Hacker News Reader

Comment by dd36 | Hacker News Reader

I thought Toyota invented not doing that.

After the 2011 earthquake they came to the conclusion that for some hard to source parts you have to keep a big stockpile.

Basically if a part can only be produced by one or two parties there is too much risk of that source going away and disrupt everything else. This also applies all the way down the chain so if you got part A that can be made by 20 contractors but if all of those contractors depend on the same source then that part is also on the list of "stockpile this part enough to get over most disruptions"

Should this be called the left-pad moment of just-in-time?

Not really, Toyota handled the shortage the best and was the last to cut production. The supply disruptions simply lasted too long to be possible to compensate for.

Not at all; rather they were only in a position to intelligently stockpile due to the deep research into safely reducing inventory elsewhere.

Yes, they more or less invented the concept of just in time delivery, but they also suffered supply chain issues after the Fukushima earth quake. After the latter incident they made the risk calculation that inventory of some parts was worth the expense.

Reminds me years ago my company used 16X2 LCD modules for a couple products. You could buy them from two dozen manufacturers. Then one day a Typhoon went over the only factory in the world that made the driver IC's.

We tried to order 100 and were told the lead time was 36 weeks. We got through it with existing stock, scrounging displays from dead boards. Buying different displays and replacing the backlights. Small orders of 10 here, 15 there. And pushing a few customers out a month or two.

And then they invented that maybe they should keep poorly sourceable parts at hand. And chips are those. Unlike nuts, bolts, metal and plastics. Which you can easily find replacements for.

Plastics have actually experienced a pretty significant shortage this year, with resins and other components being hard to source[1]. Polaris, for example, would build their entire vehicle except the seats, then build and attach them once the plastic resin for the foam became available[2].

1) https://www.ntotank.com/blog/resin-material-market-shortages... 2) https://www.wsj.com/articles/supply-chain-bottlenecks-drive-...

Yep there is an entire risk profile stage that many people leave out. And they apparently correctly estimated the risk in this case. At least for now. Probably wasn't reasonable to estimate such a long supply chain disruption though (that it's still happening now).

It is different to keep large inventory of everything because you are inefficient and can't function without it and different to decide to stockpile specific parts because of perceived potential risks while being efficient with everything else.

One big advantage of low inventories is you do not have to fix or trash lots of parts if they were made out of spec. In the case of chips the odds are they are all good. It is just the cost of ownership but not the cost of refitting.

>One big advantage of low inventories is you do not have to fix or trash lots of parts if they were made out of spec

Instead you either stop/slow production or shove them in your products and hope for the best.

You know a function called Supply Chain Management exists, right?

It should go without saying that there are nuances in implementation. What I'm describing here is a fundamental tradeoff of JIT systems. If you get the wrong thing delivered it throws a bigger wrench into things because you don't have the buffer. Can you make this rare enough that the amortized cost is low enough to make JIT overall cheaper? Of course, that's why everyone does JIT.

> If you get the wrong thing delivered it throws a bigger wrench into things because you don't have the buffer.

JIT, in Lean, does not mean no buffer, it means as little of a buffer as you can get away with. If you have issues with delivery like this on a regular basis, then you'd increase the buffer size (at least temporarily) and also take your suppliers to task for sending the wrong thing over and over.

The buffer size should be increased if any upstream supply issues exist that regularly cause a shortage. Ideally, you should address those issues themselves, but if you have and they can't (or won't) be fixed then you increase your buffer to accommodate reality. However, the shortage is itself a signal. Too high an inventory permits supply issues to persist without being addressed for a long time because you never get the signal about the issues with them (the downstream production slowdowns).

> JIT, in Lean, does not mean no buffer, it means as little of a buffer as you can get away with.

Yeah, spot on. One of my college professors used to compare it to a river with rocks in it. If you want to safely sail on the river, you can either a) keep the water level high enough or b) remove the rocks. In a production system inventories/buffers are the water level and variance is the rocks. The philosophy of JIT is to remove as much variance from your system as possible so you can lower your buffers. If you have identified areas of high variance you're forced to keep buffers until you've removed enough variance to lower your buffers.

Or, you've lowered your water to barely cover the rocks so one little mishap and you've capsized.

> JIT, in Lean, does not mean no buffer, it means as little of a buffer as you can get away with.

Eh... I would argue that JIT means making that buffer someone else's problem.

I was doing EDI at a logistics firm that contracted with Seagate who provided HDDs to Hitachi for their SANs around 2006. Hitachi was doing JIT for their manufacturing, Seagate however was just speculating Hitachi's demand and literally stockpiled HDDs in this firms warehouses geolocated next to Hitachi's factories.

We would pickup stock from Seagate and ship it to these warehouses where they would remain Seagate's property until Hitachi requested it, then we would simply transfer ownership to Hitachi.

Interestingly, we used rail shipping as a buffer to reduce warehouse size by sending freight on slow/cheap/indirect routes.

In this case you have two buffers. Seagate/you have a buffer for the outflow based on expected consumption rate. Hitachi almost certainly has some buffer of their own. This is not unusual with physical goods where you have the transportation time and cost to consider (which you/your employer took advantage of).

If Hitachi couldn't consume your delivered HDDs as fast as they were delivered and anticipated any kind of delay/disruption could ever happen, they'd have some buffer of their own.

In this case Hitachi didn't have a buffered supply. The warehouses were located literally across the street. The products were palletized at Seagate's factory in quantities matching Hitachi's product lines then delivered to the warehouse in a cadence closely matching Hitachi's consumption rate. So when Hitachi placed an order, a pallet was pulled and the ownership of the serial numbers on that pallet were transferred to Hitachi as it was delivered.

The logistics firm was the buffer allowing Seagate's product rate and Hitachi's consumption to be asymmetric in nature.

The last bit works, as long as the slow transportation is closely controlled. I tried it once, in the end warehouse space was cheaper.

EDIT: What you describe sounds more like VMI, vendor managed inventory, than JIT. Both require half way reliable forecasts and collaborative planning so to worl properly. Have to agree so that both solutions tend to push inventory risk to suppliers. Done correctly, overall inventory does decrease so.

JIT and VMI go hand-in-hand, they aren't mutually exclusive. Implementing JIT is to impose VMI on your suppliers.

The interesting thing was that Seagate avoided managing inventory by outsourcing to the logistics firm. The stock was technically Seagate's until it was ordered by Hitachi but the logistics company took immediate possession as pallets rolled out of the factory.

> The last bit works, as long as the slow transportation is closely controlled.

It didn't need to be controlled, just scheduled. You knew you need x units by d. The factory output n per week, so you could stagger shipments by way of different lines.

All of the inventory was tracked by serial numbers and it was interesting to watch it move because supply was often delivered to the warehouse out of order or shipments weeks apart arrived simultaneously.

The way I used was VMI to avoid the limited flexibility of JIT, bit yes both concepts tackle the same problem. What you describe sounds like a lot of fun to run on a daily basis, would have loved to do that!

I wonder if Toyota was a little slower at starting to expedite, for having had a large buffer stock already in hand.

They changed after 2011 earthquake when several factories couldn’t continue production because parts were not coming from factories impacted by earthquake.

JIT is probably the most mis-understood supply chain concept.

How?

Because people equal JIT with close to zero buffer inventory. JIT is basically a concept of defining the correct and minimal buffer. If that's close zero, cool. If it's more than zero, cool as well. It also puts a lot of emphasis on lot sizes and lead-times. No one in the right mind would implement JIT for stuff involving sea freight without buffer stock close to production.

There are pre-conditions for products to be set up as JIT. As you mention, short lead times, but also things like high quality, high availability, and others as well.

Toyota and a lot of the concepts that come out of Toyota are ideals to strive for. It doesn’t mean everything is like that, which is hard to understand from just reading the lean literature.

Also can depend on how high value the part is. I remember one time our purchasing agent bought 10,000 units of something. Took 5 years to use it all up. Normally they were $0.25 but some company canceled a huge order leaving the distributor with about 50,000 units overstock. We got them for $0.05 each.

Wall St in general and Jack Welch in particular embraced the “avoid owning anything” at any cost model to juice the books.

I don't think that's exactly true and actually may be a common misconception - see https://www.youtube.com/watch?v=b1JlYZQG3lI for a much better explanation than I could provide.

Chips are small, don't age fast. Make sense as an exception to the rule.

Chips also go out of production suddenly (with only a couple years warning - which isn't anywhere near enough time to replace them all). thus any major manufacture has stockpiles of chips that are no longer in production that need to last until production of the widget switches to new chips and have enough left over as spare parts for existing widgets.

> Chips also go out of production suddenly (with only a couple years warning...)

No they don't, automotive semiconductors suppliers have an "obligation" to manufacture the component for at least 15 years, which makes managing the production output planning, spare parts etc. much easier. It's not like walking into your supermarket and finding out that your favorite brand of chocolate is no longer available. There are minor exceptions, and sudden changes in the demand might affect the immediate availability, but at the very least the part is almost guaranteed to be produced for 15 years with defined notice policies. Microcontrollers don't have a pin-compatible drop in replacement when they get discounted, but many different ICs do, like power supplies, transistors etc., so discounting them is not a big deal.

e.g.: https://www.nxp.com/products/product-information/nxp-product...

> Participating products are available for a minimum of 10 years from product launch (15 years from product launch for many products developed for the automotive, telecom and medical segments), and are supported by standard end-of-life notification policies.

Some of them get that, and car manufactures have learned the hard way how important that is and so demanding it more. However there are still a lot of parts where they can't get the 15 year supply in anything that meets the other requirements.

That changed after 2011 earthquake

This is a misunderstanding of lean manufacturing. Lean is not about eliminating every stockpile, it's about thinking through your supply needs and only stockpiling what is necessary.

Were manufacturers stockpiling what was unnecessary before lean manufacturing? It's an absurd notion that anybody would stockpile things without consideration of inventory costs.

Yes, they were.

Think of each stage of the system, not just the components you bring in but also the partially assembled components you produce along the way as well as the finished product. US auto manufactures (in particular) had an operating method where they kept inventories high at all stages. This wasn't entirely deliberate. They weren't saying, "We need 5000 car doors just sitting here." They were, instead, saying, "We can't stop making doors just because everything else down the line is stopped due to <event> so keep churning them out and pile them up." The tail end inventory of "complete" vehicles were sitting there due, often, to quality issues (misaligned assemblies, missing parts, whatever the reason may be).

So inventory piles up everywhere along the chain, which also worked because there was a large turnaround period when retooling and equipment downtime (not always planned). Because Nobody Ever Gets Credit for Problems that Never Happened [0] there was an underinvestment in maintenance and improvement efforts. High inventory across the line papers over this issue. Lean discourages high inventory in order to make these issues apparent so that they can get the attention that they deserve. Also, rework is viewed as waste so quality issues should be addressed when they're discovered, not by assembling hundreds or thousands of vehicles incorrectly and then fixing them, fix the assembly line issues causing that misassembly.

[0] https://web.mit.edu/nelsonr/www/Repenning=Sterman_CMR_su01_....

Sort of, it would come down to suppliers would historically only sell in large quantities to maximize their production efficiency, so the per unit production cost would be minimized by running a long production run. But this would potentially cause scrap if the final product did not sell well to expectations or if there was a design error, and then also increased working capital as the inventory levels would spike really high with infrequent supply from the vendor instead of being more constant with more frequent replenishment.

And then multiply across everyone in a supply chain for a single product having to deal with waiting to receive giant parts orders from their vendors before they could start their own giant order to supply their customers.

They did stockpile everything due to not having a plan or consider costs. That also had the strange side effect of constantly running out of parts. In some way, stuff worked by accidentally having the right stuff on hand at the right time in sufficient quantities. TPS is all about having a plan for this stuff.

Power 101: Preach one thing, do the opposite

It is smart to project to the world that just in time manufacturing is effective and then stockpile parts. This will get you ahead of competition if there is a problem like we experience now.

Reply on news.ycombinator.com