What the Theory of Drift Tells Me About Industrial Gas Supply (And Why It’s Not a Drill)

If your industrial gas supplier isn't treating their own system as a source of error, you are one unchecked assumption away from a plant shutdown. That's the theory of drift in action. I've coordinated emergency deliveries for chemical plants and energy projects. I've seen the pattern. The 'theory of drift'—from high-reliability organization research—isn't abstract. It's a practical warning about how systems degrade. In my role, I've watched suppliers drift from safe, reliable processes into failure modes because they got comfortable. Here's what that means for anyone buying hydrogen, nitrogen, or oxygen at scale, with a real example of how to catch the drift before it costs you a day of production.

What the Theory of Drift Actually Describes

The theory of drift says that complex systems—like an industrial gas supply chain—don't fail because of a single catastrophic event. They fail because small adjustments accumulate. Each decision makes sense locally. Someone delays a maintenance check because throughput is high. A delivery slot shifts because a client needed an emergency fill. A standard procedure gets replaced by a 'workaround' that saves 15 minutes.

There's something satisfying about recognizing this pattern after watching it play out three times in one quarter. (Mental note: I really should write a checklist for this.) The drift isn't intentional. It's a response to pressure. And most suppliers don't see it until the error shows up as a miss—a wrong product, a delayed shipment, a quality variance.

Take this with a grain of salt, but from what I've observed across roughly 200 rush orders and emergency requests, roughly 60% of those emergencies trace back to a small drift that went unchecked.

How Drift Plays Out in Industrial Gas Delivery

When I first started coordinating emergency supply, I assumed the biggest risk was equipment failure—a compressor going down, a pipeline leak. That was wrong. The biggest risk is procedural drift. I used to think it was all hardware. Three years of chasing last-minute problems taught me it's usually process.

Here's a concrete example from Q3 2023. A client needed a specific nitrogen purity level for a continuous process. Normal turnaround for a specialty tanker is 3-5 days. They called on a Tuesday needing delivery by Thursday. The supplier had the product. The tanker was available. The usual process would have been: confirm spec → load → check documentation → dispatch. But the team was under pressure. Someone took a shortcut in the spec confirmation step because 'it's the same spec as last time.' The drift happened right there.

It wasn't the same spec. The product loaded was for a different pressure rating. The error was caught at 2 AM three hours before delivery—by a third-party inspector who noticed a mismatch on the paperwork. The client's alternative was a $40,000 production delay. We found a replacement tanker from another terminal, paid $2,100 in emergency logistics fees on top of the $14,000 base cost, and delivered by 2 PM. Close call. (I still kick myself for not having a dual-check spec system in place already.)

The drift wasn't malice. It was efficiency without a guardrail.

Why Big Industrial Gas Projects Are Especially Vulnerable

Projects like those run by Air Products—global hydrogen networks, massive chemical processing plants, low-carbon energy initiatives—operate at a scale where drift can compound. One small procedural deviation in a hydrogen pipeline spec can cascade into a compliance issue. A change in an ammonia cracking process that seemed minor three months ago becomes a safety variance during commissioning.

The conventional wisdom is that big companies have enough procedures to prevent this. My experience with 10+ emergency cases tied to project startups suggests otherwise. The very complexity that makes Air Products a leader also creates more points where drift can creep in. Large-scale projects have more handoffs, more subcontractors, more layers of spec verification. Each handoff is an opportunity for drift.

The counterintuitive fix is not to add more procedures—it's to create a system that makes drift visible.

Building Systems That Detect Drift Early

Based on what I've learned from coordinating emergency responses across multiple industrial gas suppliers (and a few that didn't make the cut after repeated failures), here's what works:

1. Mandatory 'Spec Confirmation' Step Before Every Load

This sounds obvious. But in practice, people skip it because 'it's the same as yesterday.' The drift starts there. One chemical plant I worked with implemented a policy in 2022: no load moves until a second person has physically checked the spec against the current order—not the previous order. Their emergency requests dropped by 35% in six months.

2. Track 'Near Misses' Like They Were Actual Failures

Most companies only log actual errors. The drift is in the near misses. A wrong spec caught before loading? That's a drift indicator. A paperwork mismatch found during audit? That's a drift indicator. An order that almost went to the wrong site? Drift. Track these. They're free data about where your system is degrading. In 2023, I documented 47 near misses from a single supplier's internal system—47 warnings they had chosen not to act on until one became a real miss.

3. Build a 'Drift Audit' Into Quarterly Reviews

Set a recurring calendar slot—don't just do it when there's a problem. Review the last 30 days of orders. Where did people take shortcuts? Where did the process deviate from the written procedure? Which deviations had no consequence (luck) vs. which ones were caught by a check (system)?

The difference between a reliable supplier and an unreliable one is often not the procedure itself—it's the discipline of looking for drift when there's no crisis.

The Boundary Conditions: When Drift Is Actually Innovation

Not every deviation is drift. Some small changes are legitimate process improvements. A team finds a faster way to confirm a spec that doesn't introduce risk. A delivery route change saves an hour without affecting safety. The challenge is telling the difference.

A rule of thumb I've landed on: if the deviation saves time and is reversible, test it formally. If it saves time but is irreversible (like changing a spec verification step), it's drift until proven safe. I'm not 100% sure this is universal, but it's caught more problems than it's caused in my experience.

The honest truth: even with good systems, you'll still have emergencies. Equipment fails. Demand spikes. Clients change specs at 4 PM on a Friday. The goal isn't to eliminate all errors—it's to ensure the errors that happen are unpredictable ones, not the ones you could have seen coming.

That's the value of treating the theory of drift as a practical tool, not just an academic concept. It gives you an early warning system for the failures you haven't had yet.