Section I. Demand from the Customer Is Too Variable


I. Demand from the Customer Is Too Variable

Overview

When a Customer's demand for entities is entirely unpredictable and highly variable, it often becomes very difficult for a process to satisfy the demand. Spikes in demand are usually dealt with using buffers of inventory (stockpiles of entities), time (lengthy delivery times), or capacity (excess resources available). None of these buffers present an ideal solution.

Examples

  • Industrial. Customer ordering patterns

  • Healthcare. Outpatient bookings, surgery/emergency department demand

  • Service/Transactional. Customer ordering patterns, customer usage rates

Measuring Performance

Variation in demand is typically measured as the Coefficient of Variation (COV) of demand, defined as


where S is the standard deviation of demand (variability) and is the mean of demand (volume). Both are measured in unit numbers of entities, not dollar value.

Entities with a COV less than 1.0 are considered to have smooth demand. Any entities with a COV above 3.0 are considered to be highly variable. For more on COV, see "Demand Segmentation" in Chapter 7, "Tools."

Tool Approach

First we must get an understanding of the current performance with respect to variability in demand:

Focus should just be on measuring validitya sound operational definition and consistent measure of COV versus a detailed investigation of Gage R&R. See "Demand Segmentation" in Chapter 7 for more detail on capturing COV data.

The data is typically captured over a period of one month to one year (depending on process drumbeat) to get a reasonable estimate. Historical data will more than likely be sufficient for the purpose.

Caution: COV data is for what a Customer requested (demand) versus what we decided to process (operations planning).

The intervals used should represent typical demand intervals from the downstream Customer. For example, if a Customer orders entities monthly, subdivide the time into monthly buckets. See "Demand Segmentation" in Chapter 7 for more detail on capturing COV data.

Map the Supply Chain with a high-level SIPOC to understand supply stream linkage issues with external processes.

This will allow us to understand the volume and variation in demand of each of the different entity types that progress through the process. High-volume, low-variation demand entities do not present a problem in this categoryonly the low-volume, high-variation demand entities are the issue.


After the Demand Segmentation is complete, it should be easy to see the culprit entity types, specifically in the bottom-right corner of the graph. The first question ought to be "Should these entity types be offered at all?" Entities in this corner cause problems in our process and are very low volume. They should be in our portfolio for a reason: either because they are very high-margin entity types or they are strategic in some way (perhaps they are a new technology that is just being introduced). Any others have obvious cause to be removed using rationalization:

If there are too many entity types, we should consider rationalizing our portfolio somewhat. To do this, go to Section J in this chapter.


If the entity types are valid and cannot be rationalized further, we need a more detailed understanding of the culprits in the bottom-right corner of the Demand Segmentation graph using Demand Profiling:

The Demand Profile plots demand volume over time and gives insight into the pattern of demand as it impacts our process. For the culprit entities, plot a Demand Profile for the Customers' ordering (demand) pattern and speak with the sales representative (or equivalent) to understand how the Customers' ordering process works.

As an optional step for large-volume Customers only, it might be worthwhile to add a mapping step at this point for the Customers' ordering process.


For the key culprits, there typically will be a single Customer driving the high variability (multiple Customers tend to smooth demand by the nature of the Central Limit Theorem[3]). There are exceptions to this, such as massive seasonal changes or singular events (the impact of the Super Bowl on beer sales), but these are typically well understood and predictable and the Single Customer Effect is usually the strongest.

[3] Most university-level statistics books give an explanation of the Central Limit Theorem. One text I regularly recommend is Statistics for Management and Economics by Keller and Warrack (South-Western College Pub, ISBN: 0534491243), which explains the theory in a practical, readable way.

The Problem Category is aptly titled "Demand from the Customer Is Too Variable." The obvious question at this point (or hopefully earlier!) is "Too variable for what?" Are there missed deliveries? Perhaps the Process Lead Time is too long? Maybe the variability is affecting the ability to forecast? Is it that large levels of inventory have to be kept on hand to buffer the variability? Before continuing, the Team should certainly be able to articulate this because the next steps will probably involve interaction with Customers, and they'll need a good story!

Use Demand Profiling again, but this time on the customer usage pattern for the entity, as opposed to ordering pattern. Quite often, customer usage is smooth, but the Customer will make large batch requests of us, the Supplier. Obviously, this will take some interaction with the Customer to achieve.

This is best-case scenario, but only if we can take advantage of that smooth usage in our process. Options include

  • Altering the demand quantities and frequency

  • Installing "blanket orders" and "call off" whereby Customers agree to order a larger amount over a long period of time, but take delivery in regular, smaller quantities, paying as they go

  • Investigating the triggers in the process to find options of earlier warnings of demand

  • Utilizing a Pull System with the Customer. For more detail see "Pull Systems & Kanban" in Chapter 7.

For whichever solution is selected, implement the solution and move to the Control tools in Chapter 5.

This is tougher. Options include

  • Reducing the Process Lead Time, in which case go to Section G in this chapter.

  • Working to increase warning time or trigger from the downstream Customer (can we know any earlier?).

  • Looking at the demand profile of the Customer's Customer and identifying opportunity (this will only work if you are considered to be a Strategic Supplierthat is, the Customer wants to listen to you).

  • Consolidating the entities across a number of sister sites in the company, so only one site deals with this specialty request and thus sees higher volume (and typically less variability, accordingly).

  • Constructing a line (process area) that specifically deals with the low-volume, high-variability entities. As such, this area would need rapid changeover ability, a responsive custom approach, and so on.

  • Holding inventory buffers at key locations throughout the process, so entities don't have to traverse the whole process to be delivered (only works if the entities aren't unique).

  • Examining options to keep the entity "vanilla" longer (not customized to a particular Customer). Is it possible to pre-prepare any parts of the entity so less work is required from the point of request?

  • Moving to a Platform Technology approach or modular entity.

  • Considering a different process technology.

    For whichever solution is selected, implement the solution and move to the Control tools in Chapter 5.





Lean Sigma(c) A Practitionaer's Guide
Lean Sigma: A Practitioners Guide
ISBN: 0132390787
EAN: 2147483647
Year: 2006
Pages: 138

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net