The Re-Ice Decision: The Cold Chain's Most Expensive Guess

Every delayed dry ice shipment produces the same moment. A notification lands: flight cancelled, connection missed, truck rerouted, weather hold at the hub. And somewhere, an operations coordinator inherits a question the pharmaceutical cold chain has never been able to answer well. Does this shipment need more dry ice before it reaches its receiving dock? If so, where along the route, and how much?

It's one of the most consequential routine decisions in cold chain logistics, and it gets made nearly blind. Nobody knows how much dry ice is left in the box. The only way to find out is to open it, and opening it spends the two things the shipment is already running short of: cold and time.

This post is about that decision. Not the delay itself, which is ordinary and mostly unavoidable, but the call the delay triggers: who makes it, what they actually know when they make it, and why so much product rides on what amounts to an educated guess.

How the call gets made today

Picture the sequence. A shipment of clinical material leaves a depot Monday night on a lane qualified for 48 hours door to door. Wednesday morning, the notification arrives: the connecting flight cancelled, the box is sitting at a hub, and the new delivery estimate is Friday afternoon. The transit just grew by two days. Someone now owns the question of whether the dry ice will last.

That someone is usually an ops coordinator at the shipper, or at one of the 3PLs and freight forwarders handling the move. Here's what they have in front of them: the pack-out record (40 pounds of dry ice, sealed Monday at 9 PM), the container's spec sheet, the elapsed hours, and the new ETA. Here's what they do with those inputs: arithmetic on paper numbers, followed by gut feel. Anyone who has sat in that seat knows the paper numbers are the smaller part of the answer.

What they don't have is anything about what actually happened to this specific box. Whether it sat in the sun on a tarmac for two hours. Whether the container is on its first trip or its ninth. Whether the pack-out was done well or done fast. The temperature logger inside, if there is one, reports that everything is fine, because the inside of a dry ice shipper holds somewhere between -100°F and -116°F for as long as any solid CO₂ remains. That reading is both true and useless. It confirms the ice hasn't run out yet. It says nothing about how close it is.

So the coordinator estimates, decides, and either starts making calls to arrange replenishment at a station or depot along the route, or lets the shipment ride. Then they wait to find out whether they were right. Usually the answer comes from the receiving dock. Sometimes it comes from a deviation report.

And "re-ice or not" is only the first of three questions. If the answer is yes, the coordinator has to pick where, from the short list of points along the remaining route that have dry ice on hand and staff cleared to handle pharmaceutical freight, and that list is shorter than most people outside the industry would expect. Then how much, which depends directly on how empty the box already is, the exact thing nobody knows. Guess low and the shipment needs rescuing twice. The where and the how much inherit all the blindness of the whether.

The three numbers the decision actually needs

Strip the re-ice decision down to its logic and it needs exactly three inputs. The trouble is not that the logic is hard. The trouble is that two of the three inputs don't exist.

First: remaining mass. How much dry ice is in the box right now, in pounds. This is the load-bearing number, and today it's unknown. It gets estimated from the pack-out record and elapsed time, which is like estimating your fuel level from when you last filled the tank, on a trip where somebody else has been driving.

Second: hours to a trained receiving dock. Not hours to delivery, hours to a person who will move the payload into a freezer. This one is knowable, more or less. Recovery ETAs slip, and dock hours matter, but a competent ops team can put a defensible number on it.

Third: the container's real burn rate. The FAA's 2024 sublimation study (DOT/FAA/TC-24/24) measured average sublimation rates between 0.53 and 0.71 percent per hour across the containers it tested, which worked out to roughly 0.26 pounds per hour for a mid-size shipper packed with about 40 pounds. Those are averages of fresh containers under test conditions. The same study found rates climb as containers are reused and insulation degrades. The box on its ninth trip burns faster than its spec sheet, and nothing on the outside of the box tells you that.

One knowable input and two guesses. And the two guesses multiply: an estimate of remaining mass built on an assumed burn rate is a guess stacked on a guess. That's the actual information content behind a decision that determines whether a shipment of clinical product survives the week.

Four ways to get it wrong

With inputs that soft, the failure modes are baked in. There are four of them, and every cold chain operation of any size has lived all four.

The first and third failures are the obvious ones: intervene too late, or don't intervene at all, and the product warms. They end the same way, with a payload at ambient and a root cause analysis that concludes, more or less, that the estimate was wrong. Which everyone knew going in, because it was an estimate.

The second failure is quieter and probably more common. A nervous team re-ices a shipment that had two days of margin left. Nothing is lost, so it never shows up in a report, but the shipment got pulled out of the network, opened, handled, repacked, and delayed, all of it unnecessary. Every touch of a pharmaceutical shipment is a handling risk, and the delay the intervention adds can itself push the shipment into trouble it didn't have before.

The fourth failure is the strangest one: opening the box just to look. It feels responsible. It's actually the worst of both worlds. The seal breaks, warm air floods in, a burst of sublimation follows, and the container's insulation performance is disturbed. The clock the coordinator was worried about now runs faster, in exchange for one eyeball reading of how much ice appears to be left. In measurement terms, the act of observing degrades the thing observed.

The ice is the cheap part

Here's the part of the economics that gets missed. Dry ice is a commodity. Pellets and blocks are among the cheapest items in the entire shipment, cheaper than the container, vastly cheaper than what's inside it. If re-icing were just a matter of buying more CO₂, nobody would agonize over the decision. You'd top off every delayed box and move on.

But the ice was never the cost. The cost is everything wrapped around it. Finding a facility along the route that stocks dry ice and has staff trained to handle a pharmaceutical shipment. Pulling the box out of the carrier's network flow, which is where shipments go to get lost. The labor to open, replenish, repack, reseal, and redocument. The delay the whole detour adds to a shipment that was already late. And above all of it, the decision overhead: the calls, the escalations, the person who has to put their name on a judgment made without data. Multiply that across every delay on every dry ice lane, and the overhead dwarfs the pellets.

Which reframes the problem. The win available here isn't cheaper ice, and it isn't even faster re-icing. It's fewer wrong calls. Every unnecessary intervention avoided, every necessary one triggered early enough to be routine, is where the money actually is.

When the guess becomes arithmetic

Now add the one number the decision has always been missing: remaining dry ice mass, measured in real time, from inside the shipment. Everything about the re-ice call changes shape.

Notice what happens to the burn rate in that equation. It stops being a spec sheet number and becomes an observed one: this container, this trip, this ambient exposure, this state of insulation wear. The reuse degradation the FAA measured doesn't have to be assumed anymore, because it's visible in the slope of the mass curve. A box burning hot announces itself hours before it becomes a problem.

Timing changes too. A mass-based alert doesn't fire when the temperature breaks, which is when the damage starts. It fires when the trajectory goes wrong: this shipment, at its current burn rate, runs dry roughly twenty hours before its dock appointment. That alert arrives while intervention is still a phone call. Someone picks a re-ice point along the remaining route, with hours of notice, and the fix happens inside the normal flow of the network instead of as an emergency extraction. And the delayed shipments that are actually fine, which is many of them, get left sealed and moving, which is exactly where they should be.

There's a quieter benefit sitting behind the operational one: the record. Today, the file on a delayed shipment shows an estimate, a judgment call, and an outcome. When the outcome is bad, the investigation ends at "the estimate was wrong," which satisfies nobody and changes nothing, because the next delay gets the same estimate. With measured mass, the file shows a number, a threshold, and the action the number triggered. Quality teams get a decision they can defend. Ops teams get one they no longer have to defend alone.

This is the reason CryoTrak measures the mass instead of the temperature. The delay was never really the problem. Delays are a permanent feature of moving boxes through airports and weather. The problem is that a delay converts a routine shipment into a blind bet, and the person holding the bet has no way to see their own cards. Give them the one number the whole decision hangs on, and the cold chain's most expensive guess stops being a guess. It becomes division.