Unlocking collective intelligence: Why manufacturers need incentives and assurance to share data

What you’ll learn:

Smart technologies, from AI to digital twins, are fueled by data and are touted as key to bringing back competitive manufacturing to the U.S.
Firms often do not perceive a clear benefit from sharing data.
It’s time for factory leaders, engineers, and tech vendors to stop talking and start testing this new playbook: federated learning, linked to blockchain.

“Data is the new gold.”

“Data is the new oil.”

Headlines such as these made rounds in various forms over the past decades. Whole new business models, companies, even industries emerged—all centered around data, with the recent rise of generative AI and large language models further pushing this narrative.

In manufacturing, data always played an important role, from experimental data to understand the physics of machining processes, to continuous improvement and coordinating complex supply networks.

Episode 2 of (R)Evolutionizing Manufacturing: Data is everything

However, with the ubiquitous availability of sensors, storage, and compute power, manifested in the Industry 4.0 and smart manufacturing paradigm, digital transformation accelerated rapidly. Smart technologies, from AI to digital twins, are fueled by data and are touted as key to bringing back competitive manufacturing to the U.S.

This should be a fruitful endeavor. After all, today’s manufacturing operations—ranging from precision machining of aerospace parts to complex assembly lines producing trucks—produce massive amounts of data every (milli-)second.

Every sensor reading, every quality inspection, every machine operation, and every supply network interaction generates time-series, pictures, or tabular data. These data points offer localized insights, helping to optimize single processes or predict maintenance needs for machines.

The real game-changing strength of large amounts of manufacturing and supply network data manifests only when it’s collected and analyzed as a single dataset, a source of truth.

Complex insights derived from patterns across thousands of similar machine tools and operations scattered across countless factories—and even across different industries—to flag looming part failures, fine-tune supply networks worldwide, or dream up brand-new, leaner production methods.

This truly is not sci-fi anymore; it’s what cross-industry machine learning and AI enables us to do. Still, most valuable data stays locked behind every company’s firewall, kept separate from everyone else. Siloed. It’s like a giant gold field in which every firm guards a tiny, lonely seam of riches, blind to the deeper shared motherlode that could be tapped if we agreed to work together and share insights for the benefit of all.

The great data divide: Where incentives fail

There are myriad reasons that companies hesitate to share data, even when theoretically everyone stands to gain. Hard policy constraints (e.g., tight export controls), competitive considerations, technological hurdles, or resource limitations all can stand in the way of sharing.

Several of these are tough to overcome, but something is often overlooked: hard cash on the table. Manufacturers, quite sensibly, keep their digital gates locked. Cyber risks are everywhere, and inside those bytes sit trade secrets, edge-winning process knowledge, and countless hidden details that drive day-to-day operations and competitiveness.

The chance that hackers, rival firms, or even a careless leak could scoop up that information turns data sharing into a minefield. Laws around privacy, export control rules, and compliance demands that are constantly changing pile on more headaches, pitfalls, and possible fines.

On top of that, cleaning, masking, standardizing, and securely sending terabytes of data costs time and money. For many outfits, the know-how or bandwidth simply is not readily available to tackle the task.

Even if all those hurdles are cleared, the underlying problem remains: Firms often do not perceive a clear benefit from sharing their data. Talk of shared smarts, faster innovation, or a slicker industry sounds great, but to a busy production manager this shrinks fast into a distant, fuzzy benefit that hardly affects today’s cash flow and operations.

Hence, the question lingers, how will handing over hard-won data ease my bottleneck, reduce quality issues and waste, or keep my products ahead of the pack? Without an answer, the data stays tucked away or will never be collected in the first place.

Federated learning and blockchain: A trust-based ecosystem for data sharing

There are various possible pathways to address this issue. Two technologies together address three key obstacles: privacy, security, and control. The two tools are federated learning and blockchain.

At its core, Federated learning turns the classic model-training playbook on its head. Instead of analyzing pre-processed raw data via a central server, federated learning lets each node (e.g., a company, production line, or machine tool) train the same AI engine locally thus keeping their private data secure and within the boundaries of the organization.

Every entity builds a local model with its own unique records, sends only the model weights—the small update, not the raw numbers—back to the global model, which gets updated by all the contributing local models.

Because the approach shares updates, not data that might be constrained, it cuts straight through the biggest data-sharing concerns, including possible leaks, compliance, or corporate spying. Factories keep their data secure and under their own control and thus avoid the risk of data breaches or cyberattacks.

The second bit, blockchain, slips into the gap as the trust layer, adding clear proof, audit trails, and enables incentives for participation. It’s public ledger, fixed in code and visible to all parties, chronologically notes every step of the individual contribution to the global federated learning model.

It can cryptographically stamp who took part, how much data they sent, how valuable their contribution to the modeling outcome, and the precise update each node uploaded, all in a tamper-proof record with full local control.

This piece adds the much-needed ability to trace valuable contributions back to its source—if your contribution improves the overall model, you should be rewarded for that. This reward can be monetary or otherwise.

Building a smarter, more resilient manufacturing future

Manufacturing finds itself at a turning point, a pivotal one not just for the sector itself but for the nation as well. If we want AI and machine learning to genuinely boost efficiency, flag maintenance needs before they disrupt production, streamline supply networks, and spark bold new products and services, we cannot keep data compartmentalized and locked away.

While AI trained on “local” data has already made tremendous impact on various manufacturing processes, systems, and machine tools, we have not unlocked the game-changing potential that models trained on large, rich, and diverse cross-industry shared datasets offer.

The way things work now, with no (financial) reward for sharing data, we hit limits that slow the whole industry down. Enter federated learning, linked to blockchain so every step is transparent and can be rewarded (e.g., monetary or tokenized).

Used together, these technologies offer the benefits of sharing data without the most critical drawbacks. Local shop-floor data can become a common springboard for innovation while still protecting trade secrets and cybersecurity.

So, plain and simple, it’s time for factory leaders, engineers, and tech vendors to stop talking and start testing this new playbook. The next moves—crafting shared standards and forming sector consortia that run these tools—are what will matter most.

Picture a world where one pooled AI model, trained on anonymized inputs from thousands of plants, spots a failing gear before the sound of grinding metal, or where a blockchain marketplace pays users for fine-grained supply data, saving entire supply networks from surprise interruptions.

Where a small manufacturer can create value (and be compensated for it) for the whole industry by running machine tools creating rich data sets, e.g., by purposely producing scrap parts to address the typical unbalanced nature of production data while the order books are low and the machine tools underutilized to enrich the global model or enable research and education.

Methodologies such as testbed-as-a-service (TaaS) provide a medium to facilitate data sharing and learning further. Envision companies providing data to research consortiums and have their valuable contribution counted as needed cost share. The tools are there, so now we need to build the business environment and the entrepreneurial mindset to change the narrative.