Archive for the ‘Optimization’ Category

Change is hard.

Collapsed Souffle

In the businesses that I help, change comes for several reasons. It may be thrust upon the business from the outside, a change in the competitive landscape for instance, or a new regulation.   It may come from some innovative source within the company, looking for cost savings to increase profitability of productivity, or a new process or product with increased productivity. Change can come from the top down, or from the bottom up. Change can come in a directed way, as part of a larger program, or organically as part of a larger cultural shift.  Change can come that makes your work easier, or harder, and may even eliminate a portion (or all) of the job that you were doing. Change can come to increase the bottom line or the top line. But primarily change comes to continue the adaptation of the company to the business environment.  Change is the response to the Darwinian selector for businesses.  Adapt or decline. Change is necessary.  It is clear to me from my experience that businesses need to change to stay relevant.

This may seem trite or trivial, but accepting that change is not only inevitable, but that it is good, is the shift in attitude that separates the best companies (and best employees) from the others.

So, you say, I see the need to change, it is not the change itself that is so difficult, but rather the way that it is inflicted upon us that makes it hard.  So, why does it have to be so hard?  Good question.

Effective managers know that change is necessary but hard. They are wary of making changes, and rightly so.  Most change projects fail. People generally just don’t like it.  Netflix is a great example.  Recently, Netflix separated their streaming movie service from their DVD rental business. After what I am sure must have been careful planning, they announced the change, and formed Quikster, the DVD rental site, and the response from the customer base was awful. As you likely know, Netflix, faced with the terrible reception from their customer base and stockholders, reversed their decision to separate streaming from DVDs. What was likely planned as a very important change, failed dead. Dead, dead, dead. Change can be risky too.

If change is necessary, but hard and risky… how can you tame this unruly beast?

The secret of change is that it relies on three things: People, Process, and Technology. I name them in the order in which they are important.

People are the most important agents relative to change, since they are the one who decide on the success or failure of the change. People decided that the Netflix change was dead. People decide all the time about whether to adopt change. And people can be capricious and fickle. People are sensitive to the delivery of the change.  They peer into the future to try to understand the affect it will have on them, and if they do not like what they see…  It is the real people in the organization who have to live with the change, who have to make it work, and learn the new, and unlearn the old. It is likely the very same people who have proudly constructed the current situation that will have to let go of their ‘old’ way of doing things to adopt to the new. Barriers to change exist in many directions in the minds of people.  I know this to be true… in making change happen, if you are not sensitive to the people who you are asking to change, and address their fears and concerns, the change will never be accepted.  If you do not give them a clear sense of the future state and where they will be in it, and why it is a better place, they will resist the change and have a very high likely hood of stopping the change, either openly, or more likely passively and quietly, and you may never know why the fabulously planned for change project failed.

Process is the next aspect of a change project that matters.  A better business process is what drives costs down. Avoiding duplication of efforts, and removing extra steps. Looking at alternatives in a ‘what-if’ manner, in order to make better decisions, these are what make businesses smarter, faster, better.  A better business process is like getting a better recipe for the kitchen. Yet, no matter how good a recipe; it still relies on the chef to execute it and the ovens to perform properly. Every business is looking for better business processes, just as every Chef is looking for new recipes.   But putting an expert soufflé recipe, where the soufflé riser higher, in the hands of an inexperienced Chef does not always yield a better soufflé.  People really do matter more than the process.

Technology is the last aspect of the three that effect change. Better technology enables better processes. A better oven does not make a Chef better.  The Chef gets better when they learn to use the new oven in better ways, when they change the way they make the soufflé, since the oven can do it.  A better oven does not do it by itself.  An oven is just an oven. In the same way, better technology is still just technology.  It by itself changes nothing.  New processes can be built that use it, and people can be encouraged to use it in the new process.  Technology changes are the least difficult to implement, and it is likely due to this fact that they are often fixed upon as the simple answer to what are complex business problems requiring a comprehensive approach to changing the business via it people, process, and technology.

Nice Souffle

Change is necessary, but hard and risky. Without change businesses will miss opportunities to adapt to the unforgiving business world, and decline. However, change can be tamed if the attitude towards it is changed to be considered a good thing, and is addressed with a focus on people, process and technology, in that order.  Done right, you can implement the change that will increase the bottom line and avoid a collapse of your soufflé.

Rich Guy

The rise of zombies in pop culture has given credence to the idea that a zombie apocalypse could happen. In a CFO zombie scenario, CFO’s would take over entire companies, roaming the halls eating anything living that got in their way. They would target the brains of supply chain managers and operations people. The proliferation of this idea has led many business people to wonder “How do I avoid a CFO zombie apocalypse?”

Supply chain managers are seeking and developing new and improved ways to exploit the volumes of data available from their ERP systems. They are choosing advanced analytics technologies to understand and design efficient sustainable supply chains. These advanced analytics technologies rely on the use of optimization technology. I am using the mathematical concept of “optimization” as opposed to non-mathematical process of making something better.

Mathematical optimization technology is at the heart of more than a few supply chain software applications. These applications “optimize” some process or decision. Optimization-base programs, for example, those frequently found in strategic supply chain network planning, factory scheduling, sales and operations planning and transportation logistics use well-known mathematical techniques such as linear programming to scientifically determine the “best” result. That “best solution” is usually defined as minimizing or maximizing a single, specific variable, such as cost or profit. However, in many cases the best solution must account for a number of variables or constraints. Advanced analytics technologies can improve a company’s bottom line – and it can improve revenue, too! CFO’s like this.

Advanced analytics technologies provide easy-to-use, optimization-based decision support solutions to solve complex supply chain and production problems.  And, these solutions can help companies quickly determine how to most effectively use limited resources and exploit opportunities.

So, from my perspective, there are seven practical reasons to embrace advanced analytics technologies:

1. Your company saves money, increases profits.
2. You get to use all your ERP system’s data.
3. It’s straightforward and uncomplicated.
4. You have the tools to discover great ideas and make better decisions.
5. At the end of the day, you know the total cost of those decisions.
6. You have a roadmap to make changes.
7. You avoid the CFO zombie apocalypse

Imagine that you’re running a business that offers a wide variety of products ranging from security, to bus and ferry transportation, to public toilets, to refreshments. You have somewhere between 45,000 and 65,000 customers, who all arrive at your various locations en mass, requiring service over a very short time span. To further complicate your situation, you don’t get the opportunity to train many of your ‘employees’ on what they should or shouldn’t do; instead the best you can do is rely on their common sense and good intentions. Actually, since many are volunteers, you can’t necessarily be sure as to the actual number of people who will come to work for you. This is the business that the New York Road Runners Club will be in on November 6 during the running of ING New York City Marathon. And the lessons that the race organizers have learned for managing the event are of universal value and applicability to many logistics and supply chain organizations.

Consider food and water. There are about 23 locations along the route (about 1 per mile starting at mile 3), where mostly volunteer staff is responsible for “stocking” cups of water, Gatoraide, and PowerGel (only available at mile 18). In the past, certain of the less desirable stations have been undermanned, while at others as the day drags along some of the volunteers simply up and leave; “I’m a volunteer, so fire me”. And of course, once the race is over, the staff that remains need to perform the thankless and definitely unglamorous job of cleaning up (at least to some extent) the litter that has been created.

And then there is security. The race attracts a number of world-class runners and security precautions need to taken to make sure that fans and paparazzi do not become too intrusive. To insure that these individuals return year after year, the organizers make sure that these athletes are able to start the race at the head of the pack so that they don’t have to dodge other slower amateurs. In addition, a large number of people typically run the race (and utilize the services) without having officially registered and paid any money. This year, there were about 140,000 applications for the 62,000 slots in the race. This means that about 78,000 people were rejected, and it is estimated that about 15% of these will run the race anyway. The Road Runners Club does not try to prevent any of these unofficial entrants due to the chaotic nature of the race start. However, security is in place at the finish, when the runners are strung-out, so that these “bandits” as they’re called, do not receive any of the available memorabilia.

Finally there’s transportation. Since the race is a “point to point” one, where the finish is at a different location than the start, runners are provided transportation to the beginning of the race. About 20,000 participants will take 522 chartered buses to the starting line. And they’ll all need this service in the space of about 2.5 hours. Since there is only 1 available bridge, if one of the buses were to breakdown at a critical location, a major traffic jam could result delaying many more runners than just those on the one disabled vehicle. Alternatively, approximately 21,100 entrants are expected to use the Staten Island ferry system. Between 5:30am and 8:30am, there will be a ferry leaving Manhattan every 15 minutes. Although the largest of these can accommodate 6,500, the organizers will try to limit each individual trip to half of the ship’s capacity. Again, this is an attempt to make sure that in the event of a breakdown, the number of runners affected (either delayed in Manhattan or stuck on a boat in the middle of the harbor) will be minimized. In fact, one of the ferries did experience mechanical problems in 2010. And when the runners disembark the ferry, a fleet of 70 buses will shuttle them the 2 or 3 miles to the starting line: 10 buses will load simultaneously at the St. George Terminal for the short trip. Last year the average load time for each group of 10 was 4 minutes, 22 seconds.

The New York Road Runners Club typically receives wide praise on blogs and other feedback forums from participants, for how well the Marathon is organized and run. In light of the unique circumstances of such an event, they have learned how to run their logistics operation like clockwork, always anticipating and planning for the worst that might happen. They’ve figured out how to manage a supply chain that is spread over a wide area and with a “work force” that presents its own unique challenges. Many businesses would do well to study their methods and take a page from the same playbook.

I recently saw this short post on a supply chain-oriented LinkedIn group:
I am responsible for 17 warehouses around the GCC. I want to create 4 major “Hubs” instead. Which would you choose and why, centralized or decentralized warehousing?
This post elicited a variety of responses. Suggestions ranged from looking at the details
Is it a circle route covering all four hubs or direct to and from specific hubs? Because backhaul opportunities can impact the overall costs within your network.
to broader advice, such as
Success of this depends on demand of product category and lead time importance.
When I am faced with a supply chain network design problem, like the one implied here, my first step is to develop a clear definition of the question to be answered – not to hastily jump to a solution. Along with defining the question, you need to determine
• the objective (minimizing cost, maximizing profit, or something else?),
• the options available (the current and potential locations for warehouses in the area, in the context of the post above),
• the constraints that are of importance (service requirements, warehouse capacity, transportation lanes),
• the time horizon, and related criteria.
Then, in the case of a complex supply chain, I find that building a model and using it to evaluate options can give insight regarding both the quantities and the qualities associated with a given solution. Modeling requires effort – you need skills, as well as various data, such as the sales forecast, warehouse location options, transportation and duty costs, etc. – but with the help of a model, and good quality information to populate it, you are able to estimate the implications of different supply chain options, and know whether operating with four hubs is more cost effective than five, or three; where should the hubs best be located; or whether you should have hubs at all. A model can help you make evaluate a variety of trade-offs so that you arrive at a better-informed, and more profitable, decision on how to proceed.

Many commercial enterprises and public agencies operate telephone call centers to provide effective and timely service for customers. By employing nearly 5% of the national workforce, call centers arguably define the “new factory floor” in an increasingly service-based economy. They are fascinating socio-technological systems which are exceptionally well-suited for the application of mathematical modeling and optimization methods.

A typical call center utilizes a computerized call handling system which can archive detailed historical information on call volume, call handling time, and other relevant attributes. This data can be analyzed and aggregated (with appropriate accounting for probabilistic variation) to generate a profile of staffing requirements across future time intervals. In theory, service agents can be optimally scheduled to closely accommodate this profile, resulting in high service levels, low customer abandonment, and efficient agent utilization. In actual practice, however, such performance represents the exception rather than the rule. Most call centers, even well-run ones, do not simultaneously achieve high levels of service quality and operational efficiency [1].

One important reason for the performance gap between theory and practice is lack of sophistication and flexibility in the standard software systems available for call center management. For example, standard systems invariably base interval staffing requirements on the classic “Erlang C” model, which is known to produce distorted results because it does not consider pertinent factors such as customer impatience [2]. Additionally, if the software has any capability for schedule “optimization,” the underlying algorithm is usually a greedy heuristic which sequentially adds agent shifts without due consideration of the complex interactions between them. Beyond these technical limitations, standard systems offer minimal capability to experiment with different shift types and customize the solution strategy.

Profit Point can provide the expertise and custom tools necessary to properly model your unique call center environment and achieve optimal performance. By applying recently-refined mathematics, interval staffing requirements can be accurately determined and optimal shift distributions can be precisely derived [3]. Efficiency improvements exceeding 10% are typical, coincident with improvement in service level performance. Many additional operational factors, such as on-line chat activity and agent specialization, can also be addressed. There is no better time than now for you to reap the rewards of optimizing your organization’s call center operations.

References

[1] Noah Gans, Ger Koole, and Avishai Mandelbaum, “Telephone Call Centers: Tutorial, Review, and Research Prospects,” Manufacturing and Service Management 5, 79–141 (2003).

[2] Lawrence D. Brown, et al., “Statistical Analysis of a Telephone Call Center: A Queueing-Science Perspective,” Journal of the American Statistical Association 100, 36–50 (2005).

[3] Dennis C. Dietz, “Practical Scheduling for Call Center Operations,” Omega 39, 550–557 (2011).

Heuristics and Optimization

 You might think the title should be ‘Heuristics or Optimization’, implying a choice.  But often the two approaches work well together with heuristics speeding an optimization process.  The Wikipedia definition of heuristic calls it an experienced-based technique for problem solving, learning, and discovery.  Wikipedia also mentions using heuristics to find a good enough solution and describes them as ‘strategies using readily accessible, though loosely applicable, information to control problem solving.’

 Those descriptions do not emphasize another aspect of heuristics – there is generally an underlying concept that informs the heuristic.  There is a good reason why we think it will work well in the majority of cases.  For example, an angle sweep heuristic is often used when designing routes for pickup and delivery from a central hub.  Those routes are candidates for selection in a formal optimization.  The designed routes look a lot like the petals of a daisy. 

 The heuristic starts out by heading north and picking locations close to that direction on the way out and back.  How far out the route goes is a property of vehicle capacity or time limitations.  The next route to be generated starts out slightly east of north and follows the same limitations and usually overlaps many of the locations on the first route.  Once the entire compass has been swept, the best set of routes to cover all locations is selected by an optimization.  In the example the heuristic becomes a front end for the optimization. 

 Another example comes from a driver scheduling problem.  Suppose a set of drivers must pick up some commodity from a set of locations for processing at a central plant.  Each trip in this example is an out-and-back because of the nature of the commodity, i.e. only one location can be visited.  Drivers pick up multiple loads in a day, and each location requires multiple visits.  The pickup times are fixed because of other problem features.  One approach is to simply allow all combination of driver-load-location pairings and let an optimizer grind away. 

 But there are other desirable features of the solution:  equalizing number of loads among drivers, and keeping driver dead time between loads to a minimum.  Specifying all the driver loads by some simple heuristic, e.g. send a driver out for the next load as soon as possible, usually ends up with some loads that cannot be covered.  A totally greedy approach fails.

 An approach that seems to work well in this case is to consider some drivers for early loads and some for late loads.  Work from the front of the early loads assigning each of the early drivers the first two loads they can feasibly complete.  Then work from the end of the late loads assigning the last two loads of the late drivers as the last two loads they can feasibly complete. 

 The loads in the middle and the drivers that are not considered early or late are handled by the optimization.  Notice that the heuristic does well on the driver gaps and guarantees that most drivers automatically get two loads, which is a good base in this application.  It also serves to speed the optimization by reducing the pairings to be searched while preserving enough flexibility to get a solution. 

 Furthermore, the heuristic is flexible in that one can choose how many drivers to consider early or late and how many of their loads to nail down heuristically.  Most importantly the heuristic gets better solutions than the optimization finds in any reasonable time.  So while the optimization must have that best solution out there, it will not find it in the time frame the scheduler has to work within.

 My experience is that flexibility is one of the key properties in any good heuristic.  Adaptability to new situations is also a feature of good heuristics.  One final example illustrating adaptability is based on an algorithm called ‘Toyoda’s Algorithm’.  I have applied this particular idea to a number of situations.

 In this example it is applied to sequencing the unloading of containers at a port.  Each container holds a selection of parts which have to be processed by various work centers prior to shipment out of the port.  A manifest shows the selection of the parts and it is known how much work is associated with a given part at a work center.  Not every work center can process every part type.  The objective is to get all the work centers to end their day at the same time and to keep all the work centers busy throughout the day.

 The approach is easy to understand in two dimensions.  The X and Y axis represent the available work time of two work centers, e.g. eight hours.  The arrows represent the amount of work delivered to a work center by a given container.  The dashed line is the ‘ideal path’ — equal amount of work at each work center throughout the day.

 The heuristic simply needs to loop through all available containers at each iteration and always try to get back onto the ideal path.  Penalties for deviation are totally flexible in that small deviations can be without penalty while sizable ones are some function that penalizes them heavily.  Other problem features can be captured, e.g., buffer space at a work center, and incorporated in the penalty function.  This is not a formal optimization, but it is speedy and good enough for the real world application.

 The watchwords seem to be these.  Look for the important features of a good solution.  See if a simple rule or concept will drive the solution toward these good features.  This is especially true when there is little or no economic benefit to the optimal solution.  Try to develop a heuristic that is flexible in adapting to normal variance in instances of the data and can be tuned to choose between competing objectives.

The summer issue of Manufacturing Today includes an article authored by Ted Schaefer and Alan Kosansky entitled Face Complexity – Making Sound Business Decisions.

“With every passing year, the amount and variety of information available to make business decisions continues its exponential growth. As a result, business leaders have an opportunity to exploit the possibilities inherent in this rich, but complex, stream of information. Alternatively, they can continue with the status quo, using only their good business sense and intuition and thereby risk being left in the dust by competitors. Top-tier companies have learned to harness the available data with powerful decision support tools to make fast, robust trade-offs across many competing priorities and business constraints.”

Read the complete article here: Face Complexity – Making Sound Business Decisions

Here at Profit Point we regularly hear from clients with well established Enterprise Resource Planning (ERP) systems that they need something more.  ERP systems are excellent for doing certain things including:

1. Providing central repositories of data
2. Enabling cross functional work processes within and across companies
3. Costing of goods
4. Planning resources and materials at a high level

However the more complicated your business work processes and manufacturing production processes the less sufficient a standard ERP system will be in providing the best decision support functionality.  Some of the complications that require decision support systems (DSS) and which we have been helping clients deal with lately include:

1. Work processes to handle make to order versus make to stock material assignments
2. Allocation of inventory to customer orders when in an oversold position
3. Sequence dependent setups / cleanings of manufacturing equipment
4. Scheduling of production sequenced through a “product wheel”

DSS are necessary because of the complexity of first finding a feasible solution and then having some means of sorting through the huge number of feasible options to find a “good” or “optimal” solution.  DSS help in these kinds of situations to:

1. Reduce costs
2. Reduce manufacturing lead times
3. Improve customer service
4. Increase revenue

ERP systems are a necessary part of being able to deliver a DSS by providing the data necessary for making the decisions in question but don’t have the following:

1. Ability to be tailored to a specific work process or manufacturing environment
2. Advanced analytical capability to sort through the complexity and volume of options to get to a “good” or “optimal” solution
3. Graphical user interface tools to be able to allow a user to visualize the data in a way that gives them the insights needed to make decisions

At Profit Point we specialize in listening to our clients needs and then building DSS to unlock improvement opportunities which enable our clients to outdistance the competition.

We recently attended a discovery meeting that was focused on how to conduct a strategic optimization planning study of an existing distribution network. The company wanted to know what changes needed to be made to lower the distribution costs. Several members of the management team were present and there were many questions regarding the ideal business process, study approach and modeling tools to be used to insure a successful project.

What was interesting to me was the overwhelming focus on the modeling tool. Questions about who would be on the project, the timeline, the types of scenarios, data gathering and validation were secondary. It may be important to have the right tool to model your infrastructure, but the real focus should be on the experience and modeling capabilities of the users of the tool.

These are the Critical Success Factors

1. Full participation in data gathering and results review by the project team and management.
2. Clear definition of the key questions to be addressed and the related scenarios required by the Project Sponsor early in the project timeline.
3. Availability of leadership resources within the company throughout the project to review assumptions and to ensure integrity and quality of the input.
4. On time delivery of a complete set of all required data by Project Team members.
5. Acceptance and agreement on the variable, fixed and capital cost assumptions of existing and potential new facilities.
6. Availability, communication, and collaboration of the Project Team members, support staff, and consultant for all working sessions, conference calls, and follow-up between meetings.

It’s important that the optimization modeling tool can incorporate the variables and constraints associated with your supply chain, but the real focus should not be on the tool, but rather on the experience of the users of the tool and their ability to deliver the results of a project. If I were to set out on a network optimization planning project to model my entire supply chain, then my primary focus would be on developing an experienced team of individuals that had the skills to minimize the above risks.

In a recent article, I discussed what are known as “Black Swan” events.  These unexpected, and extreme or severe events are frequently ignored and unplanned for by organizations.  These are occurrences that are thought to be outliers and can be safely ignored.  In that previous post, I mentioned a mathematical approach or formulation for modeling such incidents.  But even if management does not use rigorous mathematical modeling, organizations should always have strategies and plans in place in order to deal with sudden, tectonic changes in their environments.

The recent earthquake and tsunami in Japan highlight the need for businesses to acknowledge and anticipate these worst-case scenarios.  As the April 2^nd edition of ‘The Economist’ magazine outlined, the supply of certain critical components and materials has been severely disrupted by events in Japan.  For example, just 2 companies (Mitsubishi Gas Chemical and Hitachi Chemical) whose plants have been damaged, account for about 90% of the market for a specialty resin used to glue parts of microchips that go into a wide range of electronics.   Obviously, the immediate customers of these 2 will feel the impact immediately.  But ‘knock-on’ effects will ripple through Supply Chains all over the globe and as a result of the decreased output from these plants there will be companies all over the world that will find themselves unable to meet their own sales due to the constrained availability of this resin.  And these consuming companies may never have known that they relied on the 2 affected firms or realized their vulnerability to disruptions.  Another case in point is the battery in Apple’s iPod.  This uses a specific polymer made by Kureha, which accounts for about 70% of that market.  Again, their plant was also damaged in the recent disaster.

And the earthquake and tsunami have had other less obvious impacts on Supply Chains.  For example, operations at some Japanese ports have been disrupted in the aftermath of the disasters.  As a result, the worldwide availability and supply of shipping containers has been affected.  Certain containers that had been expected to arrive in the U.S. or Europe, and thus to have been available for reuse, have been damaged or delayed in Japan.  In essence, some Japanese ports have become “black holes” where containers are stuck and thus unavailable for use/reuse.  Also, due to the reduced capacity of Japan’s national electrical grid, world-wide Supply Chains that have links in Japanese suppliers, have been slowed or impeded in their ability to “keep up” with the other parts of their networks.

It is critical for managements to be more proactive, and anticipate Black Swan type occurrences.  They must remember those immortal words that “stuff happens, and then it happens again”.  Obviously, some things like earthquakes are very difficult to predict.  But disaster planning should be a regular and recurring part of the management process.  Companies should always take pains to make sure that they are not overly dependent on a single supplier, or a single region on the world.  They should take the time and effort required to investigate just where their Supply Chains are vulnerable, since the ultimate sources of some of their key raw materials suppliers may not be clear or obvious.

Just this week, IBM’s “Watson” computer showed off its impressive language processing capability by handily beating the best humans at the game Jeopardy!. This was of interest because Jeopardy is filled with tricky language such as puns, slang and wordplay; and Watson was able to process it all, figure out the context, and take it to the humans in winning handily. You can read about Ken Jenning’s firsthand account at http://www.slate.com/id/2284721/.

This story reminds me of the 1997 chess competition between Deep Blue and Gary Kasparov. I was fortunate to hear Kasparov speak at a supply chain conference just days after he had lost to Deep Blue. Despite the fact that he was deeply upset about having lost, Kasparov was able to share important insights that were relevant to the supply chain industry and business in general. What he pointed out was that the competition missed the real point. He described how machines were better at certain kinds of tasks (memorization of massive data, fast processing through the data, etc..), and humans were better at other tasks (certain kinds of inferences and relationships), and that in the future he would hope to see man-machine teams compete against one another to see who could create the best combination of person and machine to be the best at Chess.

Clearly this was an insightful comment to a room filled with supply chain experts. It is our job, every day, to make the best possible decisions in the face of an overwhelming amount of data in front of us. We know for sure, that we cannot rely completely on technology to make these decisions. We know equally well that experience and business savvy are not enough in today’s world to consistently make the best decisions for our business. So our challenge is the one of which we are once again reminded: how best to combine human ingenuity, experience and insight with the power of modern technology to make our business and supply chain be the best they can be. At Profit Point, this is our passion.

Okay. I am an anomaly. I live in Utah and drink coffee. The majority of the people that live in Utah do not drink coffee, and that is OK, but I do. So, is there a shortage of coffee Cafés in Utah? No. There are many cafés and several that serve outstanding coffee.

We have an exceptional espresso café downtown, located on a side street off of Main. They roast their own coffee and use triple certified organic grown beans. It is the type of place the local coffee lovers go to hang out and have good conversation over a morning or afternoon latté or espresso. Possibly the best coffee I have ever had. What is interesting to me is that a large percentage of the residents in my area do not even know that this café exists.

So what is my point? When it comes to outstanding services or products most people are unaware of what is available, primarily because it does not fit into their lifestyle or what they’re accustomed to. I believe you can transfer this similarity to the business world. Manufacturing logistics and transportation people become accustomed to doing things a certain way. Over time they may become blind to ideas for improving the supply chain. They are unaware of an exceptional Supply Chain Café, even when it is located just seconds from a combination of keystrokes and Google.

It is not their fault they are missing the best latté available. We, as consultants, who prepare those delightful solutions from the Supply Chain Café menu, have probably not done the finest job of promoting our services and software to your neighborhood, but that is changing.

There are many empty cups in the supply chain, waiting to be filled with successful solutions. Supply Chain and Logistic managers tackle difficult supply chain problems every day, but they are so focused on getting their job done and making it through the day that they have little time to think of alternatives that may improve their processes and well being. I am not sure how we can help everyone, so let’s focus on the window shoppers. These are the ones that are aware of the café, but have never been inside. Maybe you are one?

If you are reading this blog, then you must be a window shopper. I am guessing you are looking for a better espresso. OK, you found “Profit Point”, although you may not know what we do. Guess what? Help is on its way. We can share our menu with you. We just published four videos that will introduce you to the Profit Point team and what we do. Embrace three minutes out of your day, select one of the videos, and watch it. Learn how we help companies improve their supply chain, by serving the best coffee with a smile.

Yes, you can improve your supply chain with our help. The supply chain solution that you are looking for, is about to be yours. And if you place an order, we can fill your cup to the top, with the “good triple certified” stuff. If you cannot seem to find that special item on our Supply Chain menu, then no fear, we love special orders.

So, is there a shortage of Supply Chain Cafés? No. You just need to find the one that serves the optimal latté. I know it’s out there somewhere.

To learn more about Profit Point’s Global Supply Chain Optimzation services, please contact us.

Frequently, you might hear somebody say that the capacity of a production facility is some known and fixed value.  When this happens be very wary of what they might be trying to sell you.  Because as with so many other things, when measuring capacity ”the devil is in the details”.

The “capacity” of a factory sounds like a pretty simple notion and something that should be easy to calculate.  But this is only true for production systems that are fairly straightforward, consisting of totally independent machines and processes.  If the organization however consists of operations that are interconnected and interdependent on each other, then capacity can be a fairly difficult thing to measure.

In the vast majority of production systems, there is a very real link between capacity and three critical factors:

1. the mix of products, and how much time is required for setup/cleanup between consecutive production runs,
2. the ability to create sophisticated and optimal schedules for the production resources,
3. how much physical space exists in the factory where products that are only partially complete can be kept or stored; what’s known as Work in Process (or WIP)  Inventory.

To see these 3 relationships at work, consider the simple case where a certain department produces two products, A and B, which both use the same piece of equipment, and there is only one of these machines available.  The production rates of the machine are in the table below and there is a 4 hour setup time required when the machine switches over from producing one product to another.  Now consider the 2 scenarios below.  In Scenario A, the capacity is 170 units per day while in scenario B the capacity is 145.

This example clearly demonstrates the frist item above, that the “capacity” of the department depends to a large extent on the mix of the 2 products that are being produced.

Now suppose that management wants to produce 110 of A and 80 of B per day.  These new requirements seem to clearly exceed the capacity of the department given EITHER Scenario A or B.  But maybe the necessary capacity can still be found.

If the new requirement is to produce at this increased rate for only a single day, or to produce at this rate each and every day, then there is definitely not enough capacity on the machine.  However, if the increased production is required over a sustained length of time, then we can gain extra production by modifying the production schedule so as to eliminate or minimize the occurrence of the 4 hour setup.  If the department schedules production in long blocks spanning several days, where first one product and then the other is produced, then the department DOES have the capacity.  In the table below for example, 440 units of A is first produced followed by 320 of B, with a 4 hour setup between them.  This represents 4 days worth of the increased management requirement (100 of A and 80 of B each multiplied by 4).

With this schedule, the total required hours of 94 is less than the 96 hours available in 4 days, and so now there IS enough capacity!  By scheduling wisely (i.e. “working smarter”), the department’s average daily capacity has actually risen to (760 / 4) = 190 units per day, a good deal higher than either 170 or 145 in the two previous scenarios.

Thus, the department capacity clearly depends on the ability to implement “smart” production schedules that make the best use of the available resources, i.e. the second issue mentioned earlier.

Finally, this higer capacity schedule is an example of a “good news / bad news” situation.  Although the plant is able to produce more (and presumably company revenues will go up) the downside of this higher capacity schedule is that  the department will be maintaining a larger amount of inventory in the supply chain on average.  And if there is more “stuff” in the pipeline, then there has to be the physical space to put it.  This is an important consideration if inventory has to be stored in or on particular types of storage facilities such as refrigerators or special racks.  Therefore, although it might be possible to ”buy” extra production capacity with a better equipment schedule,  it is important to realize that different schedules put more or less demand on the spatial capacity of the actual storage facilities.

Therefore, this example illustrates the third item, that increasing ouput can put stress on the plant’s storage facilities

This last scenario also shows that maximum capacity is not necessarily the same as minimum cost.  Because notice that in this scenario there is only one 4-hour setup, and thus any costs from the setup activity are averaged over a larger number of produced items.  But offsetting this savings in setup cost is the fact that with the increased WIP, the inventory costs will have gone up.

The fact that capacity can be such a difficult thing to measure, does not mean that it is not a valuable parameter to describe a given system.  What it does mean is that when any capacity value is given for a particular supply chain, it is absolutely critical to understand the assumptions that underlie it.  The fact that capacity is such a highly maleable concept, simply reinforces the fact that managing a company’s supply chain is always a delicate balancing act between competing costs and non-monetary factors.

Cluster Analysis in Supply Chain Optimization

Cluster Analysis (CA) is a versatile tool for grouping objects based on their similarity to each other.  It is applied in a wide variety of situations.  Some example applications are similarity of faces, countries, and species of dogs.  One starts by defining attributes of the individuals (countries, faces, species…) that can be assigned a numerical value, e.g. length of nose, separation of eyes.  One of several available metrics is used to define the ‘distance’ between the individuals.   Individuals are then assigned to a cluster based on the shortest distance.  CA produces some appealing results and gives useful insights that spark further investigation even in the exotic applications.

In the relatively straightforward context of grouping locations, cluster analysis performs equally well.  This makes CA a useful tool in supply chain optimization.  Designing a distribution network often involves planning of routes over regions or deciding on locations for warehouses.  One could look at a map covered with dots representing locations and draw circles around groups.  But this would be an arbitrary approach.  CA offers a way to group locations in a systematic way and speeds up the process of exploring several different versions of the clusters.  I am going to present a few examples and describe the mechanics of the CA process.

In the examples that follow we have a set of 400+ locations from an area surrounding Denver Colorado.  The latitude and longitude of the locations are the attributes that will define the similarity of locations.  In this case the distance between objects and clusters really is a distance.  In CA you can generally use Euclidean distance or XY distance – as though you were on a sidewalk in a city.  Which to use depends on the situation.  If looking at a downtown area, the XY distance might be best.  But over a broader area the Euclidean or Great Circle distance makes more sense.

There are really two broad categories of clustering – hierarchical and non-hierarchical.  The hierarchical methods begin by considering all the individuals (locations) as separate groups.  The nearest two are joined and now we have one less group.  The location of that group can be defined in a few different ways.  For now picture it as the centroid of the group members – which could be weighted by something like demand for a product.  The next join depends on the remaining distances between members or between a member and the existing clusters.  In either case we end up with one less group.  CA keeps joining until there is one large group.  This is agglomerative or hierarchical clustering.

Along the way there is a point at which we have 434, 433… 5, 4, 3, 2, 1 groups.  For any given number of groups we can look at the distance between the groups and the distance of group members from the centroid of their own group.  One could look for a small number of clusters where the distances within a cluster are small compared to the distances between clusters. 

Most of the options within hierarchical clustering are based on the way we define the distance between an individual location and a cluster and the distance from one cluster to another.  One example is nearest neighbor.  In this case the distance of an individual to a cluster is the distance to the closest existing location in the cluster.  Farthest neighbor is also a possibility.  Neither of these is well-suited to most supply chain applications.  The typical method is to use the centroid as the cluster location for defining distances.  Let’s take a look at three methods, average distance, farthest neighbor, and Ward’s method (which will be described soon).

Figure 1 – Clustering based on farthest neighbor (Complete Clustering)

In the complete method an individual or cluster is joined to another cluster based on the smallest maximum distance between any two individual locations.  Figure 1 above is what it produces.  If the groups were to be used to define clusters for delivery routes one problem might be the unequal sizes of the five groups that were created.

The average distance method joins individuals or clusters to another cluster based on the smallest average distance to the members of a cluster.  This gives another result with unequal cluster membership.  A method based on the median distance gives a result very similar to the complete method using the sample data, though it is less sensitive to outlying locations that are far from all others.

If reasonably equal group sizes are important, then another method called Ward’s method is one of the best choices.  The results using Ward’s method are shown below.  Ward’s method makes assignments that minimize the within clusters deviations from the centroid of the cluster.  What you get depends on how many clusters you ask for.  You can compare the results of five and six clusters.

Figure 2 – Clustering based on average distance to the existing cluster members.

Figure 3 – Ward’s Method with five clusters requested.

Figure 4 – Ward’s Method with six clusters requested.

Another way to look at the results is with a type of graph called a dendogram.  The dendogram gives a visual look at the way similarity of clusters changes as individuals are merged into the clusters.  The dendogram is shown in Figure 5 below.  The vertical scale is just a measure of similarity.  It is linear and is often just a normalized scale from zero to one hundred. 

What you take away from the dendogram is this:  as fewer clusters are formed, the similarity of members within a cluster decreases.  Picture a horizontal line that cuts across the figure where it will intersect just six of the vertical dendogram lines.  Notice that with just a small change in the vertical scale we could cross four or five vertical lines of the dendogram.  That implies there is not much change going from four to six clusters in terms of the nearness of individuals within the clusters.  They are still very similar (close).  Things change much more rapidly going below four clusters.  Of course, at the bottom when there are two hundred clusters, members of each cluster are very similar (read close) to each other.

The last method to look at is different than hierarchical clustering.  It is called K Means clustering.  It is not a hierarchical method so a dendogram is not really possible – joining does not occur sequentially.  Without too many details, K Means clustering works like this.  We start with an initial division into a certain number of clusters – let’s say six.  The initial division should be done in some sensible way, e.g. Ward’s method.

Figure 5 – Dendogram based on Ward’s method.

K-Means clustering keeps examining all the observations to see if they are closer to the centroid of some group than to the group they are currently in.  If so, they are moved and the centroids of both affected groups are recalculated.  The method continues until no more improving moves can be made.  It is important to start with a decent initial grouping.  The result of K-Means clustering using Ward’s method as a starting point is shown in Figure 6.

Figure 6 – K Means Clustering with an initial group.

If we had started without an initial group the result would be a bit different.  It would look like Figure 7.

Figure 7 – K Means Clustering without an initial group.

This is somewhat different than the result with an initial group. 

Bottom line recommendations are these.  Use K-Means clustering with Ward’s method as an initial seed.  If K-Means clustering is not available, use Ward’s method.  If you use a hierarchical method take a look at the dendogram as a way to decide on a number of clusters.  You can reduce the number of clusters until the reduction of similarity is large.  Looking at the scattergram of the points for various numbers of clusters should confirm the dendogram information.  Methods like the farthest or nearest neighbor will very likely give poor results.  All methods are affected by outliers to some degree.  Consider managing a few far out points by hand.  Minitab is a very good software product for doing cluster analysis.

CA is far from a mathematical curiosity.  It can be a very useful tool for network or facility location analysis.  It certainly lets the analyst explore more solutions than could be done manually.  Rather than some arbitrary decisions on grouping, CA can contribute the analysis that leads to bottom line savings.

A good friend of mine, who works for a large employer in her city, recently told me that her department’s budget, along with every other department budget that was classified as “Administration” in the ubiquitous SAP system, had to be cut by a large and specific percentage.

It didn’t matter that the “Administration” label was not uniformly applied across her organization and that some departments that were so labeled performed functions very similar to other departments that were not stuck with that label.  It didn’t matter what services each department provided, or how efficiently they provided them, they just had to cut the budget and they had to hit the number.  Incredibly, it didn’t matter that her group was one of the few “Administration” groups that actually generated revenue; in her case three times their total annual budget spend.

Unfortunately, hers is not the first story like this that I have heard.

There is no doubt that many corporations, organizations, governments and households have been hit hard by the recent economic downturn.  Each of these groups has been forced to make some difficult decisions.  So what do I have against across-the-board (ATB) budget cuts?  Basically, I think it has to be the worst way to reduce costs in an organization, and here’s why.

Let’s take a look at something that is important and familiar to all of us; the family budget.  Sadly, many families have been forced to drastically reduce spending as a result of a lay-off or furlough over the past two years.  In those cases, an ATB cost-cutting strategy just doesn’t work.  Try telling the bank that you’ve had to cut your monthly mortgage payments by 15%.  I doubt that they will be impressed when you tell them that you’ve had to do the same with your property taxes, insurance premiums, electricity and water payments, as well.  You might get lucky and be able to renegotiate your mortgage and you might get lucky if your state provides utilities assistance for people who have recently lost their jobs, but most tax assessors and insurance companies will not be particularly sympathetic.

But my guess is that you’d probably take a very different type of approach to cost-cutting in your household.  You’d probably take a hard look at all of the money that you’re spending over a month or a quarter.  You might first examine your spending to see if you could conserve on the amount you consume or if there were ways to get the same goods and services in a cheaper manner.  If that didn’t reduce your spending enough, you’d probably divide the remaining spending into different categories.   There are many different ways to categorize your expenses, but they’ll probably come down to something like, 1) Essential; 2) Non-essential, but painful to cut; 3) Non-essential and easier to cut.  If you’re lucky, you will be able to cut enough of your spending by eliminating or reducing your expenses in the non-essential categories.  If not, you might be forced to re-examine what really is “Essential.”  For example, your mortgage payment is essential, as long as you plan to stay in your house, but if the situation calls for it, you can reduce your costs by moving into a smaller home or apartment.  Not a fun choice, but it could be the right thing to do in certain situations.

Looking back on the family budget example, what did we do?  First, we looked for opportunities to conserve and less expensive ways to purchase the same goods and services.  Next, we prioritized our spending so we could make good decisions.  To find less expensive ways to purchase the same goods and services and to prioritize the spending means that we needed to 1) understand what we were getting for the money we were spending and 2) understand what would happen when we stopped spending that money.  After prioritizing our spending we made trade-offs by deciding what we could live without.  Some of the trade-offs may have been no-brainers, but some may have been very difficult.

I would argue that this is the same process that should occur in any organization that needs to reduce its spending.  It amazes me how a manager can walk into a large organization and mandate a large cut in the budget for each and every department (as they are defined in the accounting system, but that’s a different blog) without understanding where, how and why the money is spent.  It would be laughable if the results weren’t so sad.

ATB budget cuts penalize your best managers.  These are the managers that run a lean operation, who have taken the initiative to drive out all of the waste and improve productivity.  They are already doing the job you’ve asked them to do with the fewest resources possible, but they are being treated in the same manner as the manager who is either not as effective, or who has become jaded by past ATB cuts, so that he/she keeps some “rainy day” resources in the budget for just such “emergencies.”  (… and people wonder why their best managers seem to leave after these types of budget cuts, even when their positions are not eliminated.)

Let’s not forget the knock-on effect of penalizing your best managers.  The best managers often assemble the best teams to do the work.  If one or more members of a lean, highly productive, well-functioning team is forced out in an ATB cut, the rest of the team is forced to pick up the additional work of the departing team members.  This extra work, on top of an already full workload, either forces the quality of the work to suffer, or reduces the total output of the team; that is, if the rest of the team elects to stay in an organization that doesn’t value efficiency.

ATB budget cuts often fail to achieve their savings targets or result in so much “slash and burn” damage to the organization that “add-backs” must occur after the blood-letting so the organization can survive.  It continues to amaze me that these managers have the time to perform an initial ATB cut, followed by another one or by an “add back” program; but don’t have the time to do it right the first time.

ATB cuts suggest that the value of the work performed under each of the budgets is equal to the value of the work performed in all other budgets.  I have seen a lot of different organizations over my career and I don’t think I’ve ever observed this to be the case.  Take my friend’s case: her group makes money, while others spend it.  Is a cost cut that forces a reduction in revenue equal to a cost cut that has no impact on revenue?  Probably not.

So, what’s the answer?  Clearly, many organizations are forced to radically reduce costs just to survive.  I think it goes back to our home budget example: 1) know what you’re spending; 2) understand what you get for it, 3) find ways to get the same or similar things for less money, and 4) make the hard choices about what you can do without.

In the end, my experience has been that managers who drive ATB cost reductions are incapable/unwilling to understand their business processes and organizations sufficiently; lack the imagination or skills to reengineer their business processes; or lack the courage to make the hard choices about what their organization will do and what it won’t do in the future.

To all those top level managers who have instituted ATB cuts, or for those who are planning to do so: Don’t do it! Think before you act, and save your company the added burden of bad management.

At Profit Point we are in the ‘Science of Better’, and we are always looking for new ways to do business, both for our clients, and for ourselves.  When we started, we had the challenge of being a virtual company, that is, we have never had a corporate office space. Since 1995, each of us has always worked from home.  While there are numerous benefits of this style of company architecture, including having a family that actually knows who you are, and keeping the company’s overhead to a minimum, it also has its drawbacks. Like forcing each person to make the deliberate decision about when to start work, and harder still, when to stop work each day. We knew when we started this company that we wanted to keep our overhead costs low, so a virtual office seemed like the natural choice.

More recently, we have been faced with another challenge, how to reduce the cost of the projects we do. Projects in the supply chain business require a certain amount of industry and company specific knowledge.  Until recently, we had been building into our projects ample on-site time where the project team could gel and collaborate and build the trust that is needed for the free flow of ideas.  But the world has changed, and we have changed with it.   No longer are big travel budgets a normal part of the projects we see. So the challenge was: how to reduce the travel expense line item, without sacrificing the project speed or quality?

In the consulting business, there is sometimes no substitute for ‘face-time’.  So travel to the customer site perforce happens.  Over the course of the last 15 years, I have seen a marked drop in the amount of time that we need to travel, going from 60-70% a decade ago to less than 20% currently, and this has been brought about primarily by two factors: 1) Companies simply do not want to pay the travel expenses. Since 9/11, most major companies have been slashing their travel budgets, and expect their consultants to follow suit.  One particular project comes to mind where I had seen travel expenses that were as much as the consulting bill each month. But in general, we see pressure to reduce the travel expenses that are generated by projects across the board.  2) ‘Remote Touch’ Technology has provided the means to travel less.  There are some great remote desktop control tools that allow two or more people to have a telephone or VOIP conversation, and look at the same computer screen, to discuss and collaborate on ideas and tools.  These web based telephony and remote control tools have eliminated the need for travel to a greater extent than you might think.   Many of our projects today have only two face to face meetings, one to kick it off, and one to present the results or close it out.  Some of our clients are handled successfully without any face time. I must say though, that in our experience, low face-time projects only work well within the culture and language: that is, when language and culture barriers exist in the project team, face-time is the best way to bridge these gaps, and mitigate the risk of project overruns and delays.

In business, technology comes into being as a means to enable better business processes.  The processes that we use that are enabled by this remote touch technology includes an agile approach to solving business problems or developing software solutions.  We use several readily available web based tools every day in our business, and boy have they allowed us to reduce the travel expenses.  These include:

GotoMeeting.com

This is the best remote touch tool out there in our opinion.  Until a robust free app comes along, this will remain the best value for the money.  The best part of the app is the recent addition of the integrated VOIP, where you can use a head set (I would recommend the Logitech ClearChat PC Wireless Headset: http://www.logitech.com/en-us/webcam-communications/internet-headsets-phones/devices/4226) to join the integrated telecon line.  This has the advantage of freeing up your phone, and being instantly connected to the telecon as soon as you start it.  No more long telecon numbers with their passcodes! We use this many times every day, and it is the primary reason why we can travel less.

Box.net

This is a simple to use and secure web based file storage and sharing application that fosters and supports collaboration with people both in your company and externally.  We love this app, and my clients seem to as well.  Just drop a file into this app, and share it securely with anyone with an email address. Use it when email attachments just will not do, due to size limitations, or just when the email hassle is too much.

PivotalTracker.com

This is a terrific project management tool that is designed for agile projects, and makes it simple to create and manage user stories for tool development.  While inviting new members can be a hassle (since their email seems to get caught in many spam filters), once they are in, these folks have made a stellar user interface to manage the tasks in a project of nearly any size. Use it to track bugs too. We have done several projects using this tool. and we will be using it for many more.  Great tool.

TableauSoftware.com

If you like to look at data, like we like to look at data, then you will want to look at Tableau.  You can think of it like a pivot table / chart on steroids.  You open it, connect to you data, (whereever or what ever data you’ve got, it can connect to it), and then you start to explore your data like you’ve never been able to before. Like a pivot table, you can drag and drop fields, aggregate data along dimensions,  and make sums, etc, but the really cool part of Tableau is the part where it suggests new ways of the looking at the data.   Go ahead, make maps, heat charts, time phased graphs, whatever.  Then you can assemble the graphs into a dashboard. Dashboards are the best.  Want to see a ton of data distilled down into a very compact visually stunning view suitable for management? Get a copy of Tableau, and you can make that view in minutes.

Used appropriately, these tools, and others like them, have enabled us to travel less, and work faster and better. (and more!)

If you have other great apps like these that enable better business processes, I would love to hear about them.

Did you miss Jim Piermarini talk at CSCMP about Logitech’s supply chain distribution methodology? For those that are interested, we are posting the slide deck here for your review. To download the complete presentation click the image below:

To learn more about how Profit Point can help you improve your inventory fulfillment, contact us here.

I’m a picture guy.  In our kind of work, we have to be able to take a lot of data and make sense out of the process or processes that generated it.  I used to work with a fellow named, Bill, who has a PhD in Operations Research, and is probably one of the smartest people I’ve ever met in my life.  Bill is a guy who can look at six or seven big tables of numbers and then say something like, “… and the answer is 7.563.”  He was usually right.  I don’t have that talent to create the linkages among lots of different types of information in my head to come up with a conclusion like that.  That’s why I like pictures.

Recently, one of my colleagues and I were visiting a manufacturing plant to assess their production scheduling process.  The client invited us to visit the plant because they knew they had a problem.  As we followed the scheduler through his day, we began to understand the root causes of the problem.  So how did I choose to communicate what we’d found to the client?  You guessed it; I drew a picture.

When the plant manager first opened the file containing the flowchart of their existing process, she told me she only needed to see that it took me three letter-sized pages to document to the process to know that the process was much too complex and cumbersome to be fixed with a couple of “quick hits.”  Why is it that she knew without studying the details that we needed a full redesign to fix this process?

I think many of us are just built that way.  I know there is a lot of clinical and academic research that shows how we human beings use our sense of sight as a first preference for observing the world, and that there are specific parts of our brains that are able to detect visual patterns or the lack thereof.  However, I don’t think we need to see the results of that research to know why the phrase, “a picture is worth a thousand words,” is such an enduring statement.  It rings true with all of us.

That’s why I like a software product called Tableau.  It is marketed as a visual analysis tool and I think it does its job quite well.  Although I don’t claim to be an expert user, I have found it quite useful when I need to understand what’s going on in a large dataset.  Let me illustrate using an example from a recent transportation analysis that we did for one of our clients.

Our client had grown by acquisition and managed its transportation in a very de-centralized manner.  Each of the sites contracted individually with their own set of carriers, using their own set of criteria for selecting and then awarding business to the carriers.  Profit Point was called in to help the client understand the cost-savings opportunities that would result from a more centralized approach to carrier contracting and management.

Our first priority was to find out what was going on at all of the different sites so we developed a database from the client’s freight payment records to do it.  Now, picture this (pun intended).  We now have over 63,000 individual shipment records to analyze and we needed to do it in a way that told a story that we could understand and that we could then communicate to the client.  The first thing we did was look at the spend by plant and by carrier.  The spend by plant was more of a prioritization issue, to understand which of the plants had the highest freight spend, but the spend by carrier became the first part of our story as you can see in the two pictures below.

This second chart was a very powerful image to help the client quickly see that the number of carriers being employed was out of control.  You don’t even need to be able to read the name of the carrier on the Y-axis to know that there are too many carriers in this picture.  Many of these carriers had only a single load all year long, but were still carried in the system.

We also wanted to show the client the significant different in pricing policies across their carrier base.  The following slides show how we used some more of Tableau’s functionality to make our point.

By plotting cost vs. distance for all of the shipments, we were able to see the general correlation of cost with distance that we expected, but we also saw a number of outliers that we wanted to better understand.

We then highlighted a group of very high-cost shipments and kept only those points to see what we might find out.

Using a simple stacked bar chart, it was very apparent that carrier “C-g,” the red bar in the chart at left, was the main player in this group.  Once “C-g” was identified, we were able to demonstrate that their cost was always greater than the average cost for shipments with distances greater than 200 miles and by as much as 50-66% for shipments with distances greater than 1000 miles.

Again, these pictures allowed us to find one of the smoking guns inside this mass of data.  Suffice it to say that we found many other opportunities through similar visual analysis.

Because of these pictures, and others like them, it was an easy sell.  Using a tool that makes it easy to use the built-in “intelligence of our eyeballs,” we were able to develop a convincing call to action for our client, who went out to the market with a targeted freight bid and reduced their transportation spend dramatically.

As technology continues to penetrate more and more aspects of business and our everyday lives, it makes more and more data available for us to turn into useful information.  But it’s only useful information when we can put it into a form that we understand and can communicate it to others.  That’s why I’m a picture guy.

To learn more about Profit Point’s transportation services, call (866) 347-1130 or contact us here.

Despite our egalitarian mindset in the U.S., when it comes to customers, let’s face it: They have never been ‘created equal.’ Certainly for decades, manufacturers and distributors have offered better pricing to some customers than others. We’re all familiar with quantity break pricing, column pricing with different discount levels for different categories of customers, and contract pricing. And who doesn’t visit the local supermarket today and notice the ‘buy 3 get 1 free’ offers to encourage us to increase our purchases?

Volume is valuable and warrants better pricing, we are in the habit of believing. And most often this is true. Not only does a high-volume customer drive our buying power with suppliers by helping us reach the next price break level on the purchasing side, but it can make each sale more profitable: The cost of servicing 10 orders that result in a sale of 100 units can be 10 times as great as the cost of servicing a single order for those 100 units.

This bias towards volume underlies traditional customer ranking methods. But many manufacturers today are taking a closer look at these policies and finding them lacking. Instead, they are engaging in a detailed cost analysis effort called ‘cost-to-serve.’ While cost-to-serve can be a very broad subject covering product costs, location costs, transportation costs and service costs, to name a few, this article will take a look primarily at customer costs.

It’s not that heretofore companies have ignored factors that shade the degree of profitability of a large client. Many firms, presented with the opportunity of doing business with, say, Wal-Mart or the federal government, may question whether it’s really worth doing. They’re thinking about the overhead of handling such a client and the cost of meeting client demands – with slim price margins.

What’s different today is that companies are trying to measure these costs precisely and to make informed, scientific decisions based upon them. Whether they engage consulting firms who have developed methods for tackling this measurement, purchase software to help them out, or devise their own internal approach, more and more manufacturers and wholesalers are gathering detailed costs and trying to apply them to decisions about their customers.

Consumer goods companies, for instance, are recording metrics such as the true cost of customer service. How much support time does this customer require of the customer service organization? How much sales time to we devote to him? Does the customer frequently return merchandise, and if so, what is the cost of processing that return? In the case of consumer goods manufacturers, we might also look at custom-branded merchandise: What is the true cost of providing private labeling for a retailer? Are we really capturing in the product cost all of the special handling required by the purchasing and distribution organizations? All of these costs are very important is assessing a customer’s true profitability.

On the other side of the equation, there may be some sales and marketing benefits that a customer brings, and these, too, should be weighed. Does the name ‘Wal-Mart’ on our client list provide positive benefit to the organization? Is another client who doesn’t seem to purchase very much an outstanding reference for us who sends other potential customers to us? If a business can establish a process and gain agreement across the organization on measuring true costs and benefits, it can define policies to more precisely control bottom-line revenue.
Certainly, one of the first decisions that can be made, once true costs are measured and accepted by an organization, is to eliminate customers who are really unprofitable. But cost-to-serve can also come into play in other ways. We may want to devise strategic programs that nurture our best clients to safeguard their business. We may hold special events for them or assign dedicated reps, for instance.

One of the situations where cost-to-serve becomes a critical tool is in inventory allocation, particularly in an inventory shortage situation. When there is insufficient inventory to meet demand, most manufacturers will want to serve the most valuable customers first.

This frequently comes into play in segments of the technology industry, such as computer peripherals, typically with the launch of a popular new consumer product. An extreme example of this might be the launch of a new Wii game player at the start of the holiday season. Armed with true cost-to-serve data, manufacturers could make allocation decisions scientifically to spread the available inventory across the order pool while maximizing profit.

You might ask whether this process can be automated today. The answer is ‘partially.’ Allocation can certainly be automated, but collecting cost-to-serve data on customers usually involves some manual steps, because most companies don’t have all the systems in place to collect this data automatically (and even with sophisticated systems, the data may not be collected in exactly the way you wish.) Some spreadsheet work may be required. Once the spreadsheet is in place, however, the process becomes straightforward.

Perhaps you want to rank customers sequentially from top to bottom, or group them into ‘profit’ segments. Once that is done, an algorithm can be designed to optimize the allocation of inventory according to the rules tied to those rankings or segments. The allocation algorithm might be designed to work directly from the spreadsheet, as well, automating even more of the process. In any case, executing the service decisions in accord with true costs ensures we are protecting our most valuable customers.

The application of cost-to-serve to inventory allocation takes on an even more interesting aspect for consumer goods manufacturers who ship to retailers. As those of us familiar with this industry are aware, most large retailers have very specific guidelines defining how suppliers must do business with them. The retailers specify how an order must arrive – shipped complete, packed by store, etc.; when it must arrive – ‘arrive by’ date; and a variety of paperwork details including design, content and placement of shipping labels and bills of lading. Associated with each of these requirements is a dollar penalty the supplier will incur, taken as a deduction from the supplier’s invoice, for violation of the guideline.

For a consumer goods manufacturer, these penalties or ‘chargebacks,’ can mean the difference between a profitable client and an unprofitable one. In this situation, the ability to allocate inventory defensively, to minimize chargebacks (or at least make an informed scientific decision to incur them) is critical. A powerful allocation engine, in an inventory shortage situation, can maximize profit by factoring potential chargeback costs for late or partial shipment into the equation. In this case, the allocation engine ensures that the cost to serve the retailer is as low as possible.

In addition to retailer penalties, another aspect of ‘allocation-according-to-true-cost’ involves inventory fulfillment location choices. If a company operates a single distribution center in Los Angeles and imports all its product from Asia, there may be only a single fulfillment option. But for the wide majority of consumer goods manufacturers who import from Asia, service clients nationwide, and operate either multiple distribution centers or a distribution center located in, for instance, the Midwest, there are several options and a variety of questions
arise.

If inventory is constrained at the facility that would normally handle a particular customer’s order, should the order be fulfilled from an alternate facility? To make this decision, we need to factor in not only the additional shipping cost but also to weigh that cost against the value of the customer. There may be low profit customers, viewed from the perspective of cost-to-serve, for whom we do not want to make this investment. In the case of a retailer where a potential penalty is involved, the decision might be made dynamically based on a comparison of the chargeback incurred against the additional cost of shipping. If the chargeback fee would be higher than the additional shipping cost, it may be worthwhile to use the alternate distribution center.

This type of on-the-fly fulfillment decision is often called ‘dynamic allocation.’ Another example of dynamic allocation involves intercepting shipments in transit to, say, our hypothetical Midwest distribution center. Least cost fulfillment might dictate fulfilling west coast orders by pulling off inventory required to fulfill them at a deconsolidation facility near the port – before a shipment heads out to the distribution center in the Midwest. Under what conditions is this the least-cost choice? An inventory allocation algorithm based on cost-to-serve can make this decision mathematically, using rules the manufacturer defines.

It’s important to emphasize that the decisions on exactly how to apply cost-to-serve data to inventory allocation will depend on the philosophy of the individual company. For this reason, such allocation solutions are often unique and are adjuncts to the standard capabilities of order management systems. Leading-edge firms who are structuring allocation based on true costs typically do so via point solutions that supplement their central transactional systems.

Profit Point, as the name suggest, provides these point solutions and integrates them into SAP, Oracle, and other order management systems to help clients make the best, most profitable allocation and customer decisions. Our expertise in this area can help clients drive maximal profit to the bottom line.

This article was written by Cindy Engers, a Senior Account Manager at Profit Point.

To learn more about our supply chain data integration and business optimization services, contact us here or call (866) 347-1130.

When I worked as a Branch Chief in Air Mobility Command’s Analysis Group my boss often talked about nailing the Jell-O to the wall. I took that to mean we had a project with some ill-defined performance measures and objectives. Our first task was going to be establishing the goals. He was a crusty old colonel who was usually dead right on these matters.

The phrase stuck with me, but its meaning has probably evolved a bit. Working with our clients we frequently create optimization models where the objective is a mixture of terms – some that are concrete and some that are ill-defined. The user has certain costs that are very real but also recognizes that a good solution has other desirable qualities. Sometimes these additional criteria are easily expressed as constraints. But just as often they have to be weighed against the concrete and known cost terms in the objective function.

As an example, we have a client that must pay to move heavy equipment around the country from one customer site to another. The model output is a schedule of the required movements for each piece of machinery. Any given solution will have values for a few performance measures. The cost per mile of moving the equipment is easily measured and is known. Although this cost might vary during the year and is subject to some uncertainty, it is still well-defined. We can get solutions for different values of this cost parameter when we do sensitivity analysis. This is the most directly and easily quantified performance measure for a solution.

On the other hand, there are some other factors that enter into the objectives and are harder to value. For instance, risk is a factor we add to the model to give some latitude in meeting deadlines for the arrival/departure of the equipment to/from locations where it will service a user contract. Clearly, we want a solution with little risk. Upgrades are another such factor. At times, equipment of higher quality than contracted for must be used to meet contract dates. Again, this is something to be avoided since there is wear and tear on the expensive machinery. Finally, there can be occasions when an appropriate configuration of equipment is just not available. Meeting the contract requires leasing equipment at substantial cost.

None of the additional factors is well-expressed in a constraint since there are no absolute limits. The model we use influences these performance measures via the cost terms in the objective function we are trying to minimize. The trick is to assign the right costs to get solutions that the user can recognize as good. In cases like these, Experimental Design applied to the optimization model is an efficient way to derive a useful set of parameters.

Experimental Design (DOE)
Design of Experiments (DOE) has long been used in industrial settings to quantify the way parameters affect product design. Taguchi methods and Robust Engineering are the most popular names for this approach. Inferior methods based on intuition or one-factor-at-a-time experimentation have been thoroughly discredited and largely driven out of practice by DOE approaches.

Applying DOE to optimization and simulation modeling is not a recent development, either. In the machinery example, the solution (a schedule with routes) is essentially a product we are designing with the help of the model. In the typical application DOE is used to summarize the behavior of the model with respect to resource allocation. Essentially, we see how much benefit is derived from the addition of resources. In this case we use DOE to see how the shape of the solution changes as costs are varied. The costs represent the relative importance of the various aspects of the solution. Some are concrete others really are not.

Returning to the machinery movement example you can probably picture what happens with some extreme values of the costs. If substitution and upgrades are free, then you get a solution with a small number of miles. But you find customers who paid for low-tech equipment getting a lot of free upgrades. This is not desirable for other reasons besides wear-and-tear. Those customers that paid for the best can feel abused and it hardly motivates the lucky customer to pay full price for the best equipment next time around.

Also, an overly conservative approach to meeting contract deadlines means that the company must buy more equipment than it really needs. So a very high penalty for taking risk, e.g. days allowed to reposition equipment, can be very expensive. The start and end dates for the contracts shift during the course of the year. A little risk is not entirely a bad thing — especially since the allowed risk is easily controlled in the model.

Certainly what we really want to avoid is a model that recommends inferior solutions. In a two-dimensional example that would be a solution that had the same risk value but more miles associated with movement than are needed.

Numerical Examples
Realize that as time moves on the equipment example is a model that must be rerun repeatedly. New commitments are made and existing ones may have been modified. Those solutions that have attractive summary performance measures are the ones that should be examined in detail. Ones with inferior solutions can be safely ignored. In this application the cost settings seem pretty stable – produce good solutions. But there is nothing to prevent rerunning the DOE periodically to examine solutions based on their summary performance measures.

The table below (click to enlarge) shows a set of results from a point early in the year for a subset of the equipment to be scheduled for the coming year. All of the values have been changed, but patterns have been preserved for illustration purposes. In this example we chose just three of the possible six (shown in yellow) parameters that could be varied. We were able to run all eight possible combinations of the values shown. We could investigate a larger set of parameters, perhaps by using fractional experiments that look for important effects without doing all the possible runs.

Without actually estimating the effects of the parameters you can see at a glance most of the effects. High upgrade costs lead to fewer upgrades, high risk costs lead to smaller number of risk days. High costs per route lead to a set of routes with larger total miles. One could argue that run one or run three yields a schedule which is a good tradeoff among the performance measures we considered here.

So a user would be encouraged to take a look at the details of the solutions (scheduling) from Run 1 or Run 3. Finally, notice that there is an interaction between RouteParm and RiskParm on the RiskDays performance measure. That is, changing the RiskParm value has a different effect on Risk Days depending on the setting of RouteParm – much more influence when RouteParm is large than when RouteParm is small. Another benefit of DOE is the ability to spot those kinds of interactions. One-at-a-time experimentation is completely incapable of finding this sort of information.

"http://profitpt.s3.amazonaws.com/wp-content/uploads/old/chart1-799962.png">In the heavy equipment case we had some ideas on the neighborhood to be investigated for the parameter settings. That is not always the case. Sometimes the parameters are very difficult to calibrate by reasoning from or comparison with costs that are known. For instance, we have an example from agriculture. Here the model suggests the order in which a group of farms will be harvested. Processing plants want the products in specific weight ranges and require amounts of each product for each day over the planning horizon.

Besides getting within the desired weight range there are some other goals. For instance, it would be good to hit the weight targets – not simply be within the range. We also want to reduce the number of trucks that have to be dispatched and the number of miles driven. We also want to visit individual farms a limited number of times. Furthermore, if product is not harvested while in an allowed range, it is ‘wasted’ and this is one of the most costly penalties – almost an absolute requirement. Realize that as time goes on the product is always growing and so there is a limited time window on the weight range.

Not all of a given the product on a farm is of the same weight. It varies from section to another based on the time it started growing. So in a simplest of experiments you can test the tradeoffs between distance traveled and missing the weight targets. If you are willing to visit several times, you can come closer to hitting the target weight. But this will involve more visits and usually more total mileage.

Predictions
If one goes the extra step of fitting a statistical model to the DOE results, then predictions can be made for the performance measures. When multiple performance measures are involved, one can locate regions where all performance measures are in acceptable ranges. Various statistical packages, e.g. JMP or Minitab, provide the ability to create two-dimensional contour plots that show regions where both dimensions are within ranges defined by the user.

The three figures above are examples of contour plots. Given that the experiment only examined two levels of each factor, none of the plots can really show curvature. But you can see an interaction in the Miles graph.

Conclusion
Not every model requires an experimental design. But in cases where there are multiple performance measures that need to be combined using (possibly) arbitrary weights, DOE is an excellent approach. This is true for simulation and optimization models alike. Furthermore, DOE can be used as a predictive tool. If the design is carefully chosen, it can guide you to a useful operating region and reveal interactions between the various factors.

So how do you nail Jell-O to the wall? You throw a Design of Experiments net over it.

This article was written by Joe Litko, Profit Point’s Business Optimization Practice Leader.

To learn more about our business optimization services, contact us here or call (866) 347-1130.