What we learned from a botched closed-loop trial

In the spring of 2024 we ran a closed-loop trial with a mid-sized chemical processing company in Wisconsin. The trial was supposed to last 12 months. It lasted five. The post-mortem we did afterward changed how we evaluate every closed-loop proposal that's come across our desk since.

The setup

The customer manufactured a granular chemical product and shipped it to roughly 18 downstream customers across the Midwest. Their initial pitch to us was that all 18 customers were stable, all of them were within our routed footprint, and all of them would participate in a returns program. The fleet sizing math looked clean. We agreed.

What broke

Three things broke, in this order. First, two of the 18 customers stopped buying within the first three months. Their boxes were stuck — neither at the destination nor coming back. We absorbed that as expected churn.

Second, three of the 18 customers decided to sublease their boxes to their downstream customers, who weren't in our return program at all. Boxes started traveling to places we couldn't recover them from. The customer's procurement team had no visibility into this practice.

Third, and most importantly: by month four the loss rate of the fleet was running at 6–8% per cycle, against our model's assumption of 1.5%. The economics no longer worked. We ended the program in month five at a small mutual loss.

The post-mortem

The honest takeaway was that we hadn't asked enough hard questions during the proposal phase. Specifically:

1. We didn't ask the customer how they'd actually enforce the no-sublease rule with their downstream customers. They didn't have an enforcement mechanism. They didn't know they needed one.
2. We didn't model a realistic leakage rate above the customer's stated number. The customer's number was optimistic. Ours should have been more conservative.
3. We didn't price in early-termination protection. The contract ended early, and we ate the unrecovered fleet cost. We now build in a 90-day notice clause as standard.

What changed afterward

Every closed-loop proposal since has had to clear a checklist that includes enforcement plan, realistic leakage modeling, and early-termination protection. Two proposals since have failed the checklist — meaning we declined the trial. Both, in retrospect, would have failed for the same reasons the Wisconsin program did.

Failure has a way of clarifying the questions you should have been asking. We pay for those lessons one way or another. Better to pay once, learn, and not pay again.