dav case

Q1. Why is DAV using Statistical Process Control (SPC)? What are the challenges in
applying SPC to a service industry compared with manufacturing?

Q2. The first 12 weeks of the data in Exhibit 4 represent the diagnostic period for the
Policy Extension Group. What are the 3-sigma control limits for the process? In
which of the subsequent weeks is the process out of control (if any)?

Q3. Develop specific implementation plans for solving the problems facing Annette
Kluck that are described on page 9 of the case.

For the exclusive use of R. Rodriguez, 2015.
Harvard Business School

Rev. April 15, 1997

Deutsche Allgemeinversicherung
Annette Kluck parked her chin on the heel of her hand as she watched her electronic
fishbowl, late on a Friday afternoon in January 1996. Frank Schoeck, the head of operations at DAV
Kundendienstgruppe1 (DAKG), had just made a surprise visit. “So,” he had said, in his famously
blunt style, “when do you think we’ll start seeing some visible results from all this operations
improvement work you’ve got everyone doing?” Kluck was a little chagrined. She was convinced
that the performance of the DAKG customer service operation had improved, but that the evidence
might take a little time to appear.
Kluck, the architect behind Prozessmessung und Verbesserung2 (PMV), was head of
Operations Development at Deutsche Allgemeinversicherung (DAV), one of the largest insurance
companies in Europe. The PMV project was a revolutionary effort to use manufacturing-style
improvement techniques in insurance services. It had begun six months earlier as part of a broad
initiativeto improve information accuracy and quality throughout DAV. The project had, as its name
suggested, been broken into two phases. In the first phase, methods were developed for measuring
the quality of a number of process steps at DAV (such as the process for transcribing information
from a customer application form onto the computer). In the second phase, these new measurement
methods would be used as the basis for performance improvement. DAV had now completed the
measurement phase, and was tracking the performance measures over time. It was now time to begin
improving the performance of the various processes.
Kluck, however, was facing a number of
difficult problems with the improvement phase of the project.

Deutsche Allgemeinversicherung
Founded in 1966 by Andreas Steininger, Deutsche Allgemeinversicherung was one of the
world’s largest insurance companies, writing nearly DM 48 billion in premiums in 1996 in over 32
countries. Roughly 51% of DAV’s business was in Germany. Sixty-percent of DAV’s business in
Germany was in retail insurance (including, for example health and property insurance). In
addition, its retail offerings included life insurance and disability income protection.


Customer Service Group
Process Measurement and Improvement

Professor David Upton prepared this case as the basis for class discussion rather than to illustrate either effective or
ineffective handling of an administrative situation.
Copyright © 1996 by the President and Fellows of Harvard College. To order copies or request permission to
reproduce materials, call 1-800-545-7685 or write Harvard Business School Publishing, Boston, MA 02163. No
part of this publication may be reproduced, stored in a retrieval system, used in a spreadsheet, or transmitted in
any form or by any means—electronic, mechanical, photocopying, recording, or otherwise—without the
permission of Harvard Business School.

This document is authorized for use only by Rebeca Rodriguez in MGT 653-67 taught by DR. HARESH GURNANI, University of

Miami from April 2015 to October 2015.

For the exclusive use of R. Rodriguez, 2015.

Deutsche Allgemeinversicherung

Along with companies like Allianz, Credit Lyonnais and Aetna, DAV was one of the giants of
the industry. As the second largest firm in Germany, DAV was acutely aware of its prominent
position, and had begun to think carefully about how to maintain that position as smaller insurance
companies armed themselves to attack its primary markets. While its dominant advantage remained
the breadth of its offerings and an excellent group of insurance risk managers, DAV was also aware
that individual managers were movable assets and that it needed to develop capabilities that were
not only valued by customers, but were also distributed throughout DAV. Kluck commented on the
hazards of being a successful firm:
“We see firms killed by success all the time: they don’t know how to use the
fat years to prepare for the lean ones. They get so used to their success that they let
their operations go, and don’t learn to do things better. We are determined not to
replicate their mistakes.”
Managers in other firms throughout the industry attributed DAV’s success to two key factors:
sound, traditional insurance management; and outstanding customer service. Though few managers
liked to hear it, insurance was becoming more and more of a commodity. In these circumstances,
customer service was becoming an important tool for differentiating one firm from another. Quality
in customer service had progressively become a critical element in DAV’s strategy. As Frank Schoeck
pointed out:
“Exceeding customer expectations for the quality of service is an important
way to maintain current customers and attract new ones. Customers don’t
particularly care how the company is organized and where its offices are located. All
they care about when they call someone on the phone or write a letter is that the
person is nice, and does what they want done correctly and quickly without any
An important part of delivering this responsive, “hassle-free” service was the ability to
process information and data without mistakes, and the ability to retrieve it in a timely manner.
Delivering consistent, outstanding quality was complicated by the fact that DAV ran operations in
numerous divisions in different locations. In DAV’s case, its size and the sheer diversity of its
operations meant unusually daunting challenges for its customer service group (Kundendienstgruppe).

DAV Kundendienstgruppe
The DAV Kundendienstgruppe (DAKG) was the back-office part of the operation, focusing
on the retail side of the business (with the other side servicing institutional insurance, such as
environmental insurance). DAKG processed applications from new customers for policies, changed
policy information, carried out various legal registrations, and kept track of an customer’s personal
information. While DAV aimed to provide the highest standards in customer service to the
individual, everyone with the group understood the real challenge of running such an operation.
“DAKG is a high volume-production environment,” remarked Kluck.

This document is authorized for use only by Rebeca Rodriguez in MGT 653-67 taught by DR. HARESH GURNANI, University of

Miami from April 2015 to October 2015.

For the exclusive use of R. Rodriguez, 2015.
Deutsche Allgemeinversicherung

Exhibit 1


DAV Kudendienstgruppe (DAKG): Organization Chart
Frank Schoeck

Jürgen Rentschler

Markus Scherer

Erwin Weippert

Peter Kolb

München, Köln, Hamburg

Retail Processing Support

Life Insurance Operations

Retail Transaction Processing
München, Hamburg

141 Regular
60 Temp

154 Regular
132 Temp

233 Regular
160 Temp

Jochen Härdter

Volker Maus

Beatrice Scherff

Property Insurance Services

Customer Comunication Division

Customer Problem Resolution
München and Köln

64 Regular
16 Temp

304 Regular
119 Temp

76 Regular
31 Temp

205 Regular
14 Temp

DAKG employed over 2,000 people at three primary sites: München; Köln; and Hamburg.
Divisions within DAKG included:

Systems (München, Köln and Hamburg)

Retail Processing Support (München)

Life Insurance Operations (Köln)

Retail Transaction Processing (Verarbeitung eines Geschäftsabschlüsses) (München and
Hamburg, but in the near future all in Hamburg)

Customer Communications Division (Hamburg) and

Customer Problem Resolution (München, Köln)

Because of increased business volatility and the rising costs of permanent employees DAV
managers had recently increased the proportion of temporary employees to be more flexible during
business downturns. Business volumes were exposed to seasonal fluctuations, (business being
heaviest between September 1 and December 31) and longer-term cyclical changes in the economy.
Though overtime could meet some of the up-side fluctuation, it was becoming increasingly
important to be able to respond to business downturns by shedding labor.
One of the complicating factors in running this large, distributed operation was a corporate
mandate for “same-day” processing. DAV had, for many years, ensured that certain transactions
were performed by the end of the working day in which customers requested them. This meant a
very difficult capacity management problem—heavy days (for example, during tax-planning season)
could result in twice the transaction volume of lighter days. This surge capacity was provided with
overtime and, increasingly, by the temporary labor described above. Associates responsible for
transcribing information from hand-written forms would often need to work until 8:00 p.m. to clear
the day’s volume. The technology DAV used in the various processes was generally considered to be
on the leading edge—with state-of-the-art image capture technologies being used to store customers’
This document is authorized for use only by Rebeca Rodriguez in MGT 653-67 taught by DR. HARESH GURNANI, University of

Miami from April 2015 to October 2015.

For the exclusive use of R. Rodriguez, 2015.

Deutsche Allgemeinversicherung

handwritten forms. In spite of the high level of technology employed, manual tasks such as the setup of new customers and

policies remained stubbornly human.
Though DAV had put great effort into designing forms with clear instructions, customers
were astonishingly ingenious in finding ways to fill out forms with critical information missing, with
illegible handwriting or even without an address. Such forms (deemed “nicht in ordnung” [not in
good order]) were put aside for expert rework.

A Typical Process Flow: New Policy Set-up
Individual customers who wanted to set-up a new policy would visit one of DAV’s eighty
Zweigstellen (branch offices) or make contact with an agent. They would then fill out an
application and sometimes attach a check. The branch office then sent the application package
through company mail to the Verarbeitung eines Geschäftsabschlüsses3 (VEG) division in Hamburg.
In addition, a customer might also fill out the application at home and send it directly to a number of
DAV locations, which would then transfer it to the Hamburg operation.
Once received, VEG separated the various parts of the application, then scanned it and
digitized it. The electronic image was then retrieved from a server and delivered to associates
desktop client computer. An associate in the Hamburg VEG division typically processed about 70
policy applications during a 7.5 hour shift. The associate was responsible for entering the information
on the form into the appropriate database. If the information supplied was complete, a confirmation
notice was automatically printed and sent to the customer. If the information was incomplete, then
another associate, trained to deal with customers on the phone, would call the customer to obtain the
additional information. Associates cost DM 14.50/hour and about 12% of policy application forms
were not-in-good-order, and such forms required an additional 20 minutes processing, on average.
If the customer noticed something wrong on the confirmation notice she received, she would
either call a toll-free number or send in a letter describing the problem. The Customer Problem
Resolution division dealt with problems arising at this point. Problems at this stage were often a
result of mistakes made by associates entering information into the system, and caused considerable
dissatisfaction among customers. While it might take only five minutes of the customer’s time, it
would often take an associate over an hour to rectify it in the information system. In addition,
customers were becoming increasingly intolerant of such errors.

Quality at DAV
While accuracy and quality had always been important at DAV, the looming competition
and increasingly demanding customers meant that they had recently become even more important.
A large number of processes similar to the New Policy Set-up process took place at DAV. Prior to
1994, correct transcription from forms had been assured using a method called “Tastenbestätigung”
(Double-key entry). This involved a second associate retyping the transaction in order to cross-check
the accuracy of the initial entry. A less laborious alternative (roughly translated as “sight
verification”) was also used, in which a second associate visually inspected the transaction. In both
cases, the second associate would correct any mistakes he or she discovered. There were two
problems with this method of assuring accuracy. First, it was very expensive, since it essentially
demanded that work be done twice. Second, it was found that first-pass quality actually deteriorated
over time when the double-key method was used.


Retail Transaction Processing

This document is authorized for use only by Rebeca Rodriguez in MGT 653-67 taught by DR. HARESH GURNANI, University of

Miami from April 2015 to October 2015.

For the exclusive use of R. Rodriguez, 2015.
Deutsche Allgemeinversicherung


Schoeck, Kolb and Kluck were all troubled by the way quality was being delivered at DAV.
They often used the services of an industry benchmarking firm to judge their delivered accuracy
against competitors but doubted the information was correct:
“I know we are the best in the industry at this,” said Kolb, “but when I see
our first-time accuracy numbers quoted at 99%, I know there’s something wrong with
these peoples’ data. We’re nowhere near that good – I’m sure that, say, in New
Policy Setup, there’s closer to 10% of the forms are entered into the system with some
kind of mistake. But we need to get real accuracy numbers before we can begin to

New Policy Set-up as a Model
To begin to find out what accuracy levels were like throughout DAV, Kolb and Kluck
selected New Policy Set-up as a pilot measurement project. The plan was to take a sample of the
work carried out by the associates, and use that sample to infer what the general accuracy rate was in
the New Policy Set-up process. However, neither Kolb nor Kluck had experience with statistical
sampling, so to ensure that they developed the right kind of sampling plan, Kolb and Kluck
deployed the ultimate weapon—an academic consultant.

The Consultant
Hans-Jörg Schoss was a graying professor at a famous local Technische Hochschule, famed
for its engineering prowess and nerdly atmosphere. Kluck called to ask him how she should go about
sampling to get an accurate measure of process quality—Schoss was not very helpful.
“But what will you do with the number when you get it,” responded the
socially-challenged Schoss—rather too bluntly—in the first phone call.
“What do you mean?”
“Well, let’s say your accuracy turns out to be 72.8%, what will you do?”
Kluck didn’t know.
“Hmmm—thanks,” said Kluck as she hung up the phone.

Links to Manufacturing
The phone call to Schoss had started Kluck thinking. She wondered what would happen if
they did know the actual accuracy rate of the process. There would probably be a lot of concerned
people, but she wasn’t sure what she and the other DAV managers would do with the number once
they had it. As luck would have it, Kluck’s husband owned a small auto-spares manufacturing shop
in Augsburg, and she explained the problem to him:
“It seems to me that you are much more interested in improving the accuracy
number than knowing exactly what it is. Why don’t you use Statistical Process
Control [SPC] like we do in the plant? That way you’ll be working on giving people
the right tools to improve the quality themselves, rather than just posting a number
every six months. Why not call Schoss, and see if he can help?”
This document is authorized for use only by Rebeca Rodriguez in MGT 653-67 taught by DR. HARESH GURNANI, University of

Miami from April 2015 to October 2015.

For the exclusive use of R. Rodriguez, 2015.

Deutsche Allgemeinversicherung

The following morning, Kluck called Schoss and was barraged by an idea he had after their
previous call. “Why can’t you use SPC?” spluttered the bumbling Schoss.
The next week Schoss gave a presentation to DAV’s senior managers to introduce them to the
central ideas behind SPC, and to determine how it might be used at DAV. Schoss stressed the
importance of using SPC to measure the process rather than the people who were involved in it.
Second, he emphasized the need to have people use a tool which would become part of their
everyday job, so that quality management would also be part of that job. Finally, he stressed the need
to protect people and teams (who would at last be recording honest numbers) from the wrath of
senior managers. Such managers might—in their surprise at the real numbers—resort to the age-old
practice of finding the individuals involved and subjecting them to significant emotional events. The
New Policy Set-up process was chosen as the pilot process, and Schoss and Kluck set about working
out what would be needed for the experiment.

SPC and Right/Wrong Data
SPC had traditionally been used for continuous variables such as the diameter of a piston. In
such a process, a person would measure five components every few hours, and mark the sample
average (x-bar) and range (R) of the measurement on the chart. In the case of New-Policy Set-up at
DAV, however, things were not quite so straightforward.
A new policy request had either been entered correctly or incorrectly by an associate. Items
were not on a ‘sliding-scale.’ According to SPC practice in manufacturing, a “right-wrong” (go/nogo) characteristic

demanded the use of a different kind of chart, called a p-chart. Rather like an x-bar
and an R-chart, a p-chart also tracked the sampled performance of key measures over time, but
instead of representing continuous variables, it tracked the proportion right and wrong in each
sample. An associate would take, say, 60 items, and chart the proportion correct on a chart. P-charts
also used control limits and action limits set-up on either side of the mean of the process. Crossing
the action line (or even being on the same side of the mean for five estimates in a row) would mean
that something extraordinary had happened to the process (good or bad!) and that action should be
taken to find out why (see Exhibit 2).

Technical Issues at DAV
To carry out the experiment, Kluck decided that the New Policy Set-up group should sample
each other’s work at the end of every day. The sampler on each team of six to ten people was picked
from a hat at the start of each day.
“How many items should we sample each day,” Kluck asked Schoss.
“Well, that depends on roughly how accurate the process is in the first
place,” replied the professor, with his usual academic equivocation. “If you have a
large proportion correct, you have to gather many more samples in order to get a
good estimate of how many wrong items there are.”
“Well, let’s say the benchmark is correct and we are about 99% accurate—
we’d like to be able to sample enough that we’ll find at least three wrong items in
each sample—otherwise there will be nothing to chart.”
We have to remember, though, that it takes about five minutes to check each form.
This document is authorized for use only by Rebeca Rodriguez in MGT 653-67 taught by DR. HARESH GURNANI, University of

Miami from April 2015 to October 2015.

For the exclusive use of R. Rodriguez, 2015.
Deutsche Allgemeinversicherung


PMV Background
After mulling over Schoss’s advice, Kluck made the following decisions before the
experiment began. First, she brought in an organizational consultant, Kerstin Kober, who would help
with training and with the human issues that arose during the course of the experiment. She was
confident that the experiment would succeed and was anxious that the lessons learned be captured
and leveraged for broader deployment across DAV.
As word got out in the organization that a new quality experiment would be taking place in
New Policy Set-up, many other groups asked to be involved immediately. These included a range of
processes and departments, from the sorting of mail to the resolution of legal problems. Kluck
decided to include these groups, and to use the New Policy Set-up process as a demonstration tool.
Each group would learn about the New Policy Set-up process, decide how they would measure it
then apply the general principles from the New Policy Set-up case to their own processes. In all, this
meant that there were now 15 processes and departments involved in the experiment.

Ground Rules
Kluck’s second decision was to set up some ground rules for these various groups:

Each process within the DAV was unique and would therefore need to apply SPC
principles in an appropriate way for that process.

There should be no “process inspector” jobs. Sampling should be done by the team itself.

There should be no “punitive” elements in the process, sampling should be done so that
the specific operators who carried out the task (and possibly made a mistake) would
remain anonymous (as far as was practicable).

Management Team
A cross-level team of a dozen first-line and middle managers had been brought together
from München and Köln to discuss the PMV project, and to set the “ground rules” described above.
Erwin Weippert and Annette Kluck (DAV) and Hans-Jörg Schoss and Kerstin Kober (consultants)
formed the core management team. The development process for the experiment took approximately
eight weeks. As part of this process, the management team developed the New Policy Set-up process
as a case-study example for training. Presentations of the process were made to all senior managers, a
video-tape was developed and rough scripts were written for managers to introduce associates to
PMV. The management team then split into two, with Schoss and Weippert developing a prototype
measurement template, and Kluck and Kober working on communications and training.

PMV Roll-out
PMV was rolled out in May 1995 and managers were given a month to get their department
together, develop the initial set of measures, and build a sampling checklist. In all, 15
processes/departments volunteered to be included in the pilot (including New Policy Set-up).
Initially (and in accordance with the plan) each process checklist was very different. The launch
included introductory training sessions for first-level supervisors, along with a question and answer
This document is authorized for use only by Rebeca Rodriguez in MGT 653-67 taught by DR. HARESH GURNANI, University of

Miami from April 2015 to October 2015.

For the exclusive use of R. Rodriguez, 2015.

Deutsche Allgemeinversicherung

session with senior managers. Kluck was acutely aware that many questions remained unanswered
and that the program felt “unfinished.”
“But it’s much better to get started than to sit around waiting for perfection,”
remarked Kluck. “Besides, there is a lot of excitement around the place—now is the
time to try it.”
Some problems arose immediately. For example, sample sizes were set across the board at
200/week (chart points were entered each week to smooth out day-to-day idiosyncrasies). However,
this had been set based on a very high accuracy process, New Policy Set-up. Less accurate processes
did not need to sample as much, yet were still required to do so to maintain some consistency across
processes. This meant excessive sampling by some groups.

Making It Happen
The first two months of each team’s project had been devoted to charting data, so that the
natural variance of each of the processes could be estimated (from this, action lines could be
developed). This phase also allowed people to practice keeping SPC charts. The next phase would
involve improving the processes.

Measurement Challenges
The two months went by very quickly. The measurement phase had been started by different
teams at different times, but almost all the groups quickly grasped the idea and began to make it
work. “People learned a lot about their business because they were at least all beginning to operate
and evaluate using the same tools,” said Annette Kluck.
1. Better teams do more sampling.
A few problems had surfaced in the measurement phase. First, some groups had to increase
their sample sizes so much that the sampling process had started to become burdensome. One
associate was dismayed by all the extra work. “We are the best group around—our accuracy is
running at 98% yet we have to increase our sample size, just so we can find enough errors to measure
them! It’s just not right. I don’t see why we can’t all use the same sample size.”
2. When is a mistake not a mistake? When it’s not important.
Second, the definition of right and wrong had started to become a little hazy as the plans
were implemented. One senior manager expressed his view that getting a customer’s phone number
entered wrong should not count in making a form “bad” when sampling. “I know it sounds bad, but
the reality is that we rarely use that phone number. It’s much more serious if we get the address
details wrong. It doesn’t make any sense to put these on the same level, so we’ve started counting
forms as bad only when they have critical things wrong. I think we need to use a little bit of
judgment with this quality business. If it doesn’t matter – it shouldn’t count. I’d much rather the
associates pay attention to the important things, rather than worry about trivia.
3. Measuring lawyers.
Third, some groups had had trouble measuring themselves. The legal group, for example,
who were very keen to be involved in PMV, had had great difficulty deciding what to measure. It
seemed to be very difficult to define what was good and bad work in a legal department. It was also
difficult to measure even on a sliding scale. Since their job was to resolve customers’ legal issues, the
This document is authorized for use only by Rebeca Rodriguez in MGT 653-67 taught by DR. HARESH GURNANI, University of

Miami from April 2015 to October 2015.

For the exclusive use of R. Rodriguez, 2015.
Deutsche Allgemeinversicherung


sampling process would mean one lawyer rechecking another’s work, and making a judgment about
its quality. The lawyers were frustrated—they were very keen to be involved in the experiment, but
couldn’t seem to find a way of measuring “good” legal work.
4. Automatic charting.
Fourth, one group had suggested that their process could be charted automatically. As a
problem resolution group, their primary measure was timeliness, and since problems were logged in
and out, one associate had suggested that the Information Technology department develop a system
for putting the process-control charts on-line automatically, and avoid the burden of manual
5. On the prowl.
Finally, some very senior managers had raised eyebrows when passing the charts for some of
the more modestly performing processes. “What the heck is that?” growled one as he passed out a
chart showing a 75% accuracy rate. The associates began to worry that it might be better to select
things to measure on which performance was better, so that the charts would be less likely to
provoke such reactions when such senior managers came to visit.
Everyone looked to Kluck to provide solutions to these problems.

Motivation for Change
Frank Schoeck described the motivation behind the new quality improvement initiative:
“The reason I really like PMV is that I can’t do it for them. They have got to
figure out what is important to measure, then measure it, and then figure out if there
is a better way of doing things. We’ve had phone representatives taking calls here
whom we’ve watched like assembly workers. We can’t do that anymore – they hate
it and so do we.”
“If you get 2,000 people thinking about how they can make their job better
and they are empowered to do it and they believe in that—that’s an awfully powerful
tool. The role of management is to make people believe. If a steamroller is coming
down the street towards you, you have three options: one, don’t move and get
flattened; two, step out of the way; three, get on board and control it. PMV is about
convincing people to jump on board, management just has to provide the resources
to do it.”
Monika Volz was an associate in the New Policy Set up Group, who had previously worked
as a telephone representative. She was less enthusiastic about PMV than Schoeck:
“I’ve seen these things come and go. There’s always some new fad that
management wants to try out. There are a few people in our area that are really
excited about it all, but it’s just more of the same. Every so often, management gets
some consultant in who sells them the latest idea and we all have to go along with it.
This PMV thing is just a way of getting more work out of us: now we have to spend
time making all these stupid charts. The way I look at it, if we wait long enough,
it’ll go away, just like all the other schemes. It’s a waste of time.”

This document is authorized for use only by Rebeca Rodriguez in MGT 653-67 taught by DR. HARESH GURNANI, University of

Miami from April 2015 to October 2015.

For the exclusive use of R. Rodriguez, 2015.

Deutsche Allgemeinversicherung

“I’m worried about where this is going”, said Werner Steimle, also an
associate in the New Policy Set-up Group, “first, they’ll have us measuring in each
other, then maybe they’ll use these charts to decide who to get rid of. Who knows?”
Kluck had heard some of these sentiments around the