Insights
What High-Stakes Manufacturing Teaches Us About Data Quality
May 21, 2026
Why the Industries Where Failure Isn’t an Option Have the Most to Teach us About Getting Data Right
“You Can’t Afford to Make One Bad Jet Engine.” – Jon Steffey, Tolmar Pharmaceuticals
Jon Steffey has spent his career in industries where mistakes can be catastrophic. Aerospace. Medical devices. Pharmaceuticals. The throughline connecting them isn’t immediately obvious until you think about what happens when something goes wrong.
“If we’re manufacturing a coffee mug and it goes bad, that’s okay, as long as you don’t make lots and lots of bad ones,” Steffey, Senior Director of Enterprise Software and Analytics at Tolmar Pharmaceuticals, explained during a recent conversation at FabCon 2025. “But you can’t make one bad jet engine or one bad bioreactor. The medicine that goes into our body, you can’t have that not work correctly.”
That single observation reframes everything about how we should think about data quality.
The Cost of Quality vs. The Cost of Failure
In most industries, quality is a spectrum. Good enough is often good enough. A slightly off-colour product might ship. A minor software bug might get patched in the next release. High-stakes manufacturing doesn’t work that way.
When a jet engine fails, people die. When a bioreactor malfunctions, a cancer patient doesn’t get their treatment. When a pharmaceutical batch is contaminated, the consequences ripple across thousands of lives. The cost of failure isn’t a customer service headache or a product recall: it’s existential.
This reality has forced industries like aerospace, medical devices, and pharmaceuticals to develop rigorous approaches to process control. And at the foundation of every effective process control system? Data.
“Data ends up being the foundation to have those effective process controls,” Steffey noted. “That’s what I’ve learned over the years.”
Process Controls Are Data Problems
Think about what process control actually requires. You need to know, with certainty, what happened at every step of manufacturing. You need to trace materials back to their sources. You need to verify that equipment was calibrated correctly, that temperatures stayed within range, that human operators followed protocols precisely. All of this is data.
The batch record for a pharmaceutical product isn’t just paperwork. It’s a comprehensive data trail that proves the product was manufactured correctly. Every sensor reading, every operator signature, every quality check becomes a data point that either confirms compliance or raises a flag.
When Steffey talks about “the right process controls around the manufacturing, and even everything adjacent to the manufacturing,” he’s describing a data ecosystem that extends beyond the factory floor to encompass the entire supply chain. Raw material sourcing, transportation conditions, storage environments, and distribution tracking.
In regulated industries, this level of rigor is legally required, but the underlying principle has relevance well beyond compliance-driven sectors.
What High-Stakes Industries Know That Others Don’t
Companies operating in aerospace, medical devices, and pharmaceuticals have learned lessons about data quality that every organization would benefit from understanding.
Lesson 1: Data Quality Is a Design Decision, Not an Afterthought
In high-stakes manufacturing, you don’t collect data and then figure out if it’s good enough. You design systems that produce quality data from the start. Sensors are calibrated before installation; data entry interfaces are built to prevent errors, and validation rules are embedded directly in the collection process itself.
Too many organizations treat data quality as a cleanup problem. They collect whatever data is convenient, then spend enormous effort trying to cleanse, deduplicate, and reconcile it downstream. This approach is expensive, error-prone, and ultimately inadequate for critical decisions.
Lesson 2: Traceability Isn’t Bureaucracy, It’s Insurance
Every data point in a regulated manufacturing environment has a lineage. You can trace it back to its source, verify who recorded it, confirm when it was captured, and demonstrate that it hasn’t been tampered with.
This might sound like regulatory overhead, but it’s actually a gift.
When something goes wrong, traceability lets you understand what happened. It lets you isolate problems, identify root causes, and implement targeted fixes. Without traceability, you’re left guessing. Or worse, you’re implementing broad, expensive changes because you can’t pinpoint where the failure occurred.
Lesson 3: The Entire Chain Matters
A pharmaceutical company doesn’t just worry about data quality in its own facilities. It worries about data quality at every supplier, every transportation partner, every distribution center. A contaminated raw material or a temperature excursion during shipping can compromise the final product just as surely as a manufacturing error.
This supply chain perspective on data quality is increasingly relevant across industries. Your data is only as reliable as its weakest source. If you’re making decisions based on data that flows through systems you don’t control, you need to understand the quality characteristics of that data and account for its limitations.
Lesson 4: Governance Isn’t Optional
Steffey was candid about a lesson learned during Tolmar’s data modernization journey: the importance of getting governance right early.
“I think sometimes companies might lean too much towards that one particular killer use case and skip maybe some of the fundamentals,” he reflected. “We struggled with that a little bit.”
In regulated industries, governance isn’t something you add later. It’s built into the foundation, with clear data ownership, defined access controls, established change management processes, and maintained audit trails.
This isn’t simply because regulators demand it, though they do. It’s because without governance, you can’t trust your data, and in high-stakes environments, data you can’t trust is worse than no data at all.
The Uncomfortable Truth About Your Data
Here’s the question every organization should ask: Could our data withstand the scrutiny that pharmaceutical data faces?
Not the regulatory scrutiny, but the operational scrutiny. Could you trace a decision back to the data that informed it? Could you verify that the data was accurate at the time? Could you demonstrate that the right people had access to the right information? For most organizations, the honest answer is no.
This doesn’t mean every company needs pharmaceutical-grade data infrastructure. The coffee mug manufacturer really can tolerate a different quality threshold than the jet engine manufacturer. But it does mean that many organizations are making important decisions based on data they haven’t examined critically.
Applying High-Stakes Thinking to Your Data Strategy
You don’t need to be in a regulated industry to benefit from high-stakes thinking about data quality.
Identify Your “Jet Engine” Decisions
Not every decision requires perfect data. But some do. Which decisions in your organization would be catastrophic if based on bad data? Pricing strategy? Customer segmentation? Fraud detection? Identify these high-stakes areas and apply more rigorous data quality standards there.
Design for Quality at the Source
Stop treating data quality as a downstream problem. Invest in systems that capture clean data from the start. Build validation into collection processes. Make it harder to enter bad data than good data.
Build Traceability Into Your Data Infrastructure
Know where your data comes from. Maintain lineage information. Create audit trails. When something goes wrong, you’ll be grateful you can trace the problem to its source.
Extend Your View to the Full Data Supply Chain
Your data doesn’t exist in isolation. It comes from external sources, flows through various systems, and gets transformed along the way. Understand this full chain and identify the points where quality might degrade.
Treat Governance as Infrastructure, Not Overhead
Governance isn’t something that slows you down. It’s something that lets you move fast with confidence. Invest in it early, before you need it urgently.
The Foundation That Enables Everything Else
Steffey’s career arc, from aerospace to medical devices to pharmaceuticals, illustrates a truth about data in complex organizations. The industries that can’t afford to get it wrong have developed the most sophisticated approaches to getting it right.
Their methods might seem excessive for organizations with lower stakes. But as AI becomes more prevalent, as decisions become more automated, as the pace of business accelerates, the margin for error is shrinking everywhere.
You can’t make one bad jet engine. You can’t make one bad cancer drug. And increasingly, you can’t afford to make important decisions based on data you haven’t verified.
High-stakes industries figured this out because they had no choice, but the lessons they’ve learned apply to any organization that wants to make decisions based on data it can actually trust.
Ready to Apply High-Stakes Thinking to Your Data Strategy?
Whether you’re navigating regulatory requirements or simply want to make decisions you can trust, the principles that govern aerospace and pharmaceutical data quality can transform how your organization operates.
Adastra helps companies design data infrastructure that’s built for quality from the source, with the governance, traceability, and process controls that turn data into a genuine competitive advantage.
…
This article is based on a conversation with Jon Steffey, Senior Director of Enterprise Software and Analytics at Tolmar Pharmaceuticals, recorded at FabCon 2026. Tolmar is an international pharmaceutical company recognized for advanced, long-lasting injectable drug delivery capabilities, producing products across urology, oncology, and endocrinology.
