Journal of Cost Management

Document Sample
Journal of Cost Management Powered By Docstoc
    Journal of

A WARREN, GORHAM & LAMONT PUBLICATION                 Vol. 10, No. 2 Summer 1996

Activity-Based Management

      Projects, Models, and Systems—Where Is ABM Headed?

      Implementing Activity-Based Management:
       Overcoming the Data Barrier

      A Case Study in Economic Value Added
       and Activity-Based Management

      Metrics for the Order Fulfillment Process (Part 1)

      An Introduction to the Theory of Constraints
From the Editor / Paul Sharman

Strategic Cost Analysis / John K. Shank

Investment Justification / Ed Heard

    Journal of

Volume 10 Number 2 Summer 1996

Activity-Based Management
       Projects, Models, and Systems—Where is ABM Headed?                                                              5
James M. Reeve
Activity-based management (ABM) can be implemented in an organization as an improvement methodology, a costing
model, or a cost management system.

       Implementing Activity-Based Management: Overcoming the Data Barrier                                            17
William H. Wiersema
This article explores a simple alternative approach to designing ABM systems one that leads to more efficient, high
quality implementations at a low cost.

       A Case Study in Economic Value Added and Activity-Based Management                                             21
William W. Hubbell, Jr.
One excellent measure of shareholder economic value is economic value added (EVA), which can be used in combina-
tion with activity-based costing (ABC), as shown in this case study.

       Metrics for the Order Fulfillment Process (Part 1)                                                             30
Arthur M. Schneiderman
Metrics can be categorized as results metrics and process metrics. Result metrics are what customers see and what
drives their purchase decisions, while process metrics are the drivers of improvement.

       An Introduction to the Theory of Constraints                                                                   43
Jack M. Ruhl
The theory of constraints is a systems-management philosophy. This article provides an introduction to the theory of
constraints and to the concept of throughput accounting.

From the Editor / Paul Sharman                                                                                          3
Strategic Cost Analysis / John K. Shank                                                                                49
Investment Justification / Ed Heard                                                                                    60
Calendar                                                                                                               67
Volume 10 Number 2 Summer 1996

Journal of Cost Management
MANAGING EDITOR            BarryJ. Brinker
DESKTOP ARTIST             Christiane M. Bezerra
COPY EDITOR                Debra Van Bargen

EDITORIAL CONSULTATION Consortium for Advanced Manufacturing International (CAM-I)

                                   BOARD OF ADVISORS AND CONTRIBUTORS

s P.
ers At

s A.

Ac- Ran
tivi- dolf
ty Hols
Base t
d     Man
Man ager
age So-
Insti Man
sti- age
tute ment

Rob Ac-
in coun
Coo tants
per of
Pro- Can
fes- ada
of Ro-
Man bert
age A.
The well
Cla- Pres
re- iden
mon t
t    Ho-
uate Man
Sch age
ool ment

Ni- Cor
cho- pora
las tion
h Joh
Pro- n G.
fes- Ka
sor mml
of ade
Ac- Di-
coun rec-
ting tor,
Was Au-
hing dit
ton and




  Operations Services


  Professor of Accounting Harvard University
     Law Joh
Al- renc n K
fred e S. Sha
M. Mai nk
Kin sel
g    Man
Se- ag-
nior ing
Vice Di-
Pres rec-
iden tor
t    Para
Val- ra-
ua- mou
tion nt
Re- Con
sear sult-
ch ing
Cor up
ra- Cha
tion rles
Len ner
har Ar-
dt thur
Prin An-
cipal der-
Len sen
hard LLP
Stra- CJ.
tegic Mc
Ser- Nair

    l W.
    d J.
    er &
Noble Foundation Professor of
 Managerial Accounting
Dartmouth College

Paul A. Sharman
Focused Management
 Information, Inc.

LewisJ. Soloway
Managing Consultant
A.T. Kearney, Inc.

Peter B.B. Turney
Chief Executive Officer
Cost Technology, Inc.

Gene R. Tyndall
 Distribution Consulting
Ernst & Young

Lionel Woodcock
Proxima AMS

Pete Zampino
Subscription, Advertising, and Customer Service Information: For information about advertising, call Meg Chomicz at (212) 367-6569. For subscription
information, call (800) 950-1213; for customer service, call (800) 950-1205 Foreign callers (who cannot use our toll-free numbers) should call (617) 423-2020 or
fax (617) 423-2026.
Journal of Cost Management (ISSN 0899-5141) is published quarterly by Warren, Gorham & Lamont, The RlA Group, 31 St. James Ave., Boston, MA
02116-4112. Editorial offices We encourage readers to offer comments or suggestions to improve the usefulness of future issues. Contact Barry Bunker, Editor,
The RIA Group, 395 Hudson St., New York, NY 10014; (212) 367-6376 or fax (212) 367-6305. Subscription rates: $135/year. Printed in U.S.A. Periodical
postage paid at Boston, MA. This publication is designed to present accurate and authoritative information in regard to the subject matter covered. It is sold with
the understanding that the publisher is not engaged in rendering legal, accounting, or other professional services. If legal or accounting advice or other expert
assistance is required, the services of a competent professional should be sought.
Copyright © 1996 by Warren, Gorham & Lamont, The RIA Group. All rights reserved. No part of this journal may be reproduced in any form—by microfilm,
xerography, or otherwise or incorporated into any information retrieval system without the written permission of the copyright owner. Requests to reproduce
material contained in this publication should be addressed to Copyright Clearance Center, 222 Rosewood Dr., Danvers, MA 01923, (508) 750-8400, fax (508)
750-4744. Requests to publish material or to incorporate material into computerized databases or any other electronic form, or for other than individual or inter-
nal distribution, should be addressed to The RIA Group, 31 St. James Ave., Boston, MA 02116, (800) 950-1205.

Postmaster: Send address changes to Journal of Cost Management, The RIA Group, 31 St. lames Avenue, Boston, MA 02116.
Summer 1996   C1
Metrics for the Order
Fulfillment Process (Part I)
Arthur M. Schneiderman

 Metrics constitute a small and vital—subset of the nearly infinite number of possible process measures.
 Metrics can be categorized as results metrics and process metrics. Results metrics are what customers see and
   what drives their purchase decisions, while process metrics are the drivers of improvement.
 Good metrics have the following characteristics:
—They are linked to stakeholder satisfaction;
—They have documented, operational definitions; and
—They derive their usefulness only as part of an improvement process.
 The creation of a system of metrics requires a process of its own, with built-in means for refining the metrics.

    When you can measure what you are speaking               tion is to fix its order fulfillment process. Cus-
    about, and express it in numbers, you know some-         tomers' expectations about delivery performance
    thing about it; but when you cannot measure it,
    when you cannot express it in numbers, your
                                                             have changed dramatically over the last decade.
    knowledge is of a meager and unsatisfactory kind.        Ten years ago, a good supplier delivered on time
    William Thompson, Lord Kelvin, 1824-1907                 about 70 percent of the time. Now standards are
                                                             closer to 99 percent.

                  ost people will agree with the
                  following statement: “If you               One major step in the continuous improvement
                  don't measure it, it will not im-          process is the identification of key measures of
                  prove. If you don't monitor it, it         the process, or metrics. This article focuses on
                  will get worse.” But what to               the development and use of metrics for the order
measure and how to monitor what is measured                  fulfillment process at Analog Devices, Inc.
remain more an art than a science. This first part           (ADI), a midsize semiconductor manufacturer.
of a two-part series of articles examines metrics            As ADI learned, improving the order fulfillment
in terms of how they differ from measures. The               process is a good way to “surface” (i.e., bring to
article distinguishes between results and process            light) other areas in a company that need im-
metrics, then discusses what constitutes “good”              provement. Starting with the order fulfillment
versus “bad” metrics. The article gives an exam-             process is analogous to cutting inventory in a
ple of a process for the introduction and refine-            just-in-time management system. As Taiichi Oh-
ment of metrics.                                             no ---the inventor of the kanban system at Toyo-
                                                             ta---pointed out, reducing inventory is like lower-
Metrics are illustrated in this article using a typi-        ing the level of a river: when the water level
cal order fulfillment process. The second article            falls, the rocks and boulders become visible.
in the series will describe the resulting metrics
and how to integrate metrics into the company's              Helping to overcome limitations in other
management system.                                           processes. As delivery performance improves,
                                                             the limitations imposed by other processes be-
Recognizing the need to improve                              come more visible. These processes include:
Often the single most important improvement a
company can make to increase customer satisfac-

                                        COST MANAGEMENT

                             Exhibit 1. Delivery Performance as Measured by ADI




Yield enhancement;                                      Steps to improve customer satisfaction
Credit approval;
                                                         To improve customer satisfaction, ADI es-
Quarterly revenue management (AKA
                                                         tablished the following corporate-wide cus-
“the hockey stick” or nonlinear ship-
                                                         tomer service commitment:
ments—i.e., where a large percentage of a
period's shipments occur near the end of                 1. To deliver all orders to all customers
the period);                                                complete and when promised and to mi-
Dfx (Design for x, where x stands for ma-                  nimize lateness where we fail.
nufacturability, testability, usability, servi-          2. To meet our customers' delivery re-
ceability, recyclability, etc.) and                         quirements—always.
Management review.                                      3. To offer even shorter lead times if doing
                                                            so would give us a competitive advantage
                                                            by reducing total systems costs for us
Principles of metrics                                       and our customers combined.
The comprehensive set of on-time delivery
metrics that ADI developed played a critical             ADI recognized that these objectives were
role in bringing ADI's delivery performance              likely to be met only one at a time and in the
from below 70 percent before 1986                        order given. The last commitment assumes
to 96 percent or above by 1990. Exhibit 1                that customers always value shorter lead
graphs ADI's delivery performance, as ex-                times because it helps them build to order
pressed in terms of line items on purchase               rather than forecast, thus minimizing the
orders that are shipped late. (Generally, sep-           excess inventory required to cover forecast-
arate lines are used on purchase orders for              ing errors.
different parts or for different customer re-            Policy manual. ADI also produced a detailed
quest dates.)                                            policy manual that considered market-driven

                                                                                  Summer 1996       C3
                           Metrics for the Order Fulfillment Process (Part 1)

trends in each area and expanded on each of             delay, and machine settings are examples of
the customer service commitments. Sample                the essentially limitless number of measures
topics included the following:                          associated with any process.
    Taking all customers into consideration            The science of measures is called metrology.
     rather than focusing only on the largest;          It deals with such fundamental aspects of
    Partial shipments;                                 measures as the following:
    Padded lead times with early shipments;
                                                            Accuracy;
    Informing customers in advance of ex-
                                                            Precision;
     pected late shipments;
                                                            Bias; and
    Customers' needs versus wants; and
                                                            Repeatability.
    Transit-time responsibility.
                                                         Measures can be used for both process control
The commitments, policy changes, and
                                                         and process improvement. Used in process
other improvements helped transform ADI
                                                         control, measures characterize the critical
from a company that many of its customers
                                                         nodes in the process. Critical nodes represent
considered difficult to work with to a com-
                                                         the small set of sensitive control points that,
                                                         when held within a prescribed range, assure
As delivery performance improves,                        that the output of the process is stable (i.e.,
 the limitations imposed by other                        maintained within the control limits).
  processes become more visible                          For purposes of this two-part series of articles,
                                                         the following definition of metrics is used:
pany that one of its largest and most de-                      Metrics are a subset of measures of those
manding customers ranked as number one.                        processes whose improvement is critical to
ADI was also selected as Dataquest's “Mid-                     the success of the organization.
Size Semiconductor Supplier-of-the Year”
for two years running. This award, based on             Practically speaking, a subset is at most
a survey of 300 purchasing decision-makers,             three to five measures. A company that uses
recognizes “…manufacturers who exhibit                  more than five tends to lose its focus on the
extraordinary dedication to product quality             vital few opportunities for improvement. A
and customer service.”                                  lack of focus leads to diffused efforts and
                                                        slow progress. One symptom of a failure to
About metrics                                           make the distinction between measures and
                                                        metrics is reports with hundreds of meas-
There are several important aspects of metrics:         ures—all meticulously updated each
1. Metrics vs. measures: First, a distinc-              month—yet the companies in question
   tion must be made between the infinite               show few significant improvement trends.
   number of possible process measures
   and the much smaller subset of meas-                 Stakeholder satisfaction. Metrics are surro-
   ures that are actually useful to a com-              gates for stakeholder satisfaction and delight.
   pany's improvement efforts.                          (The term stakeholders refers to customers,
                                                        stockholders, employees, communities, sup-
2.   Both process and results metrics:                  pliers, and even future generations.) As a
     Second, metrics should be categorized              good metric improves, stakeholder satisfac-
     as process metrics and results metrics.            tion will increase, either directly or indirect-
This section also enumerates the properties             ly. This relationship between improvement of
of good metrics.                                        a metric and improvement in stakeholder sa-
                                                        tisfaction must be significant.
Metrics vs. measures. Confusion often arises
over the distinction between measures and               Improvement takes effort and resources. For
metrics. A measure is a numerical representa-           there to be a payback, metrics must focus on
tion of one of the attributes of a process. For         significant gaps in performance things that
example, time, temperature, speed, quality,             can make a competitive difference.

C4     Summer 1996
                                       COST MANAGEMENT

Folk wisdom about metrics. The importance         between alternative suppliers, an option that
of metrics is captured well by often- re-         internal customers usually do not have.
peated sayings such as the following:             Results metrics. Customers generally want
    “You can expect what you inspect”;           high quality (as defined by conformance to
    “if you're not keeping score, you're only    specification and fitness for use), low cost,
     practicing”; and                             and timely availability. Measures of these
    “You get what you measure.”                  characteristics are results metrics. To
But each of these sayings is incomplete in        achieve these results, suppliers often focus
itself. Implicit in each is the concept of        on the following:
measurement and the use of measurements            Short cycle times;
in the management process. Stated in               Inventory management;
another way: “if you don't measure it, it          Scrap reduction;
will not improve. If you don't monitor it, it      Training;
will get worse.” Why will things “get              Design for manufacturability;
worse”? Without management attention,              Statistical process control (SPC); and
performance tends to drift to lower per-           Market research.
formance—perhaps because lack of atten-
tion is interpreted as lack of importance.        Process metrics. Measures of the items
                                                  listed above are process metrics. Metrics
Measurement and monitoring are necessary          associated with activity, such as the per-
but not sufficient parts of a successful im-      centage of a company's associates who are
provement effort, which must also include
                                                  on improvement teams or the hours of
things like training, goal-setting, and promo-
tion. Nonetheless, measurement and monitor-       training per year per employee, are a subset
ing are the major tools that management can       of process metrics.
use to overcome a company’s “immune sys-          Results metrics represent the real objective
tem” (i.e., resistance to change). This “im-      of any process. They are the basis for the
mune system” is triggered by the introduction     customer's decision about suppliers. They
of new performance measures. Effective            are where money can be made or lost. Ul-
monitoring (which is discussed later) is a        timately, results and process metrics are not
powerful antidote.                                independent: they characterize a measure-
                                                  ment system for delivering increasing value
Results vs. process metrics
                                                  to customers.
It is useful to recognize two types of me-
trics—results and process metrics. Results        Properties of good metrics
metrics are seen directly by the process's pay-
ing customers; they are measures of how ef-       Metrics can be tested against a set of selec-
fectively a process meets the customer's          tion criteria either individually or collec-
needs. Process metrics, on the other hand,        tively as part of a system of metrics.
are usually invisible to customers; they deal     The first requirement for a good metric is
with the inner workings of a process and de-      that it should be a reliable proxy for stake-
scribe how the results are achieved. Process
                                                  holder satisfaction. In other words, im-
metrics are more related to the efficiency of
the process. Value is created by increasing       provement in the metric should link direct-
either process effectiveness or efficiency.       ly to improved stakeholder satisfaction.
                                                  This linkage should be clear and uncompli-
Internal customers. The concept of internal       cated. It should also be what mathemati-
customers somewhat muddies this distinc-          cians call monotonic—i.e., improvement in
tion, because it can be argued that any           the metric should always produce improved
process metric is a results metric in the eyes    stakeholder satisfaction. (See the related
of an internal customer of that subprocess.       discussion of “control limits and variability
However, the usefulness of the distinction        of processes” below.) There should be no
relies on the ability of a customer to choose     nonzero optimum value for the metric.

                                                                         Summer 1996         C5
                           Metrics for the Order Fulfillment Process (Part 1)

For example, lead time defined as “ship date            Characteristics of good metrics
minus order date” fails this test because cus-
                                                        Among the characteristics of good metrics
tomers will generally tell their supplier
                                                        are the following:
when they want the product. Shipping it ear-
ly to them (a shorter lead time) simply in-                 WelI-documented, unambiguous opera-
creases their inventory.                                     tional definitions.
                                                            Continuous values: Metrics should be
It is becoming more common for customers
                                                             able to take on continuous values so that
to “ding” (i. e. penalize) their suppliers for
                                                             incremental improvement can be ob-
early as well as late shipments. A better me-
tric is excess lead time, which is defined as
                                                            Metrological standards: Metrics should
“supplier-quoted minus customer-requested
                                                             also meet such metrology tests as accu-
lead time.” Improving this metric always
                                                             racy, precision, reliability, and bias.
leads to increased customer satisfaction.
                                                        Making metrics useful. For metrics to be
Overcomplicating metrics                                useful as part of an improvement effort, they
                                                        should be all of the following:
There is a potential danger of overcompli-
cating metrics to make them a “better”                   Oriented toward weaknesses or defects
proxy for customer satisfaction. At ADI,                (i.e., metrics should measure weaknesses or
for example, a quadratic equation (using                defects in the process);
what is known as Taguchi's loss function)                Timely; and
was proposed for measuring the impact of                 Accessible to those responsible for im-
late shipments to customers. The equation               proving the process.
(which used the number of days the order                 Linked to an underlying data system that
was late and worked on the assumption                   facilitates the identification of root causes.
that being two weeks late was more than                 In other words, if the value of the metric
twice as bad as being one week late) used               prompts managerial attention, then data
the square of the number of weeks late to               should be available so that the responsible
measure the impact of the late shipment.                person can explain the cause of the varia-
Unfortunately, not all managers have the                tion.
training to work easily with such sophisti-
cated mathematical formulas. Moreover, in               Different metrics based on the same measure
this example, no one could produce either
                                                        Different metrics are often calculated using
a real or hypothetical situation where the              the same or similar measures. It is important
resulting action would be different based               that the definition of the measure be the
on the more complicated metric, so sim-                 same in each metric. A “late line” (i.e., the
plicity prevailed.                                      merchandise represented by a distinct
Truly monotonic metrics are often difficult             line-item on a purchase order) should be
to define. In practice, there can be too                defined the same way for all metrics based
much of a good thing. For example, cycle                on the number of lines late. A metrics man-
time—a key metric of the manufacturing                  ual that contains the detailed operational
process—can be reduced to a point where it              definitions of the metrics should maintain
produces increased value to customers.                  this consistency of definition of interme-
Beyond that point, however, throughput de-              diate measures.
clines, delivery performance suffers, and               Judgment-based or subjective measures
value is destroyed. Excess cycle time (i.e.,
                                                        often create difficulties. For example, visu-
actual minus optimum cycle time) trans-
                                                        al defects (e.g., scratches, chips, discolora-
forms cycle time into a monotonic metric.               tions, and bent leads on integrated circuits,
Also, combined with metrics that character-             which may have no effect on a product’s
ize the other side of the tradeoff, cycle time          performance) are often difficult to define. If
can be a useful component of a system of                they are included in the definition of a de-
metrics.                                                fect, then objective criteria should be estab-

C6    Summer 1996
                                       COST MANAGEMENT

lished to minimize variation in the metric       replacing the computer system) rather than
caused by differences in interpretation.         trying to achieve continuous improvement
                                                 of the existing process.
Control limits and variability of processes
                                                 Statistical process control (SPC). Tradi-
                                                 tional control charts provide a methodolo-
As part of the operational definition of a
                                                 gy for establishing control limits. However,
metrics, control limits should be identified.
                                                 Walter Shewhart, the inventor of SPC and
Metrics, like the underlying process they
                                                 control charts’ based most of his work on
represent, have inherent variability. Man-
                                                 what he called “a constant system of
agement action should be required only
                                                 chance causes.”1 These systems have con-
when a change in the metric is statistically
                                                 stant averages and standard deviations. Be-
                                                 cause metrics are used to drive a process
Limit values. Limit values of a metric should    that will produce decreasing averages and
be well understood. Most processes cannot        reduced variation, the Shewhart model
sustain performance with zero defects (un-       needs to be adjusted for this nonstationary
wanted outcomes) because of their inherent       (i.e., time-varying) process based on a rea-
(random) variability.                            listic improvement model. (An example of
                                                 this correction is given in Part 2 of this se-
The process capability (or entitlement or
                                                 ries of articles.)
theoretical limit) should be estimated and
metrics defined to reflect the gap between
actual and theoretical capability. This be-      Smoothing
comes more important as the limit of a
process is approached.                           Smoothing (or averaging) is often used to
                                                 reduce variability in metrics. For example'
Example at ADI. For example, the theoretical     “percent late lines” is the number of late
capability of ADl’s order fulfillment process    lines divided by the number of scheduled
was limited in reality by its computer tech-     lines during the period of measure (e.g., a
nology which constrained the availability of
manufacturing plan updates because they              Improvement takes effort and
were periodic rather than continuous. Up-        rsources. For there to be a payback,
dates occurred on weekends, which meant            metrics must focus on significant
that by the following Friday orders were be-       gaps in performance—things that
ing quoted against an out-of-date plan.           can make a competitive difference.
According to a rough estimate, this com-
puter constraint introduced a 2 percent error
into the system and limited sustainable
                                                 day, week, month, or year). The longer the
on-time delivery to about 98 percent. At 70
percent on-time delivery, the difference be-     period, the smoother the resulting metric
tween a correctable gap of 30 percent (zero      will appear over time. The longer the mea-
defects) and 28 percent (actual limit) leads     surement period, however, the longer the
to no actionable consequences. However, at       time required to detect trends. It is prefera-
97 percent on-time delivery, the difference      ble to use a measurement period that is less
between a goal of an improvement of 3 per-       than the process cycle time and to smooth
cent (zero defects) versus a goal of 1 percent   the resulting data, if necessary, using expo-
(the actual limit) is a factor of three. The     nentially weighted moving averages.
targeted incremental improvement for the         Exponential averaging differs from direct
next year might realistically be only 0.5 per-   averaging in that it weights the most recent
cent rather than the theoretical 1.5 percent.    data more heavily than the older data. Al-
If a 0.5 percent improvement is not good         though this may sound complicated, the cal-
enough to meet customer needs, manage-           culation is quite simple. First, a weighting
ment should consider reengineering (e.g., by     factor,  , is chosen that has a value
                                                                        Summer 1996          C7
                                Metrics for the Order Fulfillment Process (Part 1)

                                             Exhibit 2. Ask Why Five Times
                           Why?                                              Because:

     Why(1) are 20 percent of the orders late?           30 percent of the time they were not released by

     Why(2) were orders not released by credit?          45 percent of the time the customer was on credit

     Why(3) was the customer on credit hold?             80 percent of the time they had exceeded their cre-
                                                         dit limit.

     Why(4) did they exceed their credit limit?          99 percent of the time we didn’t know at the time
                                                         they placed the order that it would put them over
                                                         their credit limit?

     Why(5) didn’t we know?                              The order entry system does not tell us the custom-
                                                         er’s available credit.

between zero and one. It is the fractional                   managers and associates can use the infor-
weight placed on the current, unsmoothed me-                 mation for process improvement. Note that
tric. (For  =1, no averaging occurs, while for              it becomes awkward to ''ask why five times''
  =0, the current value has no impact on the                when a metric is strength-oriented (e.g., the
average.) Then the new averaged value is                    percentage of on-time shipments) rather
times the current value plus (1-  ) times the               than defect-oriented (e.g., the number of
averaged value from the previous period. In                  lines shipped late).
equation form, this is as follows:
                 xt  x t  1   xt 1
                                                             Timeliness is an important requisite for use-
A good trial value of  is 0.2, although some
                                                             ful metrics. If the metrics lag the action by
experimenting around this value is usually
                                                             too long a period, the trail grows cold and
worthwhile. Control limits for exponentially
                                                             root cause analysis becomes difficult or im-
averaged metrics can be easily calculated.2
                                                             possible. For example, a month's delay be-
Asking "why" five times                                      tween the committed shipment date and the
                                                             reporting of the late shipment makes it diffi-
Incremental improvement is based on the abil-
                                                             cult for anyone to identify the root cause for
ity to “ask why five times.”3 Exhibit 2 demon-
                                                             that late shipment. Daily reporting of yes-
strates application of this technique to late
                                                             terday's late shipments is essential for effec-
shipments. The point of the exercise is that—
                                                             tive problem solving. Monthly summaries of
by the time the fifth “why” is answered—
                                                             the same metric are usually all that is re-
corrective action (i.e., reversal of the root
                                                             quired for managers to supervise the im-
cause) usually becomes obvious. In the case
                                                             provement process.
illustrated in Exhibit 2, the solution is to add
available-credit information to the order-entry              Processes with long cycle times. Timely
field and to develop a process for working                   results metrics are particularly difficult for
with customers to increase their credit limits               processes with long cycle times. The results
when their orders are entered into the system.               of new product development (in terms, for
                                                             example, of return on investment in research
Ideally, metrics and the ability to drill down
                                                             and development) are often unknown until
smoothly to root causes can be integrated
                                                             years after the investments are made. By
into a company's information system so that
                                                             then, the original process may have changed
C8     Summer 1996
                                       COST MANAGEMENT

and the individuals involved may have             Then and only then is a metric based on the
moved to other assignments. Moreover,             deviation from plan useful in improving the
learning about what failed in yesterday's         hoshin kanri process itself.
environment may not help in today's envi-
ronment.                                          Completeness
Results-focused managers- trying to com-          A set of metrics should be complete. That is,
pensate for this delay—often use forecast or      metrics should be included for all possible
prediction-based metrics, such as forecast        undesirable tradeoffs. Many tradeoffs can
break-even time or third-year revenue and         be anticipated in advance, while others are
profits. But these proposed metrics are more      discovered along the way. With an incom-
applicable to the forecasting process than        plete set of metrics, intentional or uninten-
product generation. Most forecasts are not        tional “gaming” of the metrics can occur
based on a documented process. Even when          (i.e., doing things to improve the metric
they are, the processes have questionable         while decreasing overall stakeholder satis-
capability (i.e., high inherent variability).     faction). Thus, for example, extending
All in all, forecast metrics and results me-      quoted lead times may improve on-time de-
trics for processes with long cycle times are     livery, but it will usually decrease customer
of dubious value except for their use in          satisfaction.
quantifying forecast variability.

Focusing on process, not results                    By focusing on a core business
Does this mean that there are no useful me-        process—order fulfillment— and
trics for processes with long cycle times?
The Japanese often say, “Focus on process,
                                                    providing a conceptual frame-
not on results.” This is particularly good                       work,
advice for processes with long cycle times.       this article should equip the read-
Process metrics such as “percentage of                             er
planned milestones missed or rescheduled,”
“error-based engineering change orders,”            to extend applications to other
and “forecasting and planning process                      areas of interest;.
checklist items not completed” have been
used effectively for driving improvement.
                                                  Perhaps the area in which completeness is
It is often tempting to define a metric as the    most often overlooked is in training me-
gap between actual and planned results ra-        trics. “Percentage of employees trained” is
ther than process capability. However, that       often the only metric used. However, with-
gap is linked to defects in the planning          out metrics for the effectiveness of the train-
process or its implementation, not the under-     ing, the value of the training cannot be
lying process itself. In other words , it leads   measured. In education, test scores are used
to the question “Why did we not achieve           to measure effectiveness. In industry, how
plan?” instead of “Why is the process pro-        the training is applied tells more about its
ducing these defects?”                            value. A complete set of training metrics
                                                  includes both the percentage of people
For example, under the Japanese system of
                                                  trained and, for example, the percentage of
hoshin kanri (which is usually translated as
                                                  trainees who effectively applied what they
“policy deployment” and is a major exten-
                                                  learned within three months after training.
sion of management by objectives), imple-
mentation plans are developed based on a
                                                  The ADI order fulfillment process
thorough understanding of the underlying
process, including its defect causes and the      To illustrate the principles explained so far,
required corrective actions.4 There is a high     consider the order fulfillment process at
level of confidence that if the plan is ex-       ADI, where most bookings were for repeat
ecuted, the desired results will be achieved.     business. These repeat customers had de-
                                                                         Summer 1996          C9
                           Metrics for the Order Fulfillment Process (Part 1)

signed an ADI part into their product and               typically consumed three days for domestic or-
made periodic purchases to reflect their pro-           ders and five days on international orders. This
duction needs. Prices were generally set in             administrative lead time
advance, so the principal issue for each order          (ALT) was subtracted from the FCD to yield the
was availability.                                       date when the factory was committed to have
                                                        the product at the warehouse. At this time' the
To start that part of the order fulfillment             product was assigned to specific orders. This is
process at ADI discussed here, two items                called the ''to-be-assigned'' (TBA) date. The ac-
were entered into the system:                           tual date on which an order is shipped to the
      The order entry date (OED); and                  customer is the actual ship date (ASD). Exhibit
      The customer request date (CRD).                 3 shows a flowchart of the ADI order fulfillment
                                                        process. The various milestones on the time line
If the product was expected to be available             of this process are summarized in Exhibit 4.
on the CAD, the company committed to that
date. If not' the order was referred to the ap-         In terms of the acronyms of Exhibit 4, ADI's
propriate factory, where the production                 corporate-wide customer service commitment
planners scheduled the order and com-                   (see the previous discussion) can be restated as

  Even after detailed metrics are                       1. ASD  FCD or ASD- FCD  0
 developed for the scorecard, they                      2. FCD  CRD
 and their associated supplemental                      3. Required lead time < CRD - OED
        metrics must evolve.                            A process for defining metrics
                                                        The development of effective metrics is an on-
mitted to a date. Often they could adjust their         going process. Each organization needs to create
production plans to meet the CRD. If not'               its own approach that is consistent both with the
they quoted a later date. In either case' the           desired results and its unique culture. At ADI
factory commit date (FCD) represented                   the CEO assigned the author (who was
ADI's response to the customer's request.               vice-president of quality and productivity im-
Usually customers accepted the FCD pro-                 provement, or VP/QPI at the time) responsibility
posed and adjusted their own production                 for establishing metrics for ADI's key business
schedules accordingly.                                  processes. In today's jargon' the VP/QPI was
Factory commit dates (FCDs). The FCD                    made process “owner” for nonfinancial perfor-
represented the date by which ADI commit-               mance metrics.
ted to ship the customer's entire order. Unless         "Bottom-up" or "top-down." The first question
otherwise directed' ADI reserved the right to           the VP/QPI faced was whether the metrics
ship early up to two weeks before the CRD if            should be defined ''bottom-up'' or ''top-down''
the product became available sooner than                that is' should the various subprocess owners
planned.                                                define their own metric or should the VP/QPI
To be shipped by the FCD' a product had to              (i.e., a member of the corporate staff) do it for
arrive at ADI's central warehouse before the            them?
FCD to allow time to review the customer's              It seemed doubtful that either of these extremes
credit and complete the various shipping                would work. If subprocess owners were allowed
subprocesses. A central warehouse was used              to define their own measures, something was
because most orders contained products from             likely to be lost from the customer's perspective.
different manufacturing locations. Customers            The resulting individual metrics might not add
preferred receiving a single shipment rather            up to a system of metrics for improving custom-
than multiple packages from various ADI                 er satisfaction. Yet the VP/QPI recognized that
factories; a single shipment was also cheaper           he lacked sufficient specific knowledge of the
for the customer.                                       subprocesses and thus risked defining imprac-
Subprocesses and administrative lead times              tical metrics.5 Furthermore, those responsible for
(ALTs). The credit and shipping subprocesses            improving their processes

C10 Summer 1996
                                             COST MANAGEMENT

                            Exhibit 3. The Order Fulfillment Process at ADI

ORDER ENTRY                 DIVISION               WAREHOUSE          CREDIT           CUSTOMER

                                                                                       Customer places       Sales
                                                                                       order (w ith CRD)    process

Order entered into
  system(OED )


                                                                                        Confirm FCD to
 Refer to division

                           Schedule order


                              Transfer to
                           w arehouse ( BA)

                                                                      Is credit            Increase
                                                    Refer to credit               No
                                                                        OK?               credit limit


                           Provide missing           Paperw ork                         Provide missing
                                              No                       No
                             information             complete?                            information


                                                    Print shipping

                                                    Pick and ship
                                                     order (ASD)

                                                                                       Customer receives    Pay ment
                                                                                       order (dock date)    process

                                                                                   Summer 1996             C11
                                     Metrics for the Order Fulfillment Process (Part 1)

                                    Exhibit 4. ADI’s Order Fulfillment Process Metrics Dates
                                                    (not to scale)
                                                          (not to scale)

                                                            Early                        Ontime                 Late

                                                                                         1 week

                     Confirmed to                                          TBA
                       customer                                     To be assigned

        OED                                           CRD                                           FCD                        ASD
 Order entry date                          Customer request date                          Factory commit date          Actual ship date

                                           2 weeks

                                                   System permitted ship window


were not likely to feel a sense of ownership if                            that would be used. In this way, there was
the VP/QPI defined all the metrics.                                        little, if any, room for interpretation.
Ultimately, therefore, the VP/QPI relied on                                At first, the participants were leery of this
teams of subprocess owners for help in defin-                              process. As in most “old way” companies,
ing the metrics. He chaired these teams and                                these managers associated metrics with the
was the final decision maker when they could                               stick, not the carrot. A combination of per-
not reach a consensus. He relied on the ap-                                suasion and decree was needed to complete
peals process, described below, as the vehicle                             the definition of metrics within a reasona-
for continuous improvement of the initial set                              ble time. Even so, it took nearly 18 months
of metrics.                                                                to establish the set of metrics initially used
                                                                           at ADI. Perhaps the ultimate test of this
Coming up with appropriate metrics. The
                                                                           process is that, after several reviews, the
VP/QPI did not act as a facilitator. People
                                                                           metrics have remained virtually unchanged
were asked to volunteer to prepare a detailed
                                                                           since they were first implemented in 1988.
proposed specification for a given metric. Oc-
casionally, two or more people would volun-                                To those who are strong proponents of em-
teer (usually with opposing views), which                                  ployee empowerment, this top-down model
meant that there were alternatives to evaluate.                            may seem “disempowering.” However,
At other times, no one would volunteer, so the                             when process owners and managers form a
VP/QPI would assume responsibility for de-                                 partnership to act as architects of multipro-
veloping a proposal. In doing so, he relied                                cess organizations, both parties are ulti-
heavily on industry benchmarks and defini-                                 mately empowered.
tions used by major ADI customers. If volun-
teers failed to complete their assignments by                              Maintaining a focus on weaknesses
the agreed-on date, they were given one more                               In defining metrics, there is the option of
chance, after which the VP/QPI would revert                                measuring either what went right or what
to his own proposal.                                                       went wrong. From a psychological perspec-
The proposals were debated, refined, and                                   tive, Westerners prefer to focus on what
eventually accepted. A manual evolved from                                 they do right, because the opposite often
this process; it defined each metric in equa-                              leads to finger-pointing and blame. But, as
tion form and prescribed in detail the fields                              Chris Argyris, James B. Conant Professor
in the order-entry database (or other source)                              at the Harvard graduate schools of business
C12 Summer 1996
                                      COST MANAGEMENT

and education, has recently pointed out, “In      targeted for improved corporate perfor-
the name of positive thinking . . . managers      mance. To offset the historic bias in favor
often censor what everyone needs to say           of financial measures, an organization is
and hear.”6                                       likely to need an unbalanced scorecard at
                                                  first (i.e., in favor of nonfinancial pefor-
To find root causes one must focus on de-
                                                  mance) .
fects (or weaknesses). This conflict can be
resolved if the wisdom of Deming and Ju-          The QSC included the following members:
ran prevails. Both observed that about 95
                                                     The CEO;
percent of defects are caused by the
                                                     The president;
process- not the people. Metrics should thus
                                                     The vice-presidents of manufacturing,
be used as pointers to places in the process
                                                      sales, technology, and human resources;
that need to be improved. If this ground rule
is established, then the orientation of me-
                                                     Two representative general managers of
trics on weaknesses becomes more accepta-
                                                  This leadership group had the broad and ba-
Do not rely on averages. Another danger in
                                                  lanced perspective to assure that ADI focused its
defining metrics is the use of averages. For
                                                  improvement efforts on the right things.
example, average excess lead time could be
dropping while—for some class of products—        Nonetheless, even after detailed metrics are
the lead time is increasing, thus causing a de-   developed for a balanced scorecard, they
cline in customer satisfaction for purchasers     and their associated supplemental metrics
of these products. Therefore, it is important     must evolve. Nonscorecard (or supplemen-
that both averages and distributions (histo-      tal) metrics provide checks and can become
grams) be included where appropriate.             candidates for future elevation to scorecard
                                                  status. To this end, ''metrics boards of ap-
The metric improvement process                    peal,, were established on an ad hoc basis.
Metrics evolve over time to reflect both the
                                                  Metrics boards of appeals. For the order
changing needs of constituents and process
                                                  fulfillment metrics, a board of appeals
learning (i.e., the systematic mastery of a
                                                  chaired by the VP/QPI was established.
process). While financial measures have
                                                  This board included the credit and ware-
been in use for more than a hundred years
                                                  house managers and also operations manag-
and are relatively stable, nonfinancial
                                                  ers from any affected divisions. Representa-
measures need to change to improve an or-
                                                  tives from sales—and anyone else interest-
ganization's alignment with rapidly chang-
                                                  ed—were also invited to participate. The
ing customer requirements. Two vehicles
                                                  basis for an appeal was that a metric in use
were developed at ADI to continuously im-
                                                  inappropriately penalized a subprocess for
prove the metrics:
                                                  something either desired by customers or of
   The balanced scorecard; and                   no importance to them.
   The metrics board of appeals.
                                                  To bring an appeal, people had to show that
The balanced scorecard. In 1987 ADI pio-          the issue was a significant bar on their Pare-
neered in the development of a balanced           to chart (i.e., a rank-ordered bar chart)
scorecard.7 The scorecard contains quarter-       showing causes of defects. They also had to
ly goals for both financial and nonfinancial      specify how the definition of the metrics
metrics and is updated annually as part of        should be changed. Finally, they were re-
the business planning process. At the             quired to provide a retrospective be-
start of the process, the quality steering        fore-and-after comparison using actual data.
council (QSC) chaired by the VP/QPI re-           If broader business issues were involved'
viewed the areas addressed on the scorecard       the board's recommendation was reviewed
to assure that the right things were being        with the COO. One such issue was the

                                                                         Summer 1996        C13
                           Metrics for the Order Fulfillment Process (Part 1)

changing of FCDs when the customer                      increase unless the metric allowed early
changed the CRD. The COO, concerned                     shipment to be used to offset long quoted
that the ability to change FCDs could un-               lead times—a clearly undesirable situation.
dermine the credibility of the metrics, de-             His proposal was quickly withdrawn.
cided that no FCDs could be changed with-
out his written approval. Late shipments to             Summary
the original FCD caused by customer
changes were tracked, but were never a sig-             The creation of metrics is itself a process
nificant cause of overall lateness.                     that includes refinement cycles. This article
                                                        describes the approach ADI used to define
The boards maintained a bias toward not                 metrics, an approach that attempted to bal-
changing the metrics to keep the definitions            ance top-down alignment with bottom-up
stable whenever possible. In this way, trend            ownership. The second part of this series of
data were not distorted unduly by repeated              two articles will describe the resulting me-
changes in definitions of metrics.                      trics and the role they play in the day-to-day
                                                        management of the company. 
Examples. One example of an appeal dealt
with shipments to ADl's foreign sales affili-
ates' which were treated by the system as if            Arthur M. Schneiderman is an independent
they were individual customers. Most affil-             consultant in process management. He is lo-
iates wanted their orders held for a single             cated in Boxford, Massachusetts. The author
weekly shipment. However, orders were                   wishes to thank Ray Stata (CEO) and Jerry
committed for shipment each day of the                  Fishman (president) of Analog Devices, Inc.,
week. Therefore, an order might be sche-                for creating the environment and providing
duled for Monday shipment to an affiliate               the constant encouragement needed to make
who wanted shipment on each Friday. The                 the metrics and the scorecard a reality. Of the
warehouse would hold the order at the affil-            many others at ADI and elsewhere who con-
iate's request until Friday, which meant that           tributed to this effort, Elizabeth Derwin de-
the order would go out “late.” Clearly, the             serves special acknowledgment.
metrics tempted the warehouse to ship dai-
ly, against the wishes of the customer.
                                                          W.A. Shewhart, Economic Control of Quality of Manufac-
When this issue was first raised, it was not            tured Product (New York: D. van Nostrand Company, Inc.,
a significant cause of ware-                            1931): 146.
                                                          See for example, J. Stuart Hunter, "The Exponentially
house-controllable lateness, so action was              Weighted Moving Average, " Journal of Quality Technolo-
deferred. This decision to defer action es-             gy (Vol. 18, No. 4, October 1986): 203-210.
tablished the principle that effort should be             See chapter 3, for example, in Kaoru Ishikawa, Guide to
                                                        Quality Control (Tokyo: Asian Productivity Organization,
focused on process improvement rather                   1982), which was originally published in 1971.
than nitpicking about the metrics. As the               4
                                                          Yoji Akao (ed.) Hoshin Kanri Policy Deployment for Suc-
warehouse improved, foreign shipments                   cessful TQM (Cambridge, MA: Productivity Press, Inc.,
eventually became the warehouse's num-                  5
                                                          For a useful discussion of operational definitions see W.
ber-two root cause of late shipments. At                Edwards Deming, Out of Crisis (Cambridge, MA: Massa-
that time' the appeal was accepted and ap-              chusetts Institute of Technology, Center for Advanced Engi-
                                                        neering Studies, 1982): Chapter 9.
propriate changes were made in the order                6
                                                          Chris Argyris, "Good Communication That Blocks Learn
entry system.                                           in," Harvard Business Review (July-August 1994): 77-85.
                                                          Robert S. Kaplan, "Companies as Laboratories," in The
In another case, a general manager pro-                 Relevance of a Decade, Paula Barker Duffy (ed.) (Boston:
posed a new definition for a lead-time me-              Harvard Business School Press, 1994): 179- 182.
tric that he thought would more fairly
portray his division's performance. Careful
analysis of the proposal, however, showed
that adoption of the proposal would have
caused the manager's reported lead times to

C14 Summer 1996

Shared By: