Management Scientist

A Kaizen team is a Japanese term for a quality circle.  It is a group of people who work together on similar tasks who then share and implement improvement ideas.  Kaizen tools such as standard work, workflow diagrams, takt time, and pull systems are the main utilities of the team.  Each team documents the manufacturing process involving the baseline work setup which includes footprint and machinery layout.  Included in this documentation are the layout of the machine line, product flow, and information flow.   And also signals as to how workers and machines could understand when, what, and the quantity of work effort would be carried out.  Sketches of scale layouts and operator’s motions can be documented and line operation gets recorded by video.  The team next presents the results of their fact-finding efforts.

The next process for the Kaizen team is to generate ideas for improvement and work within the context of the improvements in order to make them error free.  And another goal is to locate and remove wasteful work while emphasizing solving specific problems.

Once someone told me that one of the best tools for consistently anticipating the future is a sound appreciation of the past.

American manufacturing has come from a meager colonial beginning to undisputed worldwide leadership in the mid 20th century.   And it has gone through a period of decline in the 70s and 80s with a revitalization in the complex global environment of the 90s.  It all began with a clean slate on a continent with unlimited resources and wide opportunities for development.  Americans were free to write their own rules.

Law, government, cultural practices, and social ideas were all choices to be made in the great American experiment.  These choices reflected the times in which they were made.  Anti-monarchist feelings fueling the french revolution and other conflicts in 1776 pushed America to choose democracy.  The end of an old mercantilist system was  put forth in a book called Wealth of Nations.  And in that, modern capitalism was described showing the benefits of the division of labor and the functions a self-propelled capitalism.  This push us in the direction of the free market system.

Also in 1776, Watt began the first industrial revolution with his steam engine.  And America took ahold of that first factory system and evolved it into a new style of manufacturing which led to transportation and communications technologies which sparked a second industrial revolution.  English common law which originated from Roman law was borrowed and became an American way of life.

Americans did not invent these concepts from scratch.  Instead, we borrowed them and transplanted them towards our needs.  Let’s explore some of the forces that permeate  our society in more posts.  Scientific management and holistic-systems are next topics.

Looking at callfire.com’s model today.  I think I’ll be writing code using their api this weekend.  I’m wondering how well they can service the midnight-6am timeslot with their work-at-home call center attendants.  Maybe they use the UK for that time slot?  I will check when I talk to their rep tomorrow.  The api looks rich and $2-4/hour per attendant is somewhat cost-effective for my uses.  I’d like to look at that breakdown in detail later….in terms of whether they mean $2 for an hour of availability or $2 for an attendant on the line for  60 minutes.  I would have to guess the former as the latter does not add up unless they use india.  As far as the api, I can integrate that into our call back processor to provide some more value added services that I otherwise would have had to code separately on Asterisk.  This makes callfire.com an interesting platform play.  What stands out with them is their documentation and sample code.  That is really what a platform play should provide to its core evangelists.  So take note…even if you aren’t doing anything with voice technology!

Yes, that is a metric that gets lost in the morass and confusion of many-a-business.   If you’ve been in failed tech startups, then you know what the heck I’m talking about first hand.  Now if you’ve been a part of some successful ones, you’d know there is a flip side to that.  Having at least a general gut sense of the metric here will lead to gold.  It is often labeled as a useless metric by the programming community and has even had a history of managers who’ve used it religiously in the 70s.

I count myself lucky because for the past 5 years….almost every line of code that I’ve written has had a positive net revenue generated back to whatever cause or business that was worked on.  I switched from java to python programming about two years ago after finding out that coding in python felt extremely fluid and ‘crankable’.  And in the past two years now I think i’ve written over 150,000 lines of code just for my own company pet projects.  And if I calculated the rev generated per line….it would definitely be over 2 euro on the low end.  This is the power of internet applications.  Web applications that make money.   Not code written for a hardware product to be sold on a unit basis.

Rev generated is not how much a developer gets paid per hour by a company that they are working at.  I’m looking at the actual revenue generated by the act of running the code which was written.  If you have coded a web app to sit out there as a facebook copycat but with only a smidgen of users and no revenue streams, yeah….each line of code there has generated negative net revenue.  My recommendation is to get out of there and find a place to write some rev generating code.  Coding is like movie script writing….choose what you focus on wisely.

If you own a tech company, try to keep track of this metric as a simple exercise.  Ask your VP or Director of Technology what it is… if you don’t have a feel for it yourself.  Just please refrain from acting like it’s none of your business by sticking your head in the sand like it is cool.  And don’t lett them give you that classic answer that it is a useless metric.  Metrics are metrics.  Telltales of what is going on.  If your head is in the sand on this one, I’d bet vegas odds that you’re burning cash.

Assuming your business model is not disruptive technology production/R&D and you’re not at day one of starting a company, the other business units will need to contribute even more than your technical units typically.  So take a look at what’s really wagging the tail.

I don’t mean to lecture here.  Just very adamant about this topic.

I can’t say enough about Kiva.  I joined the Stanford Kiva lending team about a year or two ago and it looks like it is really taking off.  $600k+ loaned out with a 100% repayment rate.  You see this kind of thing occurring in the the third world while subprime Americans here bankrupt the nation taking out $800K loans with a family income of $120k…..

An ex-girlfriend, who was CFO, always used to pitch me on the benefits of microloans in Micronesia.  I see the light on this one now.  And so I’ve also decided that my first iPhone app to publish to the Apple store will be a free kiva analytics app.  I’ve been pondering what to build in Objective-C for a couple weeks while quickly learning the language.  Conclusion is that this would be the perfect little project and should not be all that difficult.  Lord knows I am not doing any management science or true analytics these days at my day job….so this will free me from that rut.

The api that they use at kiva is json-based.  Check it out at http://build.kiva.org/

Stay tuned for more.  I think I’m feeling more obliged to post my ideas here in detail since I can’t really go into technical details when I post on facebook.

Here’s the overview of the military grade biometric access control system that I worked on during the iraq invasion.  I had a very good time working on this project from 2002-2004 and learned alot from my mentor, Tom Corder.  Good times at Washington DC, Oakland CA, Fort Belvoir VA, and Fort Huachuca AZ.

We all must handle numbers that we are not certain of, like projecting sales, costs or project durations.  One of my favorite methods of providing the answers to projections is Monte Carlo Simulation. But before we jump into that, let’s go over some fundamentals.

Random Variables

A random variable is a precise mathematical description of a number which one is uncertain of.  A random variable can be classified as continuous or discrete.

Continuous random variables can have any value between two extreme values.  Basically an infinite number of possible values.  Examples are an outcome of a spinner or the duration of a phone call.

Discrete random variables can only have distinct values.  A good example is an outcome of rolling a dice or the number of people who fill out a form.

Histograms

We can show uncertain variables as a shape known as the histogram.  It shows the likelihood that there will be different values.

The histogram can have any shape as long as the bars total 100%.  In monte carlo simulations, as more trials are run, there will be more bars in the graph.  They will still add up to 100%.  Basically more data and thus more accuracy in gaining a visualization of the uncertain variable.  The histogram has an average or mean of the uncertain variable located where at the balance point of the graph if you were to imagine the bars as blocks glued together.

As more data is simulated, the bars become narrower and the histogram become like a probability distribution which shows all the possible outcomes of the uncertain variable.  Each bar can be termed as a bin.

Cumulative Graph

Cumulative graphs give us the ability to identify the probabilities around given values.  In the chart to the right, you can see the histogram with a cumulative graph layered over it in green.  It has a y axis to the right of the graph showing probability.

Mean, Mode, Median

Our mean of an uncertain variable again will be the balance point in a histogram.  The mode will be the location with the tallest bin/bar.  And the median will be the bar with equal totals to the left and right of itself.

Variance and Standard Deviation

Variance is basically the degree of uncertainty.  It can be found by subtracting the average from the uncertain variable, squaring that.  And then taking the average of these values.  The square root of the variance is the standard deviation.  When talking about variance in revenue or costs, it will end up as dollars squared and so this is where the square root of that itself makes sense to use.  Hence the need for standard deviation.

Diversification and Variance Reduction

A histogram will become shaped in such a way that it goes up in the middle and down at the ends.  This happens when we take uncertain numbers and average them together.  The idea that the histogram takes on this shape is what we call the phenomenon of diversification.  The narrowness of the distribution defines the range of uncertainty.  Some concepts are that when the distribution is wider, we will see a greater variance.  And hence standard deviation and uncertainty.  On the flip side, narrower means smaller variance and uncertainty.  The narrowing is called variance reduction.

I’ll post the fundamentals on these graphical representations of uncertainty next.  These are the normal, binomial, poisson, and exponential distributions.  You should already know about the normal distribution or bell curve.  But we can bring in the Central Limit Theorem and how the concept applies to monte carlo simulations.

As an analyst, the most important question that we are facing in life today is something that all of our businesses and economies are interdependent upon. Natural resources. Oil. Gas. Energy. When will it run out? And what will happen to each of us when it does?

Step into a day as a military strategic analyst. Let me brief you on some information about what we’re doing about the what-if scenarios. Sentient World Simulation is a project involving the use of Purdue University’s extremely powerful real world simulator by the US Joint Forces Command in predicting and evaluating future events.

The system itself is called SEAS (Synthetic Environment for Analysis and Simulations). There are proposals to incorporate the world’s largest US computer research infrastructure using advanced computational tera-grid systems soon. For some added value, you should take a look at the research papers that have been generated in this line of study at SEAS.

So before doing a micro simulation of a scenario and the what-ifs using game theory…let’s talk about game theory and what it is. Game theory is the study of rational behavior among interdependent agents. We take these “agents” who represent people, companies, or countries and basically study how they interrelate and try to see what they tend to do in certain situations. Let’s reuse the term “agents” as it is game-theory speak.

Ok. When we’re talking about agents, several behaviors have been found to persist. Reward and motivation. A tenet in game theory is that if competition between agents is unfettered, no agent will get more than its added value in a game. And thus, added value itself will allow us to characterize the balance of power in the game. In other words, this characterization will lead us to an understanding of how a quantifiable pie is created and how it will be divided up.

Now that you have an idea of the pie analogy in game theory. Let’s set some more rules for the “game”. Agents in the game have a common interest in making the overall pie as large as possible. However, they have a competitive interest in maximizing their own share of the pie. The agent’s rational decisions will take into account the anticipation of their rival’s responses in the game. And since responses may be imperfect in that; like a chess game, the agents here will not always make the best moves. What I mean is that there is a measure of uncertainty within the game. And that it is a rule too.

In some follow-up posts, I will discuss some classic examples of Nash Equlibrium and what it dictates within some “games”/simulations. One of them is a classic prisoner’s dilemma in which two potential criminals are captured and one is being pushed to confess. Another is a large company against a small company in terms of whether which one should make the first move (eg, choose to be white or black in a chess game). And then we will get back into the analysis of the game involving agents who are dependent on natural resources.

In the study of probability, one tends towards studying what is interesting and noteworthy. I recently got sidetracked and had an eye-opening experience learning how to think as an intelligence analyst.

Let me share what I went through with you. There is a tool in the intelligence community called Analysis of Competing Hypothesis (ACH). It is used to assist in making judgements by weighing alternative explainations and conclusions. Basically by taking hypothesis that contain both consistent and inconsistent data…classifying that data. And then rejecting those hypothesis which have too many inconsistent data points.

In learning how to use this tool, you have to devote some time to understanding the mental processes of analysts. Thinking analytically is a skill like fishing or piloting an aircraft. It can be taught, learned and can improve with practice. But it isn’t something one can just absorb in a classroom without actually doing it oneself.

So lets talk about the cognitive limitations that impact the analyst’s thinking. The concept of “bounded” or limited rationality was introduced in the past century. Because of limitations in human mental capacity, the mind cannot cope directly with the complexity of the world. And so we construct a mental model of reality and then work with that model. We behave rationally within the confines of the mental model. But the model is not always well adapted to the real world. The weaknesses from the limitations gleamed from overall psychological experiments are insights into the study of applicable international political behavior.

The process of perception links us to our environment and is critical to an accurate understanding of the world around us. Accurate intelligence analysis then requires good perception. But the problem is that research has shown that human perception demonstrates that the process has many flaws.

One flaw is that impressions resist change. Our mindsets will tend to be quick to form but resistant to change. As an analyst, our mindset predisposes us to think in certain ways. Take the image below for example.

This illustrates a principle by showing part of a longer series of subtly modified drawings that change unnoticeably from a man into a woman. The right hand drawing in the top row, when viewed along, has an equal chance of being perceived as a man or a woman. However when test subjects are shown the entire set one by one, their perception of this drawing above is biased accordingly to the which end of the series they started from.

Basically those who start by viewing a picture that is clearly a man are biased in favor of continuing to see a man long after an observer who has seen only a single picture recognizes that the man is now a woman. Of the same accord, subjects who start out seeing a woman are biased in continuing to see a woman.

This principle explains why gradual, evolutionary change often goes unnoticed. You can relate to it by recalling any experience where a fresh pair of eyes have helped you find something that you weren’t able to find. So the analyst needs to realize this tendency to assimilate new information to pre-existing images is greater when the information is incomplete or false confidence in an established view is in play.

As an analyst and a human, we find it difficult to look at the same information from different perspectives. A good example is in the drawing below.

You can test your own persistence of established images by looking at the image above. Young woman or old? Now take a look again and see if you can both visually and mentally reorganize the data to form a different image opposite from what you initially perceive. An analyst must look at data and reorganize it into multiple perspectives.

I hope that gave you a taste of being an analyst. More posts to come in this area (including ACH).

I want to take the original post into more detail as I personally have worked with and within the US Army and appreciate the evolution of operations research from its origins. So naturally, the next question that I asked myself after answering the first one was, “What the heck were those guys researching back then?”

The following information is what I’ve gleamed from various sources describing British O.R….

Ok, the use of operations research in WWII involved intelligence, transportation, and supply.

Intelligence

Intelligence operations became a process whereby both the government and military collected and evaluated information for the purpose of discovering the intentions of their rivals, protecting themselves from their rivals and also exploting the weaknesses of those rivals.

They broke intel operations into two types:

1. Strategic or National Intel
2. Military Intel

#1 encompassed political, national security, greater economic, and social trends in the target nations.

#2 was gathered by specially trained analysts around data involving the enemy’s strengths, weapons technology, and estimated military capabilities.

So collection of intel was done through human intelligence, signal collection, and photography. Human intelligence was through the use of spies. Signal intel involved tapping phone lines and monitoring/decoding radio signals.

The intel collected needed to be evaluated for use. They combined the raw intel with relevant data in order to make it actionable.

So we’re talking about UK intel organizations here. The United States didnt have an organization until after Pearl Harbor.

Actionable Objectives

Anyways, several objectives were involved in the research. They were about conducting offensive operations, maintaining security, preserving unity of command, conducting surprise, and being able carry out maneuvers.

For example, the operations research questions for the D Day Invasion were:

1. How can we have a surprise attack at the least cost?
2. Where do we invade?
3. How do we invade?
4. How do we supply troops?

Let’s dive into each of these 4 questions in the next posts.