The Factory Behind the Factory

On the purchasing sheet, the switch costs less than the box.

Nobody bought the fan because of this switch. It does not feature in the photograph. The advertising team has never held a meeting about the exciting future of switching. It is small, familiar and expected to work.

Now imagine that the assembly plant has everything it needs except that part. Motors are ready. Guards are stacked. Workers are on shift. A truck is booked for Friday. The supplier has sent a message explaining that the next batch of switches will be late.

There are plenty of other switches in the world. A search produces pages of them. Some cost less. To someone looking only at the purchasing sheet, the solution appears obvious: buy another switch.

The engineer has questions. Will it fit the mount? Will the contacts handle the load? Will it last? Does the replacement preserve the safety of the approved product? Can the plant install it without changing the wiring or the housing? How long will it take to establish the answers? A switch that looks right in a photograph has not yet passed these tests.

The production manager has another question. How many can actually arrive before Friday?

The cheapest part has become the most powerful part in the room.

This does not mean that low price is a secret measure of importance. It means that an input’s price and the cost of losing access to it answer different questions. A small item can stop a much more valuable output when the item is required and a suitable replacement is not ready. The purchasing sheet is a record of spending. It is a poor map of consequences unless someone adds what happens when each item is missing.

A tier is a place in someone else’s picture

Businesses often describe suppliers in tiers. A first-tier supplier sells directly to the business doing the counting. A second-tier supplier supplies that supplier. The count continues backward.

If Lena’s shop buys from a distributor, the distributor is her direct supplier. The assembler may be one step farther away. For the assembler, the motor maker may be a direct supplier. The same company can occupy different tiers depending on whose network we are drawing.

A tier is not a rank of skill or worth. It is a position relative to a particular buyer.

Even this simple language can mislead if it makes us picture neat rows. The motor maker may buy wire from a processor that also supplies the cord maker. Two rival assemblers may use the same switch plant. The company on the invoice may be a distributor with several sources rather than the factory that actually makes the part. One firm may sell directly to the assembler and also supply one of its suppliers.

Try drawing the arrangement. The rows begin to cross.

The phrase supply chain remains useful, but the drawing starts to look less like a chain than a road map on which several roads meet at the same bridge. Counting the roads will not tell you whether you can get home when the bridge closes.

Suppose Lena divides her fan order between two brands. Their packaging differs. Their sales teams compete. One offers a better warranty; the other is cheaper. This could be genuine diversification. It might also leave her exposed to a single plant that makes a part used by both. Two names on the invoice establish two commercial relationships. They do not establish two independent sources of production.

The same problem can occur at different scales. Two factories may use one electricity substation. Two ports may depend on the same road. Two suppliers in different countries may buy a specialized material from one producer. A distributor can hold stock from several manufacturers, or several distributors can hold stock from the same manufacturer. The name at the top of the form does not settle the issue.

This is why I find the question How many suppliers do you have? less revealing than it first sounds. Ask instead which failures would affect several of them together. Ask which part could be replaced, by whom, and how soon. Ask whether the alternative has actually made the thing you need.

The answers may still be reassuring. We are not trying to discover a hidden catastrophe under every invoice. We are trying to replace a comforting count with knowledge that can guide a decision.

The shock that arrives without the storm

An earthquake can damage a factory. It can also damage the output of a factory far from the earthquake.

That second effect is easy to imagine and harder to measure. A distant business may be losing sales for other reasons. Its region may have its own problems. To show that trouble traveled through supply relationships, researchers need information about who buys from whom and a comparison that helps separate connected effects from unrelated changes.

Vasco Carvalho, Makoto Nirei, Yukiko Saito and Alireza Tahbaz-Salehi studied the Great East Japan Earthquake of 2011 using detailed firm relationships. Their research found disruptions spreading to both customers and suppliers of affected firms, including indirect connections. The damaged places did not contain the full economic effect of the disaster. Their separate estimate of its overall effect on Japan used a production-network model; it was not a direct count of every lost sale.¹

That finding gives us a better picture of a supply problem. Consider our fan assembler again. If the switch factory cannot deliver, assembly slows. That affects firms that buy the completed fan. But the assembler may also stop ordering guards because there is little reason to accumulate more guards while finished fans cannot leave. The guard maker is then hurt by trouble at the switch maker, although the two need never have traded with each other.

One side is short of a part. Another side is short of an order.

The distinction matters to the people trying to respond. More transport will not help a supplier whose customer no longer wants the shipment this week. A loan might keep it operating through a temporary pause. A clearer account of when the buyer expects to resume can prevent a needless shutdown or a pile of unwanted goods. The same initial disruption can require different responses at different points.

Jean-Noël Barrot and Julien Sauvagnat reached a related finding in research on natural disasters and firms’ supply relationships. Damage to suppliers affected their customers, with stronger effects where the inputs were more specific to the relationship. They also found effects spilling toward other suppliers. Their evidence supports a practical concern: some inputs are much harder to replace than their purchase price suggests.²

Neither study gives Lena a universal percentage to apply to her fan order. The scale of loss depends on the event, the network, the available substitutes and the time allowed. But both challenge a habit of treating each firm as though its risks end at its own fence.

A factory can pass a building inspection and still be unable to work because something it needs is missing. A supplier can have healthy machines and capable workers while its customers’ problems empty its order book. The physical site and the commercial network are two parts of the same livelihood.

This does not make connection a mistake. The same relationships that carry a shock normally carry useful materials, knowledge and income. To eliminate every path along which trouble could travel, we would also eliminate paths along which help can arrive. The task is to understand the exposure well enough to preserve the benefits while creating room to respond.

The replacement that exists in a catalogue

A useful backup must survive contact with the actual job.

Suppose the assembler identifies a second switch maker. The price is acceptable. A sample arrives in a padded envelope. The first temptation is to mark the problem solved.

There are several different achievements hidden inside that word backup. One company might be able to design a suitable part. Another can already make it, but its current output is promised to other customers. A third has idle equipment but lacks the workers to run an extra shift. A fourth can supply the part in six weeks, which may be excellent for the next season and useless for Friday’s shipment.

These firms are not equivalent alternatives.

The assembler also needs to know whether a good sample can become a dependable batch. A single carefully made item can show that a design is possible. It does not establish that thousands will be consistent. Drawings, test procedures, material specifications, equipment and trained people have to support repeatable production.

This is the slow work behind the simple request to find another source. A part must fit more than a physical hole. It must fit the product’s performance, the production process, the safety requirements and the calendar.

NIST’s Manufacturing Extension Partnership advises manufacturers to map suppliers and assess risk, and its supplier-scouting work connects buyers’ technical needs with production capabilities. The emphasis on capabilities is important. A company name is useful only when there is a reason to believe the company can do the required work.³

In our example, suppose the second switch can be made suitable by changing the mount in the plastic housing. That could be a sound redesign. It may require a tool change, another test and new instructions for assembly. Perhaps it also affects how an old fan can be repaired. A decision that appears to concern one tiny item has become a decision about the whole product.

This is one place where design and resilience meet. A designer can choose a widely available part, allow more than one approved component, or make a wear-prone item easy to replace. Each choice has consequences for cost, size, performance and maintenance. Sometimes a special part is worth those consequences. A buyer should simply know what has been exchanged for what.

Standard parts can widen the set of possible suppliers. They do not guarantee independence if nearly all of the world’s suitable output comes from a few plants. Nor is a unique part automatically reckless. It may deliver a real benefit that no common component can provide. The question is how the business plans to sustain that benefit when the original source has a problem.

Suppose the assembler wants the second maker to remain ready but places no orders for a year. The backup supplier has rent to pay. It may use the machine for another product, retrain its workers, or retire the tool. Readiness decays if it is treated as a favor that somebody else must keep doing for free.

An occasional production run can test the relationship and help keep skills current. A capacity agreement can make the duty more explicit. Either costs money. The first supplier may lose some volume, and the second may need a higher price because its batches are smaller. Those costs belong in an honest comparison with the disruption the backup is meant to prevent.

We will return to the price of preparedness. For now, the essential step is to stop calling a telephone number a production plan.

When fear becomes an order

Physical goods travel toward the customer. Requests travel back toward production. Those requests carry information, but they also carry the buyer’s worries, bargaining tactics and hopes.

Suppose Lena normally orders one hundred fans at a time. A warm week improves sales. She orders one hundred and twenty. Her distributor sees stronger demand and, worried about being short, increases its own order by more. The assembler reads the distributor’s request and adds a further cushion. Several decisions that each look sensible can turn a modest change at the shop into a much larger change upstream.

Then the weather cools.

This is one form of the bullwhip problem. In their 1997 paper, Hau Lee, V. Padmanabhan and Seungjin Whang analyzed why orders can vary more than final sales, with distortion increasing as requests move upstream. They considered demand updating, rationing incentives, order batching and price changes. The point was not that people in business are uniquely foolish. The ordering rules themselves can produce a misleading signal.⁴

Take rationing. If Lena believes she will receive only half of what she requests, she may ask for two hundred fans when she wants one hundred. Other buyers may reason the same way. The supplier sees a surge in orders and may conclude that final demand has grown sharply. If supply improves, those inflated requests can turn into cancellations. An apparent shortage of goods becomes an excess of orders, then an excess of stock.

The arithmetic in this example is intentionally simple. In practice, firms use different allocation rules and face different penalties for canceling. The lesson remains: an order can be a request for consumption, a request for insurance against shortage, or an attempt to influence allocation. The number alone does not reveal which.

Batching introduces another source of confusion. Suppose a shop sells ten fans each week but orders forty once a month to save on handling or freight. Its customers are steady. Its supplier sees a month with one large order and three empty weeks. If the supplier mistakes the ordering rhythm for the rhythm of use, it will be planning for the wrong world.

Price promotions can do something similar. A business may buy extra while the price is low and then pause purchases while it uses the stock. The deal changes the timing of its orders even if final use changes little. A busy loading dock during a promotion is not automatically evidence of a lasting rise in people’s need for the product.

The cure is not to insult forecasting and declare that nobody can know the future. Firms can share actual sales and stock information, explain unusual orders, reduce needless batching, or design allocation rules that make exaggeration less rewarding. They can also learn to ask why a number changed before treating it as a command to expand everything.

Better information does not remove real uncertainty. Next summer’s weather remains next summer’s weather. What it can remove is some of the extra uncertainty produced when each firm reacts to another firm’s reaction.

There is a human version of this problem inside organizations. A sales team hears that stock may be short and urges buyers to order now. Buyers order early because of the warning. The resulting rush is reported as proof that demand has become extraordinary. Everybody has evidence. Some of that evidence is the result of everybody’s previous message.

A manager who rewards only rapid growth in orders may encourage this confusion. A manager who penalizes every empty shelf but ignores leftover stock may encourage excessive buying. A manager who values a low purchase price but never counts the cash tied up by bulk discounts may buy an apparent bargain that the warehouse has to live with for months.

The numbers need interpretation by people whose incentives allow them to tell the truth. A system that punishes modest, accurate forecasts and applauds heroic last-minute recovery should not be surprised when it receives a regular supply of emergencies.

What a map should show

A complete map of every dependency behind a household fan would be a formidable object. The wire maker uses machinery. The machinery has parts. Its workers use roads, water and public services. Somewhere in the widening picture, we would have drawn much of an economy.

No small retailer can inspect all of that. Demanding perfect knowledge before taking any action is a reliable way to ensure that no action is taken.

A useful map begins with a purpose. Which interruptions would prevent an important service? Which inputs are hard to replace? Where could several apparent alternatives fail together? How quickly would the problem become serious? These questions help decide where detail is worth the effort.

NIST’s practical guidance on supply-chain mapping recommends gathering supplier and part information, identifying risks and choosing priorities. The map is meant to support decisions, not serve as a decorative wall covering.⁵

For our fan assembler, the switch may deserve attention even if annual spending on it is small. A common packing material with many suitable local suppliers may require less investigation. The motor could merit a deeper conversation about tooling, production sites and second sources. The point is to follow possible consequences rather than allow the largest invoice to monopolize concern.

Time belongs on the map too. An input that can be replaced in a day creates a different exposure from one that takes months to approve. A supplier with two weeks of usable stock offers a different window for action from one already shipping its last batch. A long delay is not always a disaster if it is known early enough. A small delay can be serious when every other commitment has been fitted tightly around it.

This suggests a pair of questions worth asking together. How long could we continue if this source stopped? How long would it take to restore supply or use another source? If the first period is shorter than the second, there is a gap to manage. The estimate will be imperfect. Even a rough, honest answer is more helpful than a green box labeled reliable.

The map also needs a date. Firms move production, change sub-suppliers and acquire one another. A relationship that was independent when checked can become a shared dependency later. A buyer who updates only the supplier’s bank details may know where to send money more accurately than where its goods come from.

Knowing more can create duties. If the map reveals a single flood-exposed plant making a critical component, a buyer cannot claim ignorance as easily. That may be uncomfortable. It is also the purpose of the exercise. Information becomes resilience only when someone has the authority and resources to change the risk.

The supplier has reasons too

Why would a supplier refuse to reveal its own suppliers?

There can be several reasons, and they do not all involve hiding wrongdoing. A business may fear that a powerful buyer will bypass it. Its list may include relationships built through years of effort. It may have agreed to protect confidential information. It may genuinely lack a reliable picture beyond its direct purchases. A questionnaire from a major customer can be easy to send and expensive for a small firm to answer.

This does not mean that every refusal should be accepted. It means that a request for transparency has a commercial setting. If the buyer promises to use information only to manage risk and then uses it to undermine the supplier’s position, the next request will meet less cooperation. Trust is not restored by adding more mandatory boxes to the form.

A proportionate approach may ask first about critical production sites, shared sources and recovery plans. Sensitive details can sometimes be shared under appropriate limits or checked by an independent party. The buyer can explain what decision the information will support. Suppliers can be told when their answers are outdated and how to correct them. None of this requires pretending that every business relationship is friendly. It requires making truthful cooperation more workable than concealment.

There is also a difference between asking whether the chain will serve the buyer and asking what the chain does to people along the way.

A shipment might be dependable for Lena while the workers making it endure unsafe conditions. A supplier may meet deadlines by pushing the cost onto people with little power to refuse. The buyer’s operational risk map would call that success if it looked only at arrival times.

The OECD’s guidance on responsible business conduct takes a wider view of due diligence. It asks businesses to identify and address adverse effects connected with their operations and business relationships, including effects on people and the environment. This is an ongoing process, not a certificate that makes every subsequent decision harmless.⁶

The two kinds of inquiry can reinforce each other. Workers who can report a dangerous fault may also prevent a shutdown. More stable terms may help a supplier retain skilled people. Yet we should not defend a person’s safety only by showing that it improves the buyer’s returns. There are conditions we ought to reject even when the goods arrive on time.

A decent supply chain must pass more than the customer’s stopwatch.

This is where the request to know a supplier’s supplier becomes a question about power. How far backward does the buyer’s curiosity extend? Does it stop at price and delivery, or reach the circumstances that make both possible? When a problem is found, does the buyer help correct it, require a credible change, or simply move the order and leave the people behind to absorb the loss?

There is no single response for every case. Serious harm can require a business relationship to end. Other problems may be better addressed through a plan that supports improvement and checks whether it occurs. The easy error is to treat a clean-looking supplier list as proof that the underlying work has become clean.

A smaller, better question

At the end of this inquiry, Lena has learned less than everything and more than she knew before.

She has discovered that her two fan brands share a component source. She knows which models would be affected by a delay there. One assembler has tested an alternative; the other has only requested a quotation. The distributor can identify finished stock already in the region. The next order can be adjusted while there is still time.

This is a modest result. No great speech has been delivered about the end of globalization. No executive has promised to make the entire network immune to surprise. The buyer has a few facts that change what she can do.

She may decide the shared exposure is acceptable for an inexpensive seasonal product. She may reduce the size of the next commitment, carry a different model, or pay to keep another source active. If the item were a critical repair part, she might reach a different conclusion. A risk does not become intolerable merely because we have named it. Naming it makes a reasoned choice possible.

The supplier, in turn, can ask for steadier forecasts and payment terms that allow it to prepare. The conversation has become more demanding on both sides. Good. Dependability has work to do on both sides.

The smallest part no longer disappears because it is cheap. The second supplier no longer counts as independent because its letterhead is different. The order no longer speaks with the unquestioned voice of final demand. Each has acquired the context needed to judge it.

Somewhere behind the finished fan, a worker checks a batch of switches. Somewhere else, a planner decides whether to book the next run. The customer will not see their names. The movement of air depends on their work all the same.

Inventory of a morning

A hand closes a carton.
A hand signs for the wire.
Someone keeps the old tool working.
Someone asks before the change.
The names stay off the label.
The work goes on inside.

Sources and notes