Nobody could say how many folding stools were still out there

The situation

The email came from a national retailer's compliance department, forwarding a customer complaint with the subject line 'urgent - product safety.' A folding step stool the company sold had collapsed under a child, and the retailer wanted to know, within days, how many units were affected and what the company planned to do about it. Sylvain, who kept his day job as a municipal planner while running the company on evenings and weekends, forwarded it to Micheline with a message that just said 'is this the third one.'

It was. Two earlier complaints had come in through the company's own customer service inbox over the previous month, both describing the same failure at a weld point near the hinge, both closed out by a junior staff member as isolated, unlucky incidents without anyone connecting them or flagging them upward to Sylvain or Micheline. Sylvain and Micheline had started the business as a side project a few years earlier, designing simple household goods and selling them through a handful of retail partners they had built relationships with one at a time, and it had grown steadily into an established operation with several million dollars in annual revenue. Micheline still worked days as an office manager for a local clinic and handled the company's books and customer inbox in the evenings. Neither of them had built a formal system for tracking safety complaints, because until that week, working out of a small office with a lean team, they had never really needed one.

The stool in question was manufactured under contract with an overseas supplier and distributed to retailers through Simran, who ran the wholesale relationship on the company's behalf and had pushed hard, in the earliest conversation about the complaints, against pulling stock without more certainty about how large the problem actually was. Simran's concern was not unreasonable from a business standpoint, and it came from someone who had spent years building the retailer relationships that were now at risk. A recall meant unsellable inventory, refunds the company would have to fund, and a retailer relationship years in the making suddenly under strain.

What Sylvain needed first was a straight answer to the retailer's question: how many units were affected, where were they sitting right now, and how fast could the company realistically act. Getting that answer turned out to be far harder than anyone on the team expected it to be, and harder still than the retailer, waiting on the other end of an increasingly impatient email thread, seemed willing to believe.

The complication

The company's sales records did not agree with each other, and the disagreement only became visible once someone actually tried to add the numbers up. The order management system showed one total number of units shipped. The manufacturer's production run records showed a different number, several hundred units apart. The retailer's own point-of-sale data, once they shared it, showed a third figure that did not reconcile cleanly with either of the first two. Nobody at the company could say with real confidence how many of the affected stools were sitting in warehouses, on store shelves, or already in customers' homes.

This mattered because a voluntary recall coordinated with a regulator has to be scoped accurately, not approximately. Understating the number of affected units risks leaving a defective product in circulation and facing regulatory criticism later for an incomplete response that missed known stock. Overstating it wastes company resources on units that were never actually a risk, and can trigger a broader public notice than the situation calls for, which carries its own reputational cost for a small company whose customers may remember the recall long after they forget the details. Getting the number right was not a formality to satisfy paperwork. It was the foundation the entire recall plan needed to stand on before a single notice went out.

The gap traced back to how the business had actually grown, not to anyone's carelessness. Early sales had been tracked in a spreadsheet Micheline maintained by hand back when the company sold through two retail partners. As the retail relationships expanded to a dozen partners and then more, some orders moved into a proper order management system and some did not, and a batch of units sold directly to smaller regional stores outside the main retail channel had never been consistently logged anywhere at all. Nobody had set out to lose track of inventory. It had simply happened the way it happens in a growing company, where the accounting kept pace with the business until, one particular week, it stopped keeping pace at the worst possible time.

Simran's instinct to resist an immediate blanket recall was, underneath the frustration everyone was feeling, a fair question the company genuinely could not yet answer: pulling every unit of an entire product line, when the actual defect might be limited to a single production run from one factory batch, was not obviously the right call either, for the company or for customers holding stools that were never actually defective. But without reliable numbers, there was no way to make a narrower, batch-specific recall defensible to the regulator or credible to a retailer who wanted certainty, not a guess dressed up as one.

What we did

1. Advised the company to notify the regulator promptly with what it knew, rather than waiting until the numbers were fully reconciled, since regulators generally respond better to early notice with acknowledged gaps in the data than to a late, polished report that arrives after more customers have been affected, and early notice preserved the company's credibility for the harder conversations still to come, including the eventual scope negotiation.
2. Brought in the company's accountant to rebuild the sales and shipment records from the ground up, cross-referencing the order management system, the manufacturer's production logs, and the retailer's point-of-sale data line by line to produce one reconciled set of numbers everyone in the file could actually work from with confidence, a task that took longer than anyone expected given how tangled the original spreadsheets were.
3. Traced the defect back to a specific production run using the manufacturer's batch records, once reliable shipment data finally existed to make the tracing possible, which let the company scope the recall to the affected run rather than the entire product line, addressing the legitimate concern Simran had raised about unsellable inventory without ignoring the underlying safety problem the three complaints had already exposed.
4. Drafted the voluntary recall notice in coordination with the regulator, describing the defect, the affected batch, and the remedy being offered to customers in the specific language and level of detail the regulator's own guidance expected before any notice could go public, and revised the draft twice after the regulator's initial feedback, since a notice that understated the risk or buried the remedy behind vague language would have invited exactly the kind of follow-up scrutiny the company was trying to avoid.
5. Set clear terms with the retailer for pulling and returning affected stock, including who would bear the cost of unsold inventory sitting on shelves, so the recall itself did not spiral into a separate commercial dispute layered on top of the underlying safety issue the company was already managing, at a moment when the retailer relationship could least absorb a second, unrelated fight.
6. Built a customer notification and refund process for units already sold to the public, using the retailer's point-of-sale data where it was available and a public notice campaign to reach the customers the company had no direct purchase record of at all, since the direct-to-store sales gap meant a meaningful share of buyers were otherwise untraceable through any list the company already held. Covering both channels meant the recall's real reach was not limited by the same record-keeping gap that had caused the delay in the first place.
7. Put a proper inventory and complaint-tracking system in place going forward, so that a future safety complaint would be logged, cross-referenced against the earlier complaints, and checked against actual reconciled unit numbers immediately, instead of surfacing as a mystery again in the middle of the next live recall, when there would be even less time to sort fact from assumption.
8. Reviewed the contract manufacturer's quality control obligations under the existing supply agreement to determine whether the weld defect pointed to a production standard the manufacturer had failed to meet, which mattered both for allocating the cost of the recall fairly between the two companies and for deciding, once the recall itself was behind them, whether the same supplier could still be trusted with future production runs.

The outcome

The recall closed out with the regulator several months after the retailer's first email, once the affected batch had been identified, the notice issued, and the returns fully processed across the company's retail channel and its direct customer base. No further injuries were reported after the recall notice went out. The regulator's file closed without further action, which was the best outcome realistically available once the defect had already reached customers before anyone caught the pattern.

The cost was real, and nobody at the company pretended otherwise afterward. The retailer wrote off a portion of unsold stock the company had agreed to cover as part of the negotiated terms, and the weeks spent reconciling records before the recall could be properly scoped meant the affected batch stayed on some store shelves longer than it should have. Sylvain was direct with the rest of the team afterward that the delay was not really caused by the defect itself but by the company's own inability to answer a basic question about its own inventory fast enough when it mattered most.

The company's relationship with the retailer survived the episode, helped by the fact that the company had notified the regulator early and did not try to minimize the problem once the reconciled numbers were clear and undeniable. Simran's early caution about a blanket recall turned out to be genuinely useful once the company could actually support a narrower, batch-specific recall with real data behind it, though everyone involved agreed privately that the conversation would have gone better a month earlier, before three separate complaints instead of one had already come in through the same overlooked inbox.

The review of the manufacturing agreement also gave the company a basis to push some of the recall's direct cost back onto the contract manufacturer whose weld process had produced the defect, rather than absorbing the entire expense alone. That recovery did not erase the damage to the company's own record-keeping, but it meant the company was not left carrying every dollar of a problem it had not caused on its own.

What you can learn from this