What running an Aldi RDC through Covid taught me about resilient operations.

The first thing Covid did to operations was strip out the people who knew how things worked. Not in a planned way, with handover and replacement. Suddenly, at random, with no notice. A team leader would be off for two weeks. Then another. Then three of their direct reports. The institutional knowledge of how the operation actually ran — as opposed to how it was documented to run — was being subtracted in chunks.

What I learned, painfully, is that the operations that survived this were the ones where the process was actually written down. Not the ones where the process was "in everyone's head" or "we've always done it this way." The latter operations failed quickly, because "we've always done it this way" turned out to mean "three specific people knew how to do it, and they're all off."

In normal trading, the difference between a properly documented process and an institutionally-known one is invisible. In stress, it's everything. If your operation depends on five specific people knowing how things work, you don't have a process — you have a fragility.

The second lesson is one that I've since seen in almost every operation I've audited, including ones that have never had a crisis. Most warehouse operations are tuned tightly to their normal volume profile. They can handle the average week elegantly. They struggle with the bad week.

What this looks like in practice: the operation is staffed to handle Monday's volume, with maybe 10% headroom. When Monday is genuinely 20% higher than expected, the operation falls behind. Wednesday's volume hits while Monday is still being cleared. Thursday compounds. By Friday, you're running overtime, errors are climbing, and the team is exhausted — not because the average volume changed, but because the variance broke the tight tuning.

The Covid version was extreme — variance of 200-300% on some product lines — but the principle is universal. Operations that work well over time are tuned for variance, not for averages. They have surge capacity in critical zones. They have slack in scheduling. They have process steps that flex rather than queue. Operations that look efficient on a good week are usually fragile on a bad one.

I've written about this elsewhere — it's the most common single cause of sustained overtime I see in audits — but Covid made it brutally obvious. With sickness taking out shift supervisors and team leaders at random, the only thing that kept the late shift from inheriting chaos every day was the discipline of the handover.

Before Covid, our handover was about 15 minutes, fairly informal, mostly verbal. After the third week of someone-different-running-the-shift-every-day, we rebuilt it. Written log. Standard structure. Five minutes, every transition, no exceptions. The single change that produced the biggest reduction in cross-shift problems for the rest of the period.

What I take from that, applied to normal operations: the handover is where the previous shift's problems either get absorbed by the next shift or get left for the next shift to deal with. If it's not written down and structured, the absorption defaults to the people, and the people forget, and the problems compound.

Pre-Covid, our forecasting was solid. We knew what was coming. We staffed accordingly. The plan and the floor mostly agreed.

Then March 2020 happened, and the forecast became actively misleading. Volume patterns broke overnight. Categories that were stable for years became volatile. The forecast was still being produced and still being used to plan staffing, but the underlying assumptions it was built on had become wrong without anyone updating it.

The lesson — which I've since applied to every operation I've worked with — is that a forecast is only useful when the variance between forecast and actual is regularly checked, and the check actually influences planning. When the forecast becomes a number that's produced, circulated, and not interrogated, it goes from being a tool to being a hazard. The operation runs on assumptions that nobody has validated.

This is part of why the forecast-vs-actual loop is one of the highest-impact interventions in PainKiller audits. Most operations are running forecasts that haven't been pressure-tested against reality for months. The Covid version of this was extreme; the normal-trading version is just slower-moving and more invisible.

This is the lesson that took the longest to fully land, and the one I'd argue is the most generalisable to normal operations.

During the worst stretches of Covid, we kept that RDC running because the team kept showing up. People came back from being ill to cover sickness. Team leaders worked double shifts to keep stores supplied. People did jobs they hadn't been trained for, learned them on the job, and made it work. Nobody had to ask them. They just did it.

What I learned from that — and what I now look for in every audit — is that the relationship between an operations team and an operation is something most directors don't actively manage, and it shows. The teams that show up when things are hard are the teams that have been treated as people, not as labour, in normal times. The teams that turn over fast and disengage when there's pressure are the teams that have been treated transactionally.

The operational implication is that retention, engagement, and operational performance are not separate problems. They're the same problem viewed through different lenses. The operations directors who run resilient sites are the ones who understand this and act on it long before there's a crisis.