Inventory Management Review

Cutting down on WIP

Generally speaking, there are three types of inventory that require management: raw materials, finished goods, and work in process (WIP).  Raw materials inventory can be cut down in a number of different ways, most notably by ordering smaller batches with more frequency from suppliers (JIT).  Finished goods inventory can be cut down in multiple ways as well, most notably by either producing only when you actually have an order (JIT) or by achieving more accurate demand projections (which in a sense is producing only when you have an order).  This post will cover two very practical applications of lowering WIP.

Process Pooling
The first way to cut down on WIP is to pool processes.  If a single worker is responsible for the entire production of a good from start to finish, then there will be no WIP.  By pooling workstations you can reduce queuing (wait time before a process) because there are less stations for the work to flow through.  If there are less stations, then there is less oppurtunity for work to pile up between stations.  Pooling however is not always a option.

Kanban
Another highly effective way of cutting down on WIP is through the use of kanban, Japanese for sign.  Kanban is a tagging system that is essentially a communication system designed to inform production across the entire plant about what to do and when to do it.  Hopefully the following example will help to illustrate kanban.

Imagine you work at a plant that produces two different types of footballs, NCAA size and NFL size.  Now, because the NFL likes spectacular touchdown passes, the footballs are a little bit smaller.  Unfortunately for you this requires you to create two different products.

You have to manufacture thousands of both types of footballs.  Luckily, manufacturing the footballs is, in this simplified example, a simple five step process that is completed at three work stations. At the first work station, the leather is cut, and then painted with the NFL or NCAA logo. At the second work station, the leather is stuffed with its rubber insides then stitched together.  The final stage of production is packaging.

The following table shows the cycle time (average time it takes to process a part) at each station:

Cutting and painting: 30 seconds/football

Stuffing and stiching:  1 minute/football

Packaging:  45 seconds per football

Remember from my post on cycle times (essential to understand before reading on), the longest cycle time, or the bottleneck, determines the overall speed at which a group of processes can turn out a new product.

Back to the problem at hand, the stuffing and stitching takes twice as long as the cutting and painting.  What would happen if both of these processes worked at 100% capacity utilization?  WIP would pile up sky-high in front of the stuffing and stitching station.  Until capacity increased at the stuffing and stitching station or capacity utilization decreased as the cutting and painting station, WIP would continue to pile up.  Clearly, production at the cutting station needs to be limited.

Although the cutting production needs to be limited, there always needs to be some work for the stuffing and stitching to do.  When there is no work at the stuffing and stitching station, the bottleneck is constrained.  Remember, the speed of the entire system is the speed of the bottleneck, so as the bottleneck slows, so does the entire system.  It is essential to keep the capacity utilization up at the bottleneck.

Pooling Solution
If the cutting and painting station pooled with the stuffing and stitching station, the cycle time should become the average of the two which would make it a 45 second cycle time putting it in sync with the packaging station.  This would require both stations to work at an average rate equal to each which would drastically improve performance.  Eliyahu Goldratt's The Goal is a good read that points out some problems with matching capacity across a production plant, but this would nevertheless be an improvement on the old system.

Kanban Solution
The stuffing and stitching station is working on an order.  Another order comes in for 1000 NCAA footballs.  A stack of raw materials leather gets placed in front of the cutting and painting station, behind a few other orders waiting to be processed, with a kanban tag saying "1000 NCAA footballs."  This ensures that the cutting and painting stations knows what to produce and will begin to produce it as soon as it is that order's turn.  As the WIP moves through the plant it is worked on when it is its turn and is processed to become NCAA footballs.

Essentially this is a system that tells people when to produce what.  This example was overly simplistic, but imagine that there are many different types of products and that the WIP has an oppurtunity to become a number of different types of FG at any point in the production process.  This helps to notifiy workers what exactly it is that the product needs to become.

More importantly, it limits the amount of WIP in the system.  If there is an excessive amount of WIP in front of the stuffing and stitching station, then it doesn't really matter how fast the cutting and painting can work.  For example, if the WIP is already at a high level in front of the stuffing and stitching station, then managers can simply not tag anymore leather for processing at the cutting and painting station.  Cutting and painting can sit idle until stuffing and stitching gets closer to being able to process new orders.  Ultimately, the same orders are processed at the same speed, but with less WIP.

By limiting the number of kanban tags that are released, and therefore limiting the capacity utilization of the cutting and painting station, the number of WIP is limited.  By limiting the work that can be produced by the first station, raw materials can be limited.  You can learn to reorder raw materials based on the demand the kanban tags will be placed on raw materials inventory.  Conclusively, kanban tags allow managers to determine the WIP they want at their plant.

The Risks of Being Just-In-Time

The following is a guest article written by Nick Koletic, an economics specialist at UCLA.  In addition to giving a brief background on Just-In-Time inventory system’s benefits, the article’s main focus is the risks that JIT systems face.

Just-In-Time inventory (JIT) is part of a production system whereby a firm vastly reduces inventory from its production processes so that utilization of production inputs and delivery of finished products are accomplished without incurring significant holding costs.  While JIT inventory systems are quite attractive for this reason, they are a double-edged sword. And though a JIT system might even be a necessity given the inventory demands of certain business types, its many advantages are realized only when some significant risks to healthy inventory management are mitigated.

JIT systems have several cost-cutting advantages.  As Charles mentioned in his Dell Computer case study, JIT inventory systems, a “financial imperative” for Dell, can radically reduce holding costs.  In the case of Dell Computers, this meant that the fewer finished computers Dell holds in inventory, the less money they lose per computer as they “rot” on a shelf.

In addition to these significant cuts in depreciation costs, which for Dell can be up to 1 percent per computer per week, JIT inventory can also cut storage costs.  One can imagine how Toyota, a pioneer of JIT systems, might save on storage costs as their finished computers and cars no longer sit idle in warehouses awaiting customers.  And these storage cost savings apply not only to these finished goods, but also to parts that Toyota might use as inputs in production.  These inventories are kept at a minimum through JIT systems as parts are ordered as needed.

JIT systems also cut delivery costs as finished products are shipped to where they are in demand.  Shipping the same quantity of a product to different retail outlets, for example, might not make much sense if the demand for that good is significantly greater at one location relative to another.  This approach to delivery cost savings also facilitates decreases in aforementioned holding costs by not overstocking certain locations with a product.  The same principle holds for inputs in production; parts are not delivered and held at production centers where they might lay idle.

Some positive externalities may also result from a firm’s decision to implement a JIT system.  Suppliers of such a firm, for example, might then be able handle larger orders but fulfill them with smaller shipments.  That is to say that for any given order size, supplying a customer that utilizes JIT is typically easier to do because individual shipments tend to be smaller for these customers and thus tend to be less demanding of the supplier.  So it might be possible for suppliers, merely by the nature of their customers’ JIT system, to greatly expand their ability to fill larger orders without having to increase production capacity.

Several factors, however, make JIT systems a risky proposition.  A key concern here is the extent to which firms are dependent upon particular suppliers under such an inventory system.  For example, if a firm were to commission a highly proprietary product to a single supplier (single suppliers being common in JIT), a JIT inventory system would put such a firm at an even higher risk of rip-off on behalf of the supplier because the firm would have no immediate inventory to buffer an interruption of supply.  Such an interruption of supply might be so costly that the firm might just allow the supplier to overcharge the firm up to the cost of this interruption.  This rip-off cost might completely cancel out or even exceed the savings that drove a firm to utilize a JIT inventory system in the first place.

Even more dangerous are internal issues that might lead single suppliers to be unable to fulfill a firm’s orders.  In this case, the firm has no option but to incur the costs of an interruption of its production input supply.  Internal issues might include, say, labor strikes on behalf of the supplier’s employees in which labor unions could hold the supplier for ransom up to the amount of its pending orders, again leading to an interruption of the firm’s supply of production inputs.  But internal issues can mean a host of things (and no, I'm not talking about a Webhost) that prevent a firm’s supplier from supplying.  The point is that by facilitating the interconnectedness between businesses, JIT inventory systems increases the risk that problems or failures on one end of the production chain might be felt on another end.

However, these risks associated with JIT inventory systems may be ameliorated to a certain extent.  Indeed, the evolution of the organization of firms has already taken many of these risks into account, particularly with respect to rip-offs.  For example, firms that might otherwise commission highly proprietary products to a handful of suppliers usually either produce these items themselves or in fact own the suppliers that do so in order to prevent price-gouging from occurring.

If in-house production or a supplier buy-out is not a feasible option, firms still have other common-sense ways of preventing these risks.  A firm might have to really scrutinize the integrity of their suppliers not only in terms of their trustworthiness but also in terms of the health of their business; contracting with a supplier at risk of going out of business makes little business sense in general, but firms with JIT systems are and should be even more acutely aware of these scenarios.  Taking it a logical step further, a firm might contract with several suppliers in order to lessen the harm done by any one of them failing to supply.   Furthermore, for the risk-averse firm, short-term and non-exclusive contracts with suppliers might also be attractive as they provide both insurance and punishment against a supplier’s “misbehavior”.   A supplier would have less incentive to misbehave and the firm would have more recourse under such an arrangement.

Just-In-Time inventory systems provide for an attractive, cost-cutting production system as long as risks are weighed and mitigated.  Preventative measures introduced here are by no means meant to be an exhaustive list of how firms should approach these risks, but rather are suggestions to the preliminary considerations firms should make in implementing a successful Just-In-Time inventory system.

Cycle Times and Bottlenecks

In order to figure out how much inventory you need, you should really know your capacity.  This post is going to go step by step and cover in detail the difference between throughput, cycle times, and bottlenecks.  The example we'll look at is going to be a handbag manufacturer.

The following steps, with their corresponding times, are required to produce a handbag:

Process                                  Process Time

Leather cutting                                 120 seconds

Zipper attachement                            70 seconds

Sewing                                               30 seconds

Throughput time
The times that are shown are the processing times.  They represent how long it takes an individual bag to physically pass through each step.

The throughput time is the total time it takes for a bag to physically pass through all steps.  It is the sum of all the times and it is 220 seconds.

Cycle time
There are two cycle times.  There is the cycle time for each station and there is the cycle time for the entire handbag manufacturing process.  To determine this, you will need the capacity which can be determined from the following table which shows how many workstations there are for each process.

Process             Stations available     Cycle Time

Leather Cutting            6                                20 seconds

Zipper Attachment       2                                35 seconds

Sewing                         1                                30 seconds

The cycle time for each station is the average amount of time it will take each bag to be completed.  Assuming each station can process 1 bag at a time, cycle time is the process time divided by the number of workstations.  Leather cutting for example takes 120 seconds per bag, but since there are 6 of them being worked on at a time, it is as if one bag is being completed every 20 seconds.  Now to analyze the cycle times to determine the bottleneck.

Bottleneck
The bottleneck is the slowest cycle time in the process and it determines the rate at which the entire system can produce handbags.  In other words, the bottleneck, which is the zipper attachment process, determines the capacity for the entire handbag manufacturing plant in our example.

Here's how this works:
Imagine you're standing at the end of the assembly line.  You know it takes an individual bag 220 seconds to be processed from start to finish, but how long is the wait for individual bags at the end of the line.  In other words, if you're standing at the end of the line, how much time elapses between finished bags?  I'll give you a hint, it's a lot less than 220 seconds.  Remember, during that 220 seconds, there are multiple bags being processed at any given point in time by the multiple work stations.

The answer is once every 35 seconds.  This is the time it takes on average per bag at the bottleneck and is therefore the time that elapses between each complete bag.

At station 1, there is an average of 1 cut piece of leather transferred to station 2 every 20 seconds.

Station 2, however, can only process 1 bag every 35 seconds.  This means that station three only gets a new a bag to work on every 35 seconds.

So, station 3 completes a bag in 30 seconds.  But then it waits 5 more seconds until station 2 can finish its work and send over another bag.

Therefore, station three sends one finished bag to the end of the production every 35 seconds.

No matter how fast station 1 works, station two cannot handle anymore inventory than it's already getting.  Also, no matter how fast station 3 works, the system will not produce bags at a faster rate.  If station 3 could sew the bags together at a rate of 1 bag every 10 seconds, it wouldn't make a difference because station 3 would still have to wait for station two to supply it with more work in process.

Only if the bottleneck is improved will the rest of the system improve.

That said, take a look at leather cutting.  Is it really necessary to have six workstations.  I can't see what's good about it.  Even with only four workstations, it still operates faster than the bottleneck and therefore would be sufficient capacity.

Currently, station one is over-producing.  If station 1 were to produce non-stop, there would be no way station 2 could keep up and huge work in process inventory would be the result.  A post in the near future will cover this topic in greater detail.

In the mean time, if you can't wait, I highly suggest reading Eliyahu Goldratt's The Goal.  This book is an essential (and easy, enjoyable) read for anyone who wishes to have a greater understanding of the effects of bottlenecks on inventory control and plant capacity.

Implementing Reordering

Actually reordering and knowing when to hypothetically reorder based on ROP formulas are not the same thing.  Of course, in order to actually reorder, you should know when to do it.  That is, you shouldn't reorder without ROP.  You should determine the ROP and order accordingly, which requires constant inventory level tracking.  My last post, Different Inventory Systems, detailed ways to track inventory levels.  So if you track inventory levels, and know ROP, you should be good, right?  Well, not really.  You still need two things: someone to actually reorder and the knowledge that it's time reorder conveyed to the person in charge of reordering.

First, let me give an example that should help demonstrate the issue at hand.  The firm that I wrote about in Notes from an Inventory Management Consulting Job did not have a perfectly smooth transition with the new system.  While it is working to their advantage very well now, there were some bumps in the road.

Primarily, there was a bump in the road of not reordering materials on time.  Now, this is a pretty big problem considering that the system was commissioned so they would know when to reorder.  Apparently people don't like spending money on a system that doesn't help them.  Now, I don't want to say it wasn't my fault that it wasn't working, BUT, luckily it wasn't my fault and the issues have since been fixed.

You see, the system was designed so that when materials hit the safety stock level, it was time to reorder.  Despite information in the system that very clearly alerts users that it is time to reorder by saying "NOW!" in big red letters when it is time to reorder, there were still issues regarding the actual reordering.  There was even a countdown by the system that allows users to know how many days they have left until they have to reorder.  It was really quite clear.

The problem, as I discovered, wasn't so much that they didn't know when to reorder, or how much to reorder, or what their inventory levels were at so they could decide to reorder or not.  The problem was they didn't reorder.

There are two solutions to this.  First, I could have manipulated the system so that it told them reorder a couple of weeks before they really needed to.  Basically, I could have implemented a safety net into the system that would account for their inability to reorder on time.  For some firms, this could be a viable solution.  A good example would be if there is a redesign that needs to performed each time materials are reordered.

The second solution is to reorder the damn materials when the system tells you to.  This, although not in so many words, is what I suggested to them.  Unlike the first solution, this solution reorders when it makes economical sense.  The other solution dumps in excessive inventory into the system.  It has you reorder before you should which raises average inventory, which needlessly ties up cash into the system.

Back to what I was saying earlier, you need someone to reorder and you need that someone to have the knowledge that it's time to reorder.  Here, the firm needed someone to actually do it.  People had the knowledge, but there wasn't someone who got the job done.

So how do I implement this?
There are a number of different ways to implement this.  While I won't go into detail regarding building a system, I will give the basic requirements on managing the implementation.

Most importantly, knowledge needs to be conveyed.  When your inventory tracking system recognizes that levels have reached the ROP, this information needs to be made available to people that can do something about it.  I'm not saying sirens need to go off, but something needs to occur that results in a materials orderer discovering that he has an order to place.

As I discussed eariler, someone needs to have the job of actually reordering.  I see it too often that people don't reorder when they know they should.  Don't let this be you.  Don't be the manager who invests money into a system but doesn't invest time into ensuring his employees are properly using it.

Before I conclude, I will mention that this entire process can be automated.  EDI, or electronic data interchange, used alongside RFID can accomplish such.  RFID is used to track inventory and when the levels drop below safety stock levels, EDI automatically places orders with suppliers.  Wal-Mart is a great success story regarding these two technologies.

Different Inventory Systems

Assuming a firm has performed Economic Order Quantity and Reorder Point analysis to determine how much to inventory to order and when to reorder inventory, the firm still needs to be able to keep track of inventory levels so they know when they reach the Reorder Point. This article will detail a number of different inventory systems starting with the most archaic and working up to the most advanced.

Physical Counts (Periodic System)

The most basic method of inventory tracking is physical counts. Physical counting is exactly what it sounds like; if you want to know your inventory level, you better take off your shoes so you can count on your toes when your fingers run out. Prior to my work described in Notes from an Inventory Management Consulting Job, the firm I consulted for relied on physical counts to try and figure approximately where they were at. Lucky for them they didn't bother with ROP because as you can appreciate, it's very difficult to reorder at the reorder point if you aren't constantly aware of your inventory levels.

Two-Bin System

The two-bin system is only slightly more sophisticated than the physical count system. Using the marvels of modern technology, this system uses two bins of materials. When one bin is empty, it's time to re-order.

Perpetual Tracking

A perpetual track is the method I designed in the previously mentioned consulting job. This method of counting is demand driven. Instead of counting how many items are in inventory, you count how many leave inventory. The demand can be tracked by batches of inventory usage, such as demand that is entered once a week, or they can entered in real-time which provides the ability to continuously monitor inventory levels. If you are already in the practice of counting demand, this is a great way to track inventory because it involves little additional effort. For what I set up, it was ideal because there is little variability in the products that are demanded. For a supermarket, this can be more difficult because of the variety of products sold. Luckily, the next section highlights how firms with broader product lines get over this hurdle.

Universal Product Code (UPC)

UPC is a system that supermarkets first implemented in the 70s (more info). This requires unique codes to be put on all types of inventory and is usually accompanied by a bar code that can be scanned via infrared scanning guns. If you've been to the supermarket at in the last 30 years and had a pulse at the time, you probably noticed that everything you buy is scanned into their system. In addition to helping the market determine how much you should pay, this also gives supermarkets, and other firms, the ability to track and count the movement of any and all inventory with a simple infrared scan.

My favorite example of infrared product tracking is the implementation of package tracking that Fedex and UPS have incorporated into their business processes. In case you're unaware, both Fedex and UPS now track all packages at every stage from pickup to delivery. The beauty behind the system these two companies have implemented is that not only does it directly help their operations management, but it also directly improves customer service. Now, customers can log onto a Fedex or UPS website from anywhere in the world, enter in the tracking number they received when they dropped the package off for shipping and know exactly where their package is in real time. Marketing studies have shown that informed customers are typically more satisfied and this process is a wonderful example of how the businesses are able to inform themselves and their customers with one technology.

Radio Frequency ID (RFID)

RFID, which I have written an article about, is another method of tracking inventory. Instead of using technology to track inventory as it is moved, RFID counts inventory automatically from a remote location. This is superior to perpetual inventory tracking or perpetual inventory with the UPC for a couple of reasons. Most notably is that RFID accounts for shrinkage (lost inventory, not what happens to Costanza in cold water). As much as we might not want inventory to just disappear, the fact of the matter is, things grow legs. Also, inventory is often scrapped. Perpetual counts that lower inventory levels only when there is demand don’t account for lowered inventory when a good is stolen, unless the thief enters it into the system to be a nice guy. RFID counts what is actually there, and it can tell you exactly where it is.

Different systems have their time and place so you should consider what's right for you. It may turn out that the two bin system is a great fit for your operations. If holding costs are really low, why not go for the two-bin approach. Hell, you might not want to even invest in two bins. Maybe you can easily do a physical count on a daily basis. The key is to find a system that works for your needs.