Copyright: 1996
Publisher: Dorset House Publishing
ISBN: 0-932633-36-6
Robert Austin manages to sum up in this short book the observations that bothered me for 16 years working for FedEx. As a courier for most of those years, I watched the company struggle with measurement dysfunction as they attempted to get more and more productivity from their employees. The symptoms I witnessed first-hand are so clearly described by Austin that it surprises me that no one at FedEx picked the book up and fixed the widespread problem. If you doubt me, stop your FedEx courier some day and ask him what his or her goals are. They will tell you about some metric dealing with "stops per hour" or "VAN scan compliance" or some other measurable goal. What they will not mention is making you, the customer, happy.
Dysfunction in Measurement
Dysfunction as described by Austin as "the consequences of organizational actions that interfere with attainment of the spirit of stated intentions of the organization." In other words, when an organization attempts to get some result from their employees and instead get something else. Usually though, the organization gets exactly what they asked for... the letter of the intent... rather than the spirit of the intent.
This occurred quite frequently during my time at FedEx. Upper management felt they wanted to provide the best possible service to the customer. This entailed, among other things, driving down costs while keeping service levels high. Surely it is a very simple thing to measure how quickly a courier delivers packages. The faster he works, the sooner the customer gets their package and the faster the employee gets off the clock, thus both goals are accomplished. Simple right?
Well, the measurement problem comes in when you really can't measure exactly what an employee is doing. For instance, FedEx knows exactly when and where every package was delivered. You can't game that system right? Wrong. I know of couriers who would have multiple packages for a single location. Rather than deliver all packages to a single receptionist, they would deliver a few packages to several different people all in the same office. This made it look like they were doing more work than they really were.
Rather than bore anyone with accounts of this form of tom-foolery, I will instead go over some of the ideas that Robert Austin presents about why these things take place.
Measurement by proxy
One of the first things that Austin covers it the difficulty of actually measuring what you want to measure. Typically you need to choose a proxy to measure because what you really want to measure isn't quantifiable or is difficult to measure. For instance, FedEx wanted their customers to be happy. How do you measure "happy"? Surely there are components to "happy" such as "got my package on time" and "cost was low". There are also other intangibles though such as "my courier is nice to me every day" and "my courier takes time to answer my questions".
The problem with choosing a proxy then is that it rarely equates directly with what you really want to measure. Whenever you institute a measurement system in an organization then there is a danger of displacement of goals. Austin describes this concept as taking place when the focus is on the "intrumental goals" rather than the "terminal goals". The instrumental goals of course being the proxy measurements for the terminal or ultimate goal of the organization.
Informational or Motivational?
The next concept that Austin tackles is why measurements are taken. You can use your measurements to try to motivate people to meet a particular goal or you can use your measurements to see how things stack up. Merit pay, pay for performance and Management By Objective programs all make use of measurement for motivational purposes. These are the dangerous measurement systems because they are the ones that will most likely be gamed by the employees.
Informational measurement can take one of two forms: process refinement and coordination measurement. In the first, process refinement, you find information regarding the detailed organizational structure. Understanding the inner workings of processes is necessary if you are going to attempt to undertake a process improvement project. Mary Poppendieck has advocated the use of the scientific method in conducting kaizen events. She points out that Taiichi Ohno promoted the idea of "standard work" as being a baseline measurement of how things are done now. Then after a kaizen to improve the process, you compare the way things are with the "standard" and see if things got better.
Coordination measurement on the other hand is more of a process itself. For instance, in our software shop we use a kanban system for determining what work should be done next. All of our customer stories are written on 3X5 cards and every week we meet on Monday to decide what to do for the week. The customer (or in our case the customer proxy) sits down and prioritizes based on our cost estimates. We don't do formal "hours of effort" estimates but just a rough relative number that we put on the cards. We choose from the Fibonacci sequence and usually our numbers are 1,2,3 or 5. Occassionally we throw in an 8 or a 13 just for fun. The idea though is that we know roughly how many of these "units of effort" we can get done in a week. This allows our customer to pick from a pile of cards which ones he wants done this week with a relatively great assurance that the stack he chooses will be done by Friday. This is an example of a coordination measurement system.
What do you People Want?!
Figuring out what the customer wants is one of the hardest things to do in software development. An interesting point made by Austin in his chapter on "Internal Motivation" was which things the customer will tell you about that she wants. There are three categories of requests:
- Things the customer wants, but just assumes you will write because all software should have it
- Things the customer doesn't know they want but would love if you told them about it
- Things the customer doesn't expect you to just figure out but they know they want it
The customer will tell you about features that fall into category #3 but they rarely will tell you about categories #1 or #2. For instance, most customers won't mention that they want a program that won't crash and burn after 30 minutes of running. They simply expect that feature. Also, most customers who are engaged in business automation will simply ask for a "one to one" translation of their current work-flow into an computerized format. Many times there are optimizations available that they won't know to ask for but would love if you figured it out for them.
What Style of Management Works?
Austin convincingly argues that you cannot fully measure every aspect of most jobs. Given that, and given the propensity of motivational measurements to push the worker to engage only in those activities that are measured, you are bound to have some form of dysfunction. If measurement won't work, what is the alternative?
Delagatory management is the alternative that Austin offers us. In a nut shell, this style of management does not rely on full supervision or even partial supervision. Instead it relies on allowing the agent (the worker) to engage in whatever activities he or she feels is necessary to accomplish the job. The managers job in this scenario becomes less about making sure the job gets done and more about making sure the agent has what is necessary to do the job. This includes providing accurate information about what the customer desires to have.
If you are skeptical about this method of management ever working, I understand your skepticism but I encourage you to read the book. If my only work experience involved FedEx, I would be very skeptical and even more cynical than I am now! But having worked for my current employer for over a year now and having seen this style of management actually work, I am a believer in what Austin is selling in this book.
Some of the key points in using delegatory management are:
- Works better in smaller, private firms
- Cultural homogeneity contributes greatly to the success of this style
- Longer relationships in employment are crucial
- Firm members should interact often to develop working relationships
- Organizational prestige is important to the intrinsic motivation of employees
- Employees needs should be met within the organization
- There should be mutual commitment between the employees and the organization
One interesting note about this style of management is that it often should be chosen when dealing with knowledge workers. Knowledge workers have a high propensity for automating the deterministic portions of their daily work. The non-deterministic portions that are left to evaluate performance on are not usually subject to easy measurement. Thus choosing a delegatory style of management in which you pursue high levels of intrinsic motivation can be very beneficial.
Managers of Managers
Austin in the later chapters of his book deals with a number of higher level issues. One of these is when you are faced with middle-management that is acting as both the principal and an agent. These situations can easily lead to situations where a manager implements measurement systems that can easily be gamed precisely because his or her peformance is going to be evaluated by the peformance of their direct reports.
One of the anecdotes in this chapter, "The Cynical Explanation of Dysfunction", relates the story of a Japanese manager in charge of American middle managers. The American managers were clamoring for measurements and standards that they could live up to. The Japanese manager commented that it would be preferable to not set up a standard, but rather allow the managers to set their own standard. If the managers simply understood what the ultimate goal was, they would easily see what standard they needed to set for themselves and begin to work towards that goal.
Non-attributional Culture
In the same chapter dealing with the cynical explanation of dysfunction, Austin briefly covers an interesting concept. He mentions the "non-attributional culture" in which neither approbation for a job well done or blame for a job poorly done are particularly important. In this culture, what matters is the success of the group. Measurements are not the main tool that managers use but when they do use them, they are for information and not motivation. As the Japanese proverb advocates, workers are encouraged to "fix the problem, not the blame."
This squares nicely with some of the concepts in Extreme Programming in which collective code ownership rules the day. The important thing is not who changed the code or who wrote a piece of functionality or even who wrote a bug into the system. The important thing is that the software works and if it doesn't, what part of the process can be fixed to ensure that it does work in the future? Code reviews aren't harrowing experiences where programmers make fun of each others code, but rather a continuous, on-going process of continual refinement of the code through pair programming.
Conclusion
In my opinion, this book is a "must read" for anyone in management. Particularly any manager who has ever struggled with the ideas of motivation and determining performance levels. Even if you don't agree with the principles laid out in this book, until you have read this book, you are unprepared to assess performance and / or design a motivation and measurement system.