How to quake proof a supply chain

After a long hiatus, we return with a post. This one on Toyota’s plan to quake proof its supply chain over the next five years. (Baltimore Sun, 9/6/11).  They have three ideas:

a) standardize parts across suppliers

b) hold more inventory of critical parts

c) each region should become independent in its parts procurement.

All three will indeed make the supply chain more earthquake proof. The first allows the company to switch production to non-affected suppliers. The second allows them to draw down the reserve of parts. This does rely on the transportation network recovering relatively quickly. And hopefully the inventory is stored in a quake proof facility – imagine you produce a six month supply of microchips only to have the roof cave in and crush them all (or they melt in the subsequent fire). The third will help too (create independent regions) but it may be overkill. If a part is manufactured in a low-quake risk zone, and it is a specialized part that should most cost effectively be produced in a single facility, then why insist that it be made in 5-6 plants? This initiative may have more to do with managing exchange rate risk.

So, these are surely worthwhile initiatives and they apply beyond automobile production. The real challenge will be the implementation details – how will the company decide between part A that costs 5% more but is standard and can be produced in a number of facilities and part B that costs less but is specialized and made by only one supplier in an earthquake prone area? There could be some interesting research done on this challenge.

Supply chain lessons from Japan

The devastating earthquake and tsunami in Japan have again raised the issue of supply chain robustness to disruption risk, and in particular, are they too fragile? FT.com (3/15/2011) asserts that  “Strategies that look rational for individual manufacturing companies… can create big macro-level vulnerabilities…”

The reality is that it is too costly to source every component from multiple locations throughout the world just to hedge natural disaster risks. But that doesn’t mean that companies should turn their back to the problem. The best companies follow a few intuitive steps to make their supply chains more robust. I’ll offer two.

First, map your supply chain. If you know your Tier 1, Tier 2 and Tier 3 suppliers, then you won’t have to spend one week figuring out whether you will run out of a part. Most companies know their Tier 1 supply chain, but do they know the other tiers? Do they keep track of changes to the supply chain? This information is crucial because the company that is first to work the phone to find alternative  supplies is most likely to be able to secure those supplies.  This information also gives you information that you can use to make downstream adjustments to your production. For example, should you eliminate an overtime shift or not? Should you redirect scare parts from one plant to another? Those are difficult decisions to make and are made much more complicated if you don’t even know if you have a problem – why shut down a plant for a potential part shortage that may not materialize?

Second, before disaster strikes, map out vulnerabilities. Some components can be sourced in many locations. Some components have several months of buffer inventory. You don’t need to worry about those. But if the amount of buffer inventory is limited and it is sourced in a few locations, especially a few locations that happen to be close to each other, then you need to consider finding alternative sources or alternative parts. Maybe the conclusion is that the company needs to bear the risk – there are no effective alternatives. But maybe the conclusion is that a substitution to a less risky part is actually feasible.  Finding this substitute is less costly before the disaster. There have been reports of companies that are scrambling to qualify additional suppliers, but that could have been done before disaster struck.

Finally, one risk that will hit many companies, even if they don’t have a shortage of parts, is the risk of exchange rate fluctuations – the Yen has just hit a post WWII high against the US dollar.

McDonald’s medicine

Americans want instant gratification – that is true for fast food as much as it is for healthcare. Consequently, the traditional model of general practitioner in which you make appointments and then (a week later – if you are lucky) see a doctor is getting increasingly challenged. Patients have found the McDonald’s of healthcare. It is called the ER. You go there when you want and they will get you what you want. One stop shopping for all your healthcare needs.
In the following article, two ER doctors describe their view of the problem – and make us think if we have to invent the way we deliver care: http://www.time.com/time/nation/article/0,8599,2044392,00.html

Why China – cheap capital?

In 2008 Evergreen Solar opened up a solar panel factory in Devens,  Massachusetts, but they just announced that they will layoff their 800 workers and move production to China (NY Times 1/14/11). Why? Well, of course, it is too expensive to manufacturer in the U.S. And low cost production is critical – the price of solar panels has fallen from $3.39 per watt in 2008 to $1.90 per watt now. Evergreen Solar has reduced its cost to $2.00 per watt, but Chinese manufacturers are producing at $1.00 per watt.

But labor is NOT the reason for the high cost of production in the U.S. – labor is a small portion of the cost to make solar panels. Nor does it seem a lack of technical skills. Instead, the issue is the cost of capital – a solar panel plant can cost $400 million and Chinese manufacturers have access to low cost bank loans.

It it is likely that there will be more movement to China for reasons other than the cost of labor.

How to bump passengers from flights?

Overbooking a flight enables airlines to ensure that they fly with as much capacity as possible – given that 8-10% of passengers do not show up for a flight, you want to make sure that there are not too many empty seats on the plane.  But if you overbook, you must be ready to find yourself in a situation in which you have more paid passengers than seats. Since sitting in the aisle is probably against FAA regulation, that means the airline might need to bump one or more passengers off of the flight. How should this be done?

The standard method is to offer a fixed benefit – say a $400 flight voucher – and hope that enough people take the deal. If that doesn’t work, the airline sweetens the deal. As a last resort they have to involuntarily bump a passenger (which costs them more).

According this WSJ 1/14/11 article, Delta has another idea – when passengers check-in, ask them how much they are willing to accept to be bumped from a flight (a $ amount) and choose the lowest bidding passengers if necessary.  They have apparently be trying it out for one month.

This is an interesting idea.  There are two obvious advantages: (i) it reduces the time and effort to solicit passengers off of a flight when you discover that you need to bump some people and (ii) it is hard to imagine that they would end up paying more to bump passengers. I suspect that passengers are willing to accept less cash to be bumped when they check in than when they are about to walk onto a plane – the cost of getting bumped is more tangible when you are watching the plane pull away from the gate. Furthermore, I think some passengers will bid less than the $200 or $400 that the airline would normally pay.  (There are many ways to run this auction, but it seems that passengers receive their bids, rather than the highest clearing bid or the bid of the lowest non-bumped passenger.)

However, as an airline, do you want to ask every passenger to report how much they want to be paid if you fail to serve them? Imagine booking a vacation at a Four Seasons Resort and receiving an email like this: “We are excited that you will be staying with us and we look forward to catering to your every needs so that you will have a relaxing an memorable visit with us. On occasion, we are unable to accommodate all guests with reservations. If that were to occur, could you please let us know what the minimum cash you would accept to be booked into a partner hotel nearby to our facility?” It doesn’t scream customer service. But “customer service” may not be the first thought when people think of airlines these days. Air travel is a commodity and efficiency matters. I wonder if this experiment – like charging passengers for bags – will stick.

 

 

Taxi capacity in NYC – the 4 O’Clock Blip

Apparently it is hard to find a taxi between 4 and 5pm in Manhattan. Why? Because that is the time when many taxis switch drivers. This turns out to be a fascinating capacity management issue.

The first surprise is that a substantial number of taxis are used around the clock in two 12-hour shifts. I guess NYC really does never sleep.

So, given that you have two 12-hour shifts, and clearly one driver can do only one of those shifts, you need to decide a time and place to switch drivers. The place turns out to be, in all likelihood, a garage in Queens. (The use to do it in Manhattan, but real estate became too expensive, so they moved to Queens.) The time, you would think, would be chosen at a relatively light period. Or, it would be chosen so that earning potential of the two shifts are equated in some sense (accounting for the fact, I would assume, that the day shift is more pleasant than the night shift).  Take those factors into account and apparently 4 pm is a good time to swap drivers – one person gets the morning rush hour, one the evening rush hour. The result, according to NY Times 1/11/11 is a 4 O’Clock Blip of empty taxis:

What is really interesting about this is that the spike is so pronounced. If 20% of the taxis are taken out of service 4-5pm, then you would think that there would be a noticeable drop in competition. Wouldn’t more taxis want to make the switch at 3pm or 6pm to take advantage of this? In other words, why isn’t the blip smoother? Given that it is spiky, it must be that it is too costly to switch at the other times relative to the 4-5 pm slot.

One explanation has to do with coordination. If some cab drivers make the switch at 3, others at 4, others at 5 and yet others at 6, a cab driver needs to know when he will make his switch – when to drive the taxi back if he is doing the day shift and when to show up at the garage if he is doing the night shift. Keeping track of those times might lead to the inevitable late cabby, which will lead to idle taxis. Having everyone show up at 4-5 is a simple rule that is easy to remember – you show up at 4-5. When do you show up? 4-5.  But this seems like it could be fixable with technology. A computer keeps track of the schedule and sends reminders to cabbys as to when they will show up. Or, it could be fixed with some consistent scheduling – for a given month a given cabby will have the same drop off/pick up time but the times are staggered to smooth the spike.

An alternative explanation could be due to demand. Maybe demand drops significantly at 4-5, so being a taxi at that time is not as advantageous as the lower supply would make it seem. This strikes me as unlikely, but what do I know about taxi demand in Manhattan between 4 and 5pm?

The Bloomberg administration is looking into this issue, being a data-driven type of administration. To their credit, they don’t plan to do anything until they identify the root causes. And even then, they say they are hesitant to tell businesses how to do their business – maybe the taxis know something that we don’t know. “Fixing” this issue (if it should be fixed) will not change the world much, but it is a fun capacity management exercise.

 

 

What Disney does to reduce wait times

Understanding (and reducing) waiting lines (Queues) is central to Operations Management. Waiting in line is often one of the most memorable experiences when visiting theme parks such as Disneyland. The long wait times limit the number of rides a visitor can experience per day. Disney research shows that the average visitor only can enjoy nine rides per day. From a “value-add” perspective, this is a rather poor performance.  A ride might take five minutes. So we are looking at 45 minutes of value add (ride) time relative to an eight hour day. What can be done to improve this? Improvement actions in this environment can be categorized into three groups: (a) Increase capacity: in some cases, it is possible to increase capacities at the attraction. In the Pirates of the Caribbean attraction, Disney is able to increase the launch frequency of boats as demand increases. (b) Make the waiting more pleasant: Disney dispatches greeters and entertainers to reduce the perceived waiting times when it is not possible to reduce the actual waiting time. Simple video games are now added to Space Mountain as a way to turn unproductive (and not amusing) wait time into something fun. (c) Divert demand to less crowded areas: if you can’t increase capacity, try to reduce demand. There are 40 rides / attractions in a typical park. But demand tends to focus on some high visibility attractions as those are perceived as being the most fun. When wait times in those areas grow long, Disney tries to create some “buzz” in other areas of the park, trying to divert demand and thus balance capacity utilization.

For a more detailed story, see: http://www.nytimes.com/2010/12/28/business/media/28disney.html?scp=1&sq=disney%20lines&st=cse

Robots or people?

Maybe the biggest challenge for e-commerce retailers is dealing with the huge surge in sales in the fourth quarter.  How can you build enough capacity cheaply enough to satisfy the rapid growth in demand during October, November and December, only to have most of that demand disappear by January? The traditional approach is to hire lots of seasonal workers. The trick with this is to be able to train them quickly enough for them to be productive in time for when they are actually needed. The Wall Street Journal reports that one company, Kiva Systems, has a different idea – instead of hiring workers, install robots (Dec 19, 2010).  To see these robots in action, check out the video (click here).

You might assume that these robots would “walk” around a warehouse picking products, putting them into a basket and bringing them to a place to be packaged. That is what humans do. Instead, these robots move shelves of inventory around. (See the photo – the robot is the orange contraption at the bottom of the shelf.) One advantage of this system is that you don’t need permanent aisles between the inventory – the shelves can be packed in tightly with the computer controlling the sequence (so that the one pink doll you need isn’t buried deep within a sea of shelves).

The next thing you may notice is that these robots are not particularly fast. It is not like the robots move product through the warehouse at twice the speed a human can walk. However, assuming these things are reliable (e.g., treads don’t need replacing every couple of days) they don’t need to take breaks, and they are instantly trained. One downside of this system is that the robot must move the entire shelf and not everything on the shelf may be needed at one time. Humans pushing a cart around a warehouse only put into their cart what is needed at the time.

But the point of the article is how to deal with the holiday surge in demand. While a robot might replace a human, it doesn’t eliminate the problem – the company simply needs a lot more capacity in the 4th quarter. If it buys these robots, then they are likely to be idle most of the rest of the year. Seasonal employees are just that – seasonal – that is, they go into the deal with the expectation that their work will be temporary.

The article ends with an idea for making the robots more cost effective for the retailer – Kiva Systems will rent the robots to the company for just the peak demand period. But I don’t see why this solves the problem – now Kiva Systems is sitting on expensive and idle capacity for most of the year (even in the Southern Hemisphere, Christmas falls in December).  Rental systems work well when potential customers need the product at different times. Given that the 4th quarter is the same for all retailers, I am not seeing this as an idea that works. Interestingly, the founders of Kiva Systems worked previously at Webvan. If there was ever a company that invested too much in replacing human workers with technology, it was Webvan – they may have survived if they didn’t blow all of their capital on hugely expensive warehouses. That said, I suspect there are surely applications of the Kiva Systems for some retailers. But as a solution to the 4th quarter demand surge, I am skeptical.

Intel to invest in U.S. manufacturing

Intel announced today that it will invest $6-8 billion in the U.S. to build a new plant in Oregon and to improve existing  manufacturing facilities.  (See WSJ 10/19/10).  This investment will give them the capability to produce chips with 22 nanometer circuits, whereas the previous wave of investments gave it 32 nanometer capability. (A nanometer is one billionth of a meter – a couple of billion dollars to produce lines that are a billionth of a meter wide!)

The interesting part of this story is that Intel is sticking with U.S. production.  The potential allure of overseas production is not from the typical “cheap labor” issue, but rather because foreign governments are willing to hand out considerable subsidies for building a chip plant in their territory.  Intel estimates that the potential subsidies are equivalent to $1 billion, a considerable portion of the cost to build the plant.

So why is Intel sticking with the U.S.? Unlike apparel, they are not concerned about “made in America” – they will gain no premium in the market for made in America.  Unlike autos, their product is not expensive to move around the world, so local production to avoid higher transportation costs is not a concern. Nor is there a concern about tariffs or labor unions.

The issue is production yield – when you first start making chips, some of them, maybe even most of them, don’t work. The trick to make an investment in capacity pay off is to improve yields as fast as possible. And yields don’t improve because of better equipment, but rather because the equipment is used more effectively – engineers and line workers need to find out what particular recipe generates the best yield. Thus, improving yields requires intelligent experimentation, i.e., a skilled workforce.  It follows that Intel must believe that a U.S. workforce can improve yields faster than an overseas workforce, fast enough to  justify the additional $1 billion in cost.  It will be interesting to see how long that U.S. advantage can last.

Apple’s amazing run

Apple just announced some impressive profit numbers: it’s first quarter with $20 billion in revenue and profits that exceed those of  IBM, HP and Intel.  Only Microsoft had higher profit, but not a higher market capitalization.  The following graphic comes from the WSJ 10/19/2010:

The focus in the popular press on Apple’s success has been on how the consumer market loves their products and technology.   For example, the WSJ article discusses how Apple’s bet on the consumer market has paid off relative to IBM’s focus on the business market. I agree that Apple has been remarkable at design. But to me, they don’ t get enough credit for how well they execute.

Consider the iPad. It went on sale April of 2010. We have sales numbers from the first two quarters – by the end of Q2 (Jun 26) they sold 3.3 million and in Q3 (ended Sep 26) they sold 4.2 million.  That is 7.5 million units in the first 6 months of selling a product.  The fact that they sold that many units means that they were able to produce that many units, which is the amazing untold story.

How many iPads might they be able to sell in Q4 of this year. Let’s look at the sales history of the iPod.  Notice that they sold about 2.5 times as many iPods in Q4s as they did in the previous Q3. That means that Apple is on track to be able to sell about 10-11 million iPads in Q4.  How can they ramp up the necessary capacity to make all of them? They don’t talk about their capacity management strategy, but I suspect the following are the keys to their strategy. First, they must be tracking sales weekly if not daily and making updated forecasts and comparing them to previous sales trajectories, like those of the iPod.  Second, they have outsourced production to a flexible firm who can switch labor from other products to the iPad if necessary. Third, either they are securing component supplies or their supplier (Foxconn) is securing component supplies – you cannot assemble a product if you don’t have the components. Fourth, the iPad must be made with components that are either standard (hence there is lots of available capacity in the market) or easy to make (in the sense that the yields are high and stable already). On the last point, if the iPad were made with a component that is unusual or hard to make, Apple simply would not be able to ramp up production as quickly as they seem to be able to do.

Apple executes so well that they make it look simple. But it is a mistake to assume it so simple. To relate this to baseball, Mariano Rivera (the closer for the Yankees, i.e., the pitcher that comes in very late in the game with the job of defending a small lead)  has two pitches that he throws with deceptive ease. It might be tempting to conclude that there should be many pitchers that could throw Mariano’s two pitch combo. Many can, but none have thrown them as successfully as Mariano. Quality execution is often underrated.