General Mills plots supply-chain revamp in $3B cost-cutting plan

The Analysis

Three billion dollars, and somewhere inside that number is a road that gets longer before it gets shorter. General Mills has folded a supply-chain network revamp into a USD 3 billion cost-cutting program, and the executive line is that the current network was built for a lower-volume era and now needs upgrading to lift productivity and simplify distribution. I have read that sentence more times than I can count, and every time it reads to me like the line I hear on the landbridge when the volumes thin: the road was graded for heavier traffic than it now carries, and somebody has to decide which junctions stay open.

A distribution network is not a list of buildings. It is a road system: a set of nodes with lanes between them, and every lane has a length measured in days and a cost measured in dollars per case. When you redesign that system, you are not rearranging furniture. You are changing how far a case of product travels and who pays for the last stretch of that trip. That is the part a cost-cutting headline quietly skips, and it is the part that lands on the desk of anybody who ships into or out of a network like this one.

Take the headline apart first, because the money is the least interesting thing in it. A USD 3 billion cost-cutting effort is a big number, but it is a pot, not a plan. The two phrases that actually describe the work are the ones stuck at the end: built for a lower-volume era, and simplify distribution. Those two lines tell you this is not a belt-tightening exercise. This is a network that was sized for a flow it no longer has, being resized to the flow it does. Teams of people do not restructure a supply chain because they enjoy it. They do it because the fixed base has become too heavy for the traffic, and the arithmetic of a heavy road under light traffic is unforgiving.

Here is the arithmetic in one line. A distribution node costs roughly the same whether it handles a full flow or a thin one. Rent, labour, systems, the manager, the yard, the lease on the forklifts: that bill does not shrink when the volume does. So when the flow through a network falls while the number of nodes stays put, the fixed cost per case climbs, and it climbs quietly, node by node, until somebody looks at the total and decides the road has too many junctions on it for the traffic it carries. That is the sentence in the press note. It is a fixed-cost story wearing a strategy hat.

Now put the map on the table, because the second phrase is where the real work hides. Simplify distribution sounds like tidying up. What it actually means, on the ground, is fewer nodes and longer lanes. If you close a distribution centre and serve its territory from a node further away, you have just made two opposite things happen at once. The middle of the road gets shorter and cheaper: fewer handoffs, fewer touches, fewer times a pallet is picked up and set down, less inventory sitting in more places at once. And the end of the road gets longer: the customer who used to sit close to the closed node now sits at the far edge of somebody else's territory, and their last stretch of the journey got the distance that the middle stretch lost. You did not shorten the road for everyone. You moved the length from the middle to the end, and you changed who pays for it.

I have watched this exact move on the landbridge, and it is the same physics whether the node is a border crossing or a warehouse. When the traffic thins, the small crossing gets closed and everything is funnelled through the big one. On the balance sheet that is a clean win: one less facility, one less shift, one less set of overheads. On the road it is not clean at all. The trader who used to be twenty kilometres from the small crossing is now eighty kilometres from the big one, his transit day is longer, his fuel bill is higher, and he is paying for a decision that looked tidy from a spreadsheet in another city.

A closed node does not delete its traffic. It moves that traffic onto a road that is still open, and the road that stays open is the one that has to absorb the strain. I am not saying this is what will happen to every lane in this network. I am saying it is the shape of the thing, and it is the shape the headline will not show you.

So why now, and why this program? The honest answer is that the volume change already happened and the network never followed it. A network built for a higher-throughput era does not become wrong the day the volume drops. It becomes wrong slowly, which is why it survives for a few years and then shows up as a line item in a cost-cutting plan. And once a company has announced a USD 3 billion target, it has also, in practice, announced which pots the money will come from: procurement, overhead, and the physical footprint. Announce a number that big and the network becomes a candidate, because a network is one of the few things you can genuinely change in the middle of a business like this. That is the causal chain, and it is worth holding onto, because it tells you this is not a demand story and not a freight-rate story. It is a fixed-asset story.

Who feels this, and how much? The company itself takes the first hit, straight on its own margin. Then there are the suppliers who ship into these nodes: their delivery addresses, their transit promises and their appointment windows all change the moment a node closes, and they usually find out later than they would like. The logistics providers holding the nodes are next in line, where one site loses the flow and another site inherits it, along with the congestion that arrives with inherited flow. And the one that is always left out of the press release is the customer at the far end of the surviving node's territory, whose goods arrive a day slower after a decision that was sold as a cost saving. If you ship for a network like this, or into one, you are in one of those four circles, and the only question is which.

Here is the point the coverage tends to miss, and it is the one I would underline. The headline frames this as a cost story, but the operating decision underneath it is a routing decision, and a routing decision is answered one node at a time. The company is not choosing between cost and cost. It is choosing, node by node, which lane stays and which lane goes, and every one of those choices has a version that saves money on the middle mile and a version that quietly loads the final mile.

A program measured by how many nodes closed will close the cheapest node. A program measured by landed cost per case, door to door, will close the node whose removal makes the whole road cheaper end to end. Those two yardsticks can point at different buildings, and only one of them is right. If you are on the other side of this decision, shipping into or out of the network, you want to know which yardstick is in the room, because it decides whether your lane survives.

There is one more missed point, and it is about service. Simplifying distribution shrinks complexity, and complexity is expensive, so far so good. But simplification also consolidates flows onto fewer, busier lanes, and a busier lane is a lane that queues first. The node that inherits the volume inherits the pressure. So a redesign that looks like pure savings on the cost sheet can show up six months later as longer lead times, more missed appointment windows and a service penalty that nobody wrote into the business case. None of that is an argument against the redesign. It is an argument for pricing the service side of it honestly, and for not mistaking a cost number for the whole picture.

Now let me run the counterfactual, because it tells you where the plan is either right or wrong. If the volumes come back, or if the category demand turns up again, a network cut too deep runs out of road. And a closed node is not a switch you flip back on. Once you sell a site and let the lease go, re-opening it costs months and money you did not budget, if you can re-open it at all. So the whole exercise is safe only if the nodes you cut are ones whose traffic is structurally lower, not ones whose traffic is cyclically soft. Cut the structurally thin node and you have done the right thing. Cut the node that is thin only because this quarter is thin, and you have sold a junction you will want back in a year, which is the most expensive way to save money I know.

The other side of that condition matters too. A network that refuses to close anything, in the name of flexibility, pays for that flexibility forever, and eventually pays more than the occasional wrong closure would have cost. So the safe answer is not zero closures and it is not maximum closures. It is closures matched to structure, and a documented reason attached to every node that stays and every node that goes.

Let me put numbers on it, and I will say clearly that all the splits below are my assumptions and not the company's. Take the USD 3 billion target. A distribution-network program inside a consumer-goods cost effort typically carries something in the range of fifteen to twenty-five per cent of the total, so call it twenty per cent. That is about six hundred million dollars of the target sitting in the network, at least in the way I would frame it. Now spread that across the physical system. I do not know how many nodes this network runs, so I will assume thirty for the sake of the arithmetic. Six hundred million across thirty nodes is twenty million dollars of savings attributed to each node on average. Seen that way, this is not a one-time clean-up. It is a node-by-node campaign, and each of those twenty million dollar decisions is made by people who have to guess at the traffic on both sides of the move.

Or take it another way, the annualized one. Assume the program runs over two to three years, which puts it at roughly one billion to one and a half billion a year. Twenty per cent of that, from the network, is two hundred to three hundred million a year. Now assume the system carries something on the order of a thousand distinct lanes, which is not a big number for a consumer-goods footprint. Two hundred to three hundred million across a thousand lanes is two hundred to three hundred thousand dollars per lane per year.

That number is the one to sit with, because it says something very practical: consolidating a handful of the biggest lanes moves the total far more than sanding down hundreds of the small ones. It also tells you where the axe lands, and therefore which of your lanes are exposed. The big, busy lanes get reworked. The thin, awkward lanes get merged or dropped. If you ship on either kind, you are about to feel this.

So what do you do with it, and when, if you are the one shipping into a network that is being redesigned? Get your own lanes on a map this week, and mark, for each one, which node serves it and whether that node is big enough to survive a consolidation. You do not need the internal plan to do this. You only need to know, for every lane you own, whether the node behind it is a junction that traffic is flowing toward or a junction traffic is flowing away from. The ones flowing toward will get busier and will queue. The ones flowing away may close, and their lanes will get longer.

Then get the calendar. A network revamp like this does not happen overnight; it happens in a sequence, and the sequence is decided months before it is announced to suppliers. Ask your buyer or your account contact for the consolidation and closure dates for the next two quarters, in writing, and re-price the affected lanes within a couple of weeks of each date you are given. A lane that has just been re-pointed at a different node is a lane whose transit promise has probably changed, and the people who find that out late are the people who pay for it.

Where the middle mile shortens and the final mile lengthens, renegotiate the split. If you carry any part of the distribution leg, this is the moment to put both halves of the cost on the table and argue about the total, per case, door to door, not just the line-haul portion. It is entirely possible that the redesign makes the whole road cheaper and your particular leg more expensive at the same time. Both things can be true, and only one of them will be in the proposal somebody sends you.

Then hold a buffer at the nodes that inherit the flow. This is the part people forget. The surviving node is not just bigger on a diagram; it is the node everything is now funnelled through, and funnel points are where the queue lives. If your SKUs run through a node that is absorbing volume from a closed or merged site, keep a little more safety stock on the fast movers and expect longer appointment lead times while the new flow settles. A node that gains traffic spends a few months learning how to carry it.

Where the moving goes badly is when you copy the closure logic without checking your own road. It is tempting to look at a company cutting nodes and decide that smaller is automatically better, and to start shaving your own network to match. But your network is not their network. What is right for a high-volume consumer-goods footprint, with dense flows and a national DC estate, is not automatically right for a smaller or more fragmented book, and a lane that is thin for them may be a lane that is load-bearing for you. Match the principle, not the plan.

Mode is the next place it goes badly. When a lane gets longer and the middle mile stretches into rail range, somebody will suggest moving that leg onto an intermodal service, and sometimes they will be right, because a longer haul at a steady flow is exactly what rail is good at. But rail only wins when the volume on the lane is dense and steady enough to fill it, and when both the origin and the destination are close enough to a terminal that you are not just moving the trucking pain to the wrong end of the trip. Compare the two options per case, door to door, and keep whichever one holds the transit you promised. Do not compare a truck rate with a rail rate and call it a saving, because the truck rate and the rail rate are not the same journey.

The quietest place it goes badly, though, is different. A network redesign looks like an operations event, but for anyone shipping into it, it is a date-sensitive commercial event. Your transit promise, your delivery window, your penalty clauses if you miss a slot: all of them ride on a road that is being re-cut under your feet. So the real risk is not that the redesign is wrong. The real risk is that it is right, and moves underneath you, and you are the last to hear about it.

So back to whether this road can be walked. It can, and the direction it is being walked is fairly clear: fewer junctions, longer last miles, denser flows down the lanes that remain, and a fixed base that finally matches the traffic it carries. That is a sensible thing for a network to do, and it is the same move every corridor makes when the volumes thin. The road that fewer trucks use is the road that gets narrowed, and narrowing is only the wrong decision when you shut a junction you will need later or leave one open that is carrying nothing.

The people who get hurt in this are not the ones who made the wrong call about a node. They are the ones who were shipping through it, found out late, and had to explain to their own customer why a road they had been walking for years suddenly got longer. Watch which node your lane runs through, get the date it changes, and price the whole road, both ends, before it does. After that, it is, as it always is, a question of the route and not the rate.

↑ Back to the key points

— By Ömer Kaya