Agile Product Development For Mass Customization, JIT

by Dr. David M. Anderson, P.E., Management Consultant, Lafayette, CA

This article is derived from the book, Agile Product Development for Mass Customization, Niche Markets, JIT, Build-to-Order, and

Flexible Manufacturing, by David M. Anderson, with an introduction by B. Joseph Pine II (McGraw-Hill, 1997).

Developing products for the agile paradigms is the ultimate, and most challenging, application of concurrent engineering. In many companies, concurrent engineering means no more than "keep manufacturing in mind" when developing products. In agile product development, product development teams must concurrently develop flexible processes that are so agile that companies can quickly develop a broad portfolio of niche market products, build products to order, mass customize individual products at mass-production speed and efficiency, and rapidly introduce a steady succession of "new" products that are really planned "variations on a theme" based on common parts and modular product architecture.

This article will show how to develop products, and concurrently, how to eliminate setup by design, how to standardize parts, materials, and processes, how to fabricate parts for the agile environment, and how to design for agile assembly.

The phrase "Agile Product Development" can be interpreted in two ways, both of which are correct and applicable to a wide range of products and industries:

1. An agile product development process that can rapidly introduce a steady succession of incremental product improvements which can be called "new" products — that are really planned "variations on a theme," based on common parts and modular product architecture. This capability results in ultra-fast time-to- market, much faster than possible with independent products that do not benefit from product-family synergies in design and manufacture.

2. Development of agile products that can be manufactured in the following agile environments: Agile Manufacturing, Just-in-Time, Build-to-Order, and Mass Customization.

Few products today can be manufactured like mass-produced Model T's, without any need for product variation or customization. Most manufacturers are forced to customize products, to some degree, for increasingly selective customers or to compete in niche markets. However, if products are not designed well for this and manufacturing is not flexible enough, then

The next paradigm, after the century-old Mass Production, is Mass Customization, ¹ which is the ability to design and manufacture niche market or customized products at mass production efficiency and speed.

This is accomplished by applying the agile product development techniques described herein and implementing agile manufacturing.² Mass customization is a proactive approach that anticipates the range of customization and designs it into product families and production system. This replaces the reactive approach where sales "takes all orders" and then Engineering responds by modifying previous products with change orders and Manufacturing jumps through hoops to get them built.

To be agile, companies must be able to understand customer needs, quickly adjust to changing market opportunities, develop products rapidly for fleeting windows of opportunity, build products quickly in response to those needs, and mass customize products when appropriate.

As a first step, companies can improve their responsiveness by implementing Just-in-Time (JIT) programs to reduce setup, batch sizes, and WIP (work in process) inventory. A JIT study of 1,165 companies, reported in the APICS journal, Production & Inventory Management Journal, showed throughput time decreased an average of 59.4% as a result of implementing JIT.³ A Coopers & Lybrand study of ten clients that it helped implement JIT showed an average lead time improvement of 83.1%.⁴

However, it may be difficult to implement JIT for existing products if products have been designed with too many different parts and raw materials to distribute at all points of use. This is even more important for very agile manufacturing environments which do not rely on MRP systems to procure and deliver parts, instead using automatic replenishment systems like kanban or breadtruck deliveries. This article will present very cost-effective methodologies to standardize parts and raw materials.

A cornerstone of JIT is setup reduction which allows batch sizes to be smaller. It may be difficult to implement setup reduction for existing products if their designs are inflexible. If there are too many parts to be distributed at the points of use, then companies are forced to perform the setup called kitting. Kitting is the practice of gathering all the parts for the assembly run and putting them in a "kit," which is then delivered to the assembly area. Designing products around standard parts can eliminate this setup.

Concurrent engineering of product families and flexible processing can eliminate setup by designing versatile tooling and fixturing that can accept entire groups of parts without setup. Tool plates and fixtures can be designed to be versatile enough to accept a common blank which then can be machined by computer controlled machine tools to produce many shapes. The blank must quickly be positioned in the fixture without any need for measuring and manual positioning. Thus, the blank must be designed with common fixturing geometries. An example of this principle in electronics involves designing all printed circuit boards in the same "form factor" (size and shape) with the same tooling holes so that all of them use the same tooling plate, pallet, or material handling devices. Some companies build multiple circuit boards into panels just to keep the processing size standard.

Workpiece location setup can be eliminated with standard location geometries on parts and tooling. Computer Numerical Controlled (CNC) machine tools can eliminate extra machine loadings and setups by performing several operations on the same machine.

Agile companies will also have to develop products quickly to respond to rapidly changing market conditions. Agile enterprises can employ ultra-fast time-to- market techniques to design "new" products with most of the design work coming from previous engineering, software code, modules, and parts, instead of "reinventing the wheel" every time. Parametric CAD can accelerate any new engineering by creating part drawings with "floating" dimensions which can easily be changed for subsequent parts. The changed part drawings can, in turn, update assembly drawings. Parametric CAD can also automatically update analyses of stress, motion, vibrations, and thermal behavior.⁵

Ultra-fast product development depends on developing versatile product family architecture that optimizes versatile modularity, utilization of standard modules and parts, maximum use of off-the-shelf hardware, design by suppliers, and easy-to-modify parametric CAD templates. To fully capitalize on this ultra-fast product development, agile enterprises must be able to get new products into customers' hands rapidly. Build-to-order helps this happen by shortening the "pipeline" to the customer and eliminating the finished goods inventory from the pipeline. Traditional manufacturers have to empty the pipelines of older products before new products can flow to customers. Rosendo G. Parra, Group Vice President of Dell Computer Corporation, summarized the advantage of built-to-order for Dell: "We were probably the first vendor to transition into the new Pentium FPU processor, simply because we didn't have a hundred and some days of inventory out in distribution that we had to move first."⁶

Standardization is a key prerequisite to any of the agile paradigms.⁷ The challenge for the agile enterprise is to reduce the internal variety to the point where products can be built flexibly without the cost and time delays of setup changes. This is accomplished by standardizing on parts, features, materials, and processes in the design stage.

To operate flexibly in high volume, parts must be common enough to be readily available at all the points of use, fixturing geometry must be common enough so that all variations of a part can be positioned the same, design features must be common enough to use the same manufacturing tools, materials must be common enough to avoid stopping production to change materials. Standardization is easy to accomplish in the design stage and standardization efforts can quickly pay for their investment in reduced material overhead and purchasing leverage. However, the parts, materials, tools and processes must be standardized before product development teams begin their designs. The simplest form of standardization is simply to make it clear which parts are already in use. This helps to stop the typical proliferation of parts where "new" parts are added that are identical or similar to existing parts.

Many parts lend themselves to listing in a logical order as examples show in Figure 1. For these parts, simply list all existing parts in order, circulate the lists to the design community, and encourage engineers to use existing parts whenever possible. A significant standardization effect can be obtained by identifying the "high-runner" parts that are used over a certain quantity per year. A procedure will be presented below for determining preferred status for parts.

Part Type	Listing Order
Threaded fasteners	Thread diameter, pitch, length, head type, material/coating, grade
Washers/spacers	O.D., I.D., thickness, material, finish
Gears	Pitch, number of teeth, face width, material
Gearboxes	Ratio, horsepower, shaft orientations, shaft diameters
Motors	Horsepower, voltage, phase, shaft diameter, mount
Pumps	Pressure, Flow rate
Power supplies	Output voltage, wattage
Resistors	Ohms
Capacitors	Microfarad
Integrated circuits	Generic numbering system (e.g. 74F00)

The following methodology is an easy-to-apply approach that can be quite effective at reducing the number of different parts (part types) by standardizing on certain preferred parts.⁸ This usually applies to purchased parts but it could also apply to manufactured parts. Part commonality methodology is based on a zero-based principle that asks the simple question: "What is the minimum list of part types we need to design new products?" Answering this question can be made easier by assuming that the company (or a new competitor) has just entered this product line and is deciding which parts will be needed for a whole new product line. One of the advantages of new competitors is the ability to "start fresh" without the old "baggage" of too many parts. Just imagine a competitor simultaneously designed your entire product line around common parts. Now imagine doing the same thing internally. This is the zero based approach.

The commonality approach determines the minimum list of parts needed for new designs and is not intended to eliminate parts used on existing products, except when the common parts are functionally equivalent in all respects. In this case the common part may be substituted as an equivalent part or a "better-than" substitution, where a common part with a better tolerance can replace its lesser counterpart in existing products.⁹

The first step in the commonality approach is to ascertain the company's inherent part commonality for each category of parts. It would be very unlikely if all existing parts had identical usage histories. Every category of parts has some "high-runners" that are used far more often than others. Rather than choosing arbitrary values for new parts, choose values that correspond to existing high usage parts as much as possible. One reason for this approach would be to avoid adding new common parts to a bloated list of parts already in use. Another reason is that the values already in widespread use probably were chosen for reasons that may apply to old and new products alike.

To determine the company's inherent commonality, first obtain prioritized lists of each part category either based on total quantity used per year or based on which parts are used on the most products. Then plot part usage, for each category of part type, in the format of Figure 2 where the vertical axis represents volume and the horizontal axis represents part numbers prioritized by volume.

Next, apply judgement to the curves and determine which existing parts should be candidates for the commonality list. For the data presented in Figure 2, the left eighth of the parts would be good candidates. This will become the baseline list based on the inherent commonality of existing parts. New generation parts, with no usage history, are then added to the list.

Then eliminate duplicates and consolidate parallel lines of parts, for instance, standardizing on 1% resistors instead of having both 5% and 1% tolerances. Some companies report that the increased purchasing leverage of buying all 1% resistors compensates for the increased cost of the higher tolerance parts.

After the list is structured in some logical order (as discussed above), circulate the list to engineering groups for feedback and "buy-in" before implementation.

The design community may need to be educated on the strategic importance of using standard parts. Design engineers need to realize that no matter how simple a part appears, every part number inhibits agility and incurs a material overhead burden to document, procure, store, distribute, re-supply, and manufacture in low volume.

This part commonality approach was implemented by the author at Intel Corporation's Systems Group. Starting with 20,000 parts for printed circuit boards and computers, this commonality approach generated a preferred parts list of only 500 parts. For the category of "passive" components (resistors, capacitors, and diodes), 2,000 values were reduced to 35 values, one set for leaded axials, and another set for their surface mount equivalents. Fasteners for computer systems were standardized on one screw!

Tektronix used parts management software from CADIS¹⁰ to eliminate 32,000 duplicate part numbers from an active base of 150,000. Bob Vance, Tektronix VP and Chief Information Officer, summarized the return on eliminating excess parts: "There are few areas where a manufacturing company can make such a significant impact to its bottom line with so little effort. We want to invest our resources in product innovation and customer services, not carrying an overburdened parts inventory."¹¹

While training a company that manufactured air conditioners and furnaces, the author discovered the company used 152 different types of motors. When he challenged the designers, they insisted they needed every size to specify "just the right" motor for every application. Then he asked their supplier, GE Consumer Motor Division, "What would be the savings if those 152 motors could be reduced to five or ten?" The one word answer was "Massive!" Why? Because each of those five or ten motors would be ordered in volumes that would be ten times their current order volumes, thus resulting in greater economies of scale. Further, the five or ten chosen would have been the most cost-effective in GE's line — the motors that they produce in high volume for other customers too.

Similar procedures can be used to standardize tools, raw materials, and "features" which are defined as any shapes that require separate tooling to make: e.g. drilling or reaming holes, bending sheetmetal to a specified radius, and machining with profile end-mills. Fixturing geometry should be standardized to accommodate all parts in the product family without setup to position or clamp parts, dies, or molds. Processes should be standardized to eliminate setups to change adhesives or oven temperatures.

In agile operations, standard parts can be "pulled' into assembly operations on demand. Small, inexpensive parts can be resupplied automatically without sales forecasts with breadtruck deliveries. A local supplier is contracted to simply keep the bins full and bill the company monthly for what has been used, much like the way bread is resupplied by the breadtruck to a small market. All the MRP/purchasing expense is eliminated. Further, this type of delivery can assure a constant supply of parts, thus avoiding work stoppages.

Internally fabricated parts do not necessarily need to be made in a lot-size- of-one to support built-to-order product assembly. If the parts are small, inexpensive, and not likely to deteriorate or become obsolete, they can be made in batches as part of a kanban resupply. The simplest form of kanban uses two bins for each part. When the first bin is depleted, a second full bin moves into place. The empty part bin then is returned to its "source," which could be the machine that made the part, a sub-assembly workstation that assembled the part, or a supplier. The source fills the bin and returns it to this assembly workstation behind its counterpart bin which is still dispensing parts.

Kanban bins for small, cheap, and stable parts can be sized with enough parts to never cause a work stoppage, even at peak demand. Only rough estimates of this peak demand are necessary, so part resupply does not require forecasts and MRP-based procurement cycles.

Parts that are large, expensive, or prone to deterioration or obsolescence may have to be built flexibly as needed to minimize space, cost, or risk. Programmable machine tools can provide a versatile source of agility by programmably cutting bar stock, tubing, sheetmetal, cloth, leather, and so forth.

To eliminate setup and make these machines flexible, fixturing geometries and raw materials must be standardized. Further, all operations for all parts built must be within the capabilities of the machine including automatic tool changers, for instance, for machining centers and punch presses.

Versatile programmable machine tools can reduce processing steps by performing several functions in one machine, for instance, using laser cutters to blank, drill, punch, notch, and cut sheetmetal and using machining centers for drilling, taping, and milling. Flexible operations must have the ability to quickly and easily change CNC programs.

If plastic molded parts and cast metal parts can not be supplied easily by kanban, then the molding or casting process may need to be more flexible. The first step to make these processes more flexible would be to consolidate molded and cast shapes to the minimum number of different parts that are used in the maximum number of products. To maximize versatility, each of these parts may have extra metal or functions. Similarly, standard bare circuit boards may have extra traces and pads.

Agile companies design products and their assembly process to enable the assembly/configuration of all product variations in the family without any setup to retrieve parts, position parts, download programs, calibrate anything, or find and understand instructions. Assembly, adjustments, and configurations may all occur at some assembly operations for sub-assembly or final assembly. For agile assembly:

Assembly instructions must be quickly changed. As with the automation programs, the manual instructions can be changed by input from bar code technology by "wanding" the bar code of a key part (like a circuit board), pallet, or some kind of "traveler" card or documentation.¹⁶ This is preferable to entering work order numbers on a keyboard because it is faster and not prone to entry errors. In a "push" system (where work is scheduled and sequenced), the next set of instructions could be displayed after the previous job is finished.

Companies that use these agile product development techniques will experience: ultra-fast development of modular products; quick response to changing markets and opportunities; the lowest overhead costs; price premium opportunities; and better satisfaction of customers needs.

Dr. David M. Anderson, P.E., is a Lafayette, CA, based management consultant who specializes in seminars, workshops, and consulting on Agile Product Development and Design for Manufacturability. At the Haas Graduate School of Business at the University of California at Berkeley, he created and taught the course "New Product Development" as part of the Management of Technology Program.

Dr. Anderson has over 23 years of industrial experience and counts as clients many leading companies including several divisions of Hewlett-Packard, Emerson Electric, GE, Bausch & Lomb, Northern Telecom, United Technologies, Loral, Guidant, Freightliner, and many others. When he was Manager of Flexible Manufacturing at Intel's Systems Group, he initiated successful programs for Design For Manufacturability (DFM) and standardization of parts and tooling.

He wrote the book, Design For Manufacturability, Optimizing Cost, Quality, and Time-to-Market (CIM Press, 1990), the opening chapter in the SME handbook on DFM (TMEH, Vol 6), and the chapter on DFM and Mass Customization in the Quality Function Deployment Handbook (Wiley, 1997). His new book, Agile Product Development for Mass Customization, has just been published by McGraw-Hill Professional Publishing (Burr Ridge, IL). His next book will be on low-cost product development.

From 1977 to 1983, his company, Anderson Automation, Inc., generated design studies and built special production equipment for companies such as IBM, FMC, Clorox, and Optical Coating Labs. As the ultimate concurrent engineering experience, he personally built the equipment he designed and is proficient at machining and welding.

He holds professional registrations in Mechanical, Industrial, and Manufacturing Engineering and a Doctorate in Mechanical Engineering from UC, Berkeley.

Dr. Anderson can be reached at (510) 253-0900; fax: (510) 283-1330; e-mail: andersondm@aol.com. Web site: www.mcninet.com/GlobalLOOK/anderson.html

1 B. Joseph Pine II, Mass Custom-ization, The New Frontier in Business Competition, (Boston, Harvard Business School Press, 1993).

2 David M. Anderson, with an introduction by B. Joseph Pine II, Agile Product Development for Mass Customization, Niche Markets, JIT, Build-to-Order, and Flexible Manufacturing, (Burr Ridge, IL, Irwin/McGraw-Hill Professional Publishing, 1997), Chapter 7, "Building Products Quickly with Agile Manufacturing."

3 Richard E. White, "An Empirical Assessment of JIT in US Manufacturers," Production & Inventory Management Journal, v34, n2 (Second Quarter, 1993) pp. 38-42.

4 William A. Wheeler, III, JIT Client Engagement Results, Coopers & Lybrand Center for Manufacturing Technology (Burlington, MA, 1988).

5 Anderson and Pine, Agile Product Development for Mass Customization, see page 209, "Parametric CAD."

6 Niklas von Daehne, "Database Revolution," Success, v42, n4 (May 1995), pp. 38-42.

7 Anderson and Pine, Agile Product Development for Mass Customization, Ch. 5, "Early Steps and Prerequisites: Standardization."

8 David M. Anderson, Design for Manufacturability, Optimizing Cost, Quality, and Time-to- Market, (Lafayette, CA, CIM Press, 1990), Ch. 5.

9 Anderson and Pine, Agile Product Development for Mass Customization Ch. 7, "Building Products Quickly with Agile Manufacturing."

10 CADIS-PMX (Parts Management eXpert); CADIS, Inc., 1909 26th Street, Boulder, Colorado 80302; (310) 440-4363.

11 Tim Stevens, "Prolific Parts Pilfer Profits," Industry Week, v 244, n 11 (June 5, 1995), pp. 59- 62.

12 Anderson and Pine, Agile Product Development for Mass Customization, Ch. 10, "Designing Mass Customized Products."

14 AME (Association of Manufacturing Excellence) plant tour, Compaq Computer, Houston Texas, February 2-3, 1989.

15 Kiyoshi Suzaki, The New Manufacturing Challenge; Techniques for Continuous Improvement, Video program (Dearborn, MI, Society of Manufacturing Engineers).

16 Anderson and Pine, Agile Product Development for Mass Customization, see Figure 2-1 and p. 211, "Manual Assembly Instructions."

Agile Product Development For Mass Customization, JIT, Build-to-Order, and Agile Manufacturing