Charlie Alfred’s Weblog

Value Modeling Story 2

A few years ago, I had the opportunity to work with a division of a global medical products company. This division designed and manufactured equipment to treat a chronic medical condition.

At that time, they sold a number of similar product lines, which they wanted to consolidate into a product family.  My assignment was to evaluate the existing software architectures and assess the technical and economic feasibility of creating a common framework, which would host all the members of the combined product family.

During the assessment, I spent a significant amount of time with the technical leadership: the program manager, the software engineering manager, and hardware engineering manager.  When asked questions about their products, their answers were immediate, concise, and on the mark.  These questions addressed a wide range of subjects:

o  how the products were organized into hardware/software components

o  how the entire process worked – hardware, process software, and control software

o  what the responsibilities of each component were

o  how the components collaborated to ensure treatment effectiveness

o  how the components collaborated to ensure patient and operator safety

o  how the components dealt with differences in treatment approaches

o  how the components dealt with differences in environmental conditions 

In short, these were engineers who were intimately familiar with all of the critical details of their product, and could explain precisely how it worked.

During the assessment, I also asked a number of questions about the hospital wards and outpatient clinics where their products were used:

o  how did clients handle scheduling patients?

o  how did this translate into scheduling equipment and clinical staff?

o  how were these resources rescheduled when a service exception occurred?

o  how did they forecast how long treatment would take?

o  how did they manage the inventories of materials used in treatment?

o  how were trends in equipment performance tracked?

o  how was this data used to predict equipment maintenance?

o  how did the answers to each of the above affect the service quality and profitability of their customers’ clinical operations?

o  how did these answers affect what their customers wanted and expected from products like theirs?

In these cases, the responses generally took significantly longer to formulate and were not nearly as precise.  In many cases, their response was that these concerns fell outside of the scope of their responsibility.

In a strict sense, they were justified.  None of these things were directly related to theirs products’ primary use cases – to setup for patient therapy and provide it.  Other software products deal with patient scheduling, clinician scheduling, inventory management, and service tracking.  Their focus was on a family of medical devices, and they did not have the resources to be diffuse their efforts into these neighboring areas.

However, in a broader sense, these areas were directly related to the types of activities that their clients needed to excel at to provide high quality care in an efficient manner.  While resource scheduling, inventory management, and service diagnostics software systems are quite capable, they are only as good as the quality of information they have to work with.

For example, if medical devices at a site can report their inventory consumption, an inventory management system can do a better job of keeping track of inventory levels, and do this in a more timely manner than if separate user input is required.  Similarly, if the medical devices can monitor the progress of a treatment, they might be able to estimate the completion time of that treatment and report it to a resource scheduling application.  This enables scheduling of three groups: patients, clinicians, and medical devices.

In short, this engagement led me to wonder:

How could people who were so clearly expert in all aspects of their product, be so out of touch with the contexts where it was used?

The answer is touched on in my earlier blog entry titled Introduction to Value ModelingSpecifically, the first set of questions were “what” and “how” questions about the medical device.  The second set of questions were “what” and “how” questions about usage contexts, which meant they link to “why” questions about the medical device. 

Russell Ackoff and Fred Emery (On Purposeful Systems, Wiley 1972) address this area when they discuss the concepts of Synthesis and Analysis.  These concepts date back to the time of Aristotle.

Analysis is a three-stage process

1) taking apart the thing to be understood,

2) trying to understand the behavior of the parts taken separately, and

3) trying to assemble this understanding into an understanding of the whole.

Synthesis is also a three-stage process, but with a different orientation:

1) identify a containing whole, of which the thing to be explained is a part

2) explain the behavior or properties of the containing whole

3) then explain the behavior or properties of the thing to be explained in terms of its roles or functions within its containing whole.

Ackoff and Emery elaborate with the following succinct observation, “Analysis focuses on structure, it explains how things work.  Synthesis focuses on function; it explains why things work the way they do.  Therefore, analysis yields knowledge; synthesis yields understanding.  The former enables us to describe, the latter to explain.”

My client’s employees had done an excellent job of analysis.  They understood how their medical device worked, inside and out.  They understood how it interacted with the human body and provided therapy.  However, they had only started to scratch the surface with synthesis.  They hadn’t been exposed to their client’s clinical operation, and did not have a complete picture of how their medical devices fit in.  They couldn’t picture the difference between a busy environment and one that was less busy.  They couldn’t picture the implications of an environment containing two devices and one containing eighty.

While there were people in their organization who could see the bigger picture, they lacked the same grasp of how the medical devices worked – what their constraints were and how hard or easy it would be to improve its capabilities.  And while the two groups talked from time to time, there wasn’t enough common ground to communicate what needed to happen in a meaningful way.

Synthesis and analysis are two cornerstones of value modeling.  Synthesis enables us to view the problem from our customer’s perspective, and see what things will add the most value.  Analysis enables us to dissect our customer’s problem, model the parts, and diagnose the areas that can benefit from change.  In combination, they are very powerful tools.

In my earlier entry titled “Significance of Context” I identified five factors that would make a group of people have very similar views about value: 

o  They play the same (or very similar) role in the system.

o  They want the same small set of key benefits

The amount of value they perceive from different levels of these benefits is similar.

o  They have the same view when ranking these key benefits in order of importance

o  They are subject to similar environmental factors (e.g. constraints and uncertainties)

Let’s take a closer look at the third factor.

Economists have used the concepts of Ordinal Utility and Cardinal Utility as a means for expressing consumer preference.

Ordinal utility tries to eliminate the notion that utility is quantifiable by asking people to rank preferences.  For example, a hungry person might be asked to rank the following lunch items: Quarter Pounder, Whopper, slice of sausage pizza, tossed salad with chicken, or a ham and cheese sandwich.  Some measure of relative preference can be gotten by grouping the items and requesting choices.  For example, over the next 3 days, would you prefer a Quarter Pounder, Whopper, and salad, or 2 ham and cheese sandwiches and a slice of sausage pizza?  Combinations which leave the person indifferent (or confused?) identify equivalent value points.

Cardinal Utility seems to map a measurable quantity (such as a level of some benefit) to an abstract measure of its value, called a util.  This approach had fallen out of favor with economists for many years, until it was revived by von Neumann and Morgenstern in their analysis of behavior under uncertainty.  We will use the notion of Cardinal Utility in this post because the notion that value is a subjective perception is a key underlying concept for value modeling.

In 2003, the Software Engineering Institute at Carnegie Mellon University, published a method for software architecture analysis called CBAM (Cost Benefit Analysis Method).  This method uses a technique which is familiar to anyone who had their high school or college exam scores “graded on the curve”.  In particular, respondents are asked to pick the level of measurable benefit which corresponds to five levels of value:

F    Unacceptable      Benefits at or below this level are equally poor (as bad as it gets)

D    Adequate             Benefits at this level provide value that is barely passable

C    Satisfactory         Benefits at this level provide the expected level of value

B    Desirable             Benefits at this level provide more value than expected

A    Ideal                     Benefits at or over this level reach saturation (as good as it gets)

As a result, utility curves for fuel economy from a car might look something like this:

This artificial example shows three utility curves: one linear, one shaped like an “S”, and the third parabolic.  All three curves agree that 10 MPG is undesirable, and 50 MPG is ideal (this example must be artificial, since in the real world you’d need a pretty large group of people to find 3 reaching this much of an agreement).

What makes this example interesting is what happens between the undesirable and ideal levels.  The parabolic curve reaches the “acceptable” level with only 12 MPG fuel economy, while the s-shaped curve requires 25 MPG to reach this level.  The high frequency of “parabolic curves” in our society might explain why only one out of 5 vehicles on the road today seems to be a Hummer, Esplanade, Navigator or other tank-class SUV.

Acquiring meaningful utility values can be problematic.  Since the values cannot be directly measured, they are saturated with interpretation.  This makes it difficult to compare values from one individual or group to another.  It’s very difficult to know whether “acceptable” or “satisfactory” for one person means the same as for another.

Even so, the curves do have some usefulness.  In particular, they provide a sense of how the slope of the utility curve changes.  In the above example, the parabolic utility curve requires progressively more MPG for every increase in benefit level.  Economists refer to this concept as marginal utility (i.e. the rate at which utility changes w.r.t. benefits).

Let me illustrate with an example.  Imagine that you were blindfolded and driven in a car to a remote location.  Then you were taken for a 30 minute walk.  At any time during this walk, you might have a difficult time answering the question, “What is our present altitude in feet about sea level?”  You might estimate, but your guess could be off by hundreds of feet.  On the other hand, you might be able to provide a pretty accuate assessment of the slope of the land you just walked – mild uphill, mild downhill, steep uphill, flat, etc.  In general, people seem to be more attuned to changes, than to absolutes.

The reason that marginal utility is so important is that it highlights the places where tradeoffs are desirable (or undesirable).  When the marginal utility curve is steepest, it takes a relatively small change in benefit levels to achieve a relatively large change in value.  By contrast, when the marginal utility curve is flattest, it takes a relatively large change in benefit levels to achieve a relatively small change in value.  In other words,

Marginal Utility             Sacrifice Other to get This    Sacrifice This to Get OtherSteep                                  Generally a Good Idea          Generally a Bad Idea

Flat                                     Generally a Bad Idea            Generally a Good Idea

 

 

 

 

 

 

The investment community has a mantra known as “Buy Low, Sell High”.  For value models, the equivalent is “Sacrifice Flat to Gain Steep.”  Trade what I’m ambivalent about in order to gain what I desire.  While this explains the seller side of yard sales and flea markets, I’m not sure I can explain how it relates to the buyer side.

Advertisements
TrackBack URI

Create a free website or blog at WordPress.com.

%d bloggers like this: