge, while OT contributes deep understanding of factory operations. A collaborative team incorporates both perspectives into data architecture decisions.

The Deloitte panelist recommended a dedicated digital transformation team to serve as the bridge, focusing on transformation goals and ensuring alignment with broader business objectives rather than letting the initiative stall between two departments with different priorities.

This article is based on a recorded panel session at IIoT World with John Harrington, Chief Product Officer at HighByte; Ravi Subramanyan, Director of Industry Solutions Manufacturing at HiveMQ; Brendan Conrad, Senior Solution Engineer at CrateDB; and Brian Zakrajsek, Smart Manufacturing Specialist Leader at Deloitte at the time of this session. Moderated by Hamish Mackenzie. AI tools were used to help summarize and organize the content. Reviewed and edited by the IIoT World editorial team.