Waiting for a company that can not plan

Today, I am at home waiting for TDC – Denmark’s national phone company – to stop by and connect an ADSL connection. They indicated that they would be here between 7:30 and 16:00. When I asked them if they could be a bit more precise they said they were unable to do so.

It surprises me that this company is running an operation that is focused on doing routine jobs at client’s home addresses is unable to plan those visits better and thereby increase the customer satisfaction rate. However, I doubt whether they are unable to do so or simply not willing to do this.

Wouldn’t you assume that since the company is private capital owned (KKR among others), the owners would campaign for efficiency also in this operation? I assume that they visit the clients that they promised to visit during a day. In my view you can do this in two different ways:

1. There are large margins in a service professional’s daily schedule so that he/she has little problems fulfilling the assigned jobs during a day and he does not have worry about the optimal route from client A to client B. In other words, the system is highly inefficient.
2. The system is efficient and the service professional’s daily schedule is tightly planned.

I assume that the second scenario is the case, but why they cannot inform their customers better is a question that remains to be answered. I believe that they are simply not willing to; behavior you see more often with former monopolists.

Patenting for society

In a number of European countries we have so-called pirate parties elected into the countries’ and continent’s parlements. The pirate party movement was born due to a frustration of many – especially younger people – with the outdated legal framework for intellectual property. Most people will directly think about music piracy or illegal downloading of music. However, the implications of the intellectual property laws are beyond just music and are becoming a serious problem for important aspects in society such as: health.

Often the pharmaceutical industry is criticized for asking a fair amount of money for the products that they are selling to society: medications. Fact is that many so-called blockbuster products need to compensate for basic research or products that never yield anything. Thus, the compensation that the pharmaceutical industry receives for their products might actually be a just fair return; when you look at the entire portfolio.

One of the areas where they ceased development of new drugs is anti biotics. This is due to become a serious problem for society at large. Anti biotics are widely used in the healthcare system and more and more people become resistant to the existing cures, all with the consequence that we soon run out of antibiotics that do not work anymore!

Developing new antibiotics takes a lot of effort and has a high degree of failure and therefore is not considered to be a value adding activity by many pharmaceutical companies. The main reason why this is not value adding is because the companies have too little time to generate money from the invested capital before the patent expires and price drops due to entrance of copy-cat generic companies. Fortunately, this can be changed by reforming the intellectual property laws. The reformed intellectual property laws should differentiate in durations for patents so that they can be used as to provide the necessary incentives to innovators again. Aspects that should be considered in defining patent durations are things like:

1. General payback time in the industry the patents is filed for. For example, there is a huge difference between technological developments in pharmaceuticals in contrast to developments in information technology.
2. What incentive for innovation is needed in a particular industry.

How important is retention for performance?

In my last post I outlined that the winning team in competitive battles is build on people, computers and superior procedures of interfacing between the two. I even highlighted that in order to achieve outstanding performance teams do not need the best people. This might be a somewhat controversial statement as almost every organization from time to time claims that people are there most valuable asset. This might be true to a certain extend, but processes might be more valuable.

The management consultancy industry shows this very well. The staff turnover in this industry is very high. Every year many recent graduates from business or engineering schools are hired and put to work in advising senior management of their clients. This is not value adding because of the IQ of the consultants but the firms models they trained to use; hence, processes and human interfaces to those processes. Also note that the consulting firms like McKinsey, Bain, BCG etc can replace its consultants relatively easy with recent graduates.  In other words, when the value adding part of the business is in the processes and interfaces to those processes the retention of people is not all that important for the organization’s long term performance. Of course, people the retention of people that interface well with the processes may work to the firm’s advantage but what I like to say is that they can be replaced.

Thus, in order to secure long term performance organizations would better invest in processes and training its people in interfacing with those processes. As discussed in my previous post those processes could very well be based on computers. How the interfaces between the people and the artifacts should be depends on the particular circumstances of each organization. However, in general we can say that it is the objective that the human focuses on intuition and creativity and that the artifacts ensure that the human’s reasoning are error-free. With error free, I mean that they do not see obvious flaws that can have major consequences from the firm’s performance. In chess terms this might be giving away a Queen by mistake.

The winning team: Men and Machine

The winning team: men and machine. While that is a powerful combination a critical aspect is missing and that is a superb way of using the two together. Grandmaster Gary Kasparov highlights this in an article the New York Review of Books. I found this article via a post by Professor Andrew McAfee.

The story line of Mr. Kasparov’s article is about a period a bit longer than a decade during which he went from beating chess computers without too many problems to the point where he had no chance against them. Over the scope of this decade the computers became soo much ‘better’ (read: more powerful to calculate everything) that it could compensate for the absence of human intuition and creativity. This resulted in machines that become unbeatable by humans.

The most interesting part of the article is that he describes chess tournament, which was helt after he was made ‘irrelevant’ by computers. The appealing aspect of this tournament was that the competitors were free to create teams of humans and computers; by now desktop based chess computers became powerful enough to beat most humans (including grandmasters). The participants of the tournement included both grandmasters and amateur chess players assisted by computers. Let’s quote Gary Kasparov about the conclusion of this tournament:

“The surprise came at the conclusion of the event. The winner was revealed to be not a grandmaster with a state-of-the-art PC but a pair of amateur American chess players using three computers at the same time. Their skill at manipulating and “coaching” their computers to look very deeply into positions effectively counteracted the superior chess understanding of their grandmaster opponents and the greater computational power of other participants. Weak human + machine + better process was superior to a strong computer alone and, more remarkably, superior to a strong human + machine + inferior process.”

The interesting take-away of the whole article is that average people in combination with machines and good methods of interfacing between computers and people are performing better than the competition. In the summer of 2008 I was introduced to advanced planning and scheduling software solutions that are used by some companies in the maritime industry; in fact this type of solutions are used in many industries where you find complex planning and optimization problems. During the evaluation process of these solutions, I have seen three type of solutions that organizations deploy to maximize the performance by planning.

The first approach is the traditional approach in which human planners are utilizing their ability to take advantage of their intuition and ability to make judgements on the basis of casual dependencies, which are in many cases are not mathematically defined. The planning and scheduling is often done on a whiteboard or in a spreadsheet program. To revert to the chess metaphor: grandmasters competing which each other is the best-of-breed you get in this solution.

The second approach is a more what I would call “hit-the-button-calculators” (MarinTek, Route66). This approach basically works as following. The humans line up the input of the model and once they are ready they hit the button and algorithm outputs the optimal planning. To many users this is somewhat of a black magic. Furthermore, they have little ability to alter the output without making changes to the input and running the algorithm again.  To revert to the chess metaphor: this can be looked at as the Deep Blue (IBM’s chess computer that won against Kasparov).

The third approach I evaluated was an interacting approach between human intuition and brute calculating power (Quintiq). The main idea behind this solution was that humans could make changes and see immediate response from the algorithm. This approach facilitates a iterative cycle when it comes to creating schedules and adjusting them when the knowledge of the future becomes more precise or unexpected events occur. Of course this type of solutions include a “hit-the-button” function, but if I recall correctly none of their customers ever used this.

After reading the above quoted text, I realize that they will never use it either. Firms that want to lead the industry by creating sustainable competitive advantage need to combine human intuition and creativity and brute computing power to eliminate human errors. Companies that manage to find the right interaction between the two will outperform their competitors that depend solely on humans or computers, or those that depend on the combination but did not manage to get the interaction between the two right.

Reading online

In my previous post I wrote about e-readers. Newspapers and magazines are obvious types of content that will be read on this type of devices. However, discussing how newspapers/magazines will be read on e-readers is not the topic of this post. The only thing I have to say is that I am glad to see that there is finally some serious innovation happening in the way we are reading newspapers only. To be honest, reading a website has never been very pleasant. The pages are cluttered with adds, you always have the feeling that you read a quickly put together presentation of the content. I am one of those persons that cares about presentation when I read something (that is why I prefer PDF over websites). Anyway, recently The New York Times launched an Adobe AIR based application that presents their newspaper in a nice way, which I believe will even be nice to read behind a computer screen. I hope more newspapers, magazines and other content owners will follow this example and create similar products.

Read more about the Times Reader.

Reading and the enviroment

Over the past few weeks I have been looking a bit into e-reader devices and the wide range of opportunities they offer. I am not sure whether I am ready to buy one yet, but the notion of electronic paper attracts me. I believe I would enjoy reading the blogs and news sources I am following away from the computer on such a device. But also the uses in professional environments attract me. I am amazed how many prints we (me and my girlfriend) are making a week with stuff we would like to read. It just surprises me over and over again. It would be great to print them to an e-reader so we can read it whenever we want.

The following video is good to see some of the applications that such e-readers offer for us humans and for the environment.

How to do a Monte Carlo simulation with QlikView

Over my summer holidays I read three books that I can recommend to everybody working with uncertainty and risk in the broadest sense of the definition. The books are the following:

1. “Flaws of Averages” by Sam Savage
2. “The Failure of Risk Management and how to fix it” by Douglas W. Hubbard
3. “How to Measure Anything” by Douglas W. Hubbard

Following reading those books I started a discussion in a LinkedIn Group about implementing some of the lessons of the above mentioned book with the help of the business intelligence software I am using at work: QlikView. After some back ‘n forth between the participants of the discussion I have created a simple draft proof-of-concept. In this post, I would like to describe the core principles that I implemented in the proof-of-concept Monte Carlo analysis that I developed with QlikView. A detailed description of the model and its context can be found in “How to Measure Anything” by Douglas Hubbard (p. 74-78). However, in brief the model deals with the lease of a piece of equipment that should result at least in a 400,000 cost saving in order to break-even.

Before I continue, I would like to introduce QlikView to those of you who are not familiar with it. QlikView is so-called in-memory Business Intelligence platform that is available as a desktop application for Windows (works fine via VMWare Fusion on the Mac) and there is an complete range of server products for Enterprise implementation. The beauty is that BI tools developed in the desktop environment can easily deployed on the server and therefore be used enterprise wide.  Last but not least, from QlikView 9 onwards you can download a copy and play within for free; the only problem is that you may have some difficulties exchanging files with colleagues and friends (buying a license will resolve this).

I set myself the following success criteria for the Monte Carlo application I want to build in QlikView:

• Users should be abel to insert values through the user interface;
• Based on the new user input and predefined model, the application should be able to generate a set number of new scenarios.

The user interface

The user interface has two component. First, the input section which you can see in the left column in the enclosed image. Here the user can define the lower and upper bound of the 90% Confidence Interval (Hubbard, 2007 and 2009). Second, the column on the right provides some basic information about the results of the scenarios. The entire data set is available so you could also think about creating histograms and many other reports that allows you to analyze the results of the simulations. The report with the title “Risk of not break-even” shows  what the chances are based on estimated intervals defined on the left that his project will break-even.

The script and work-flow

In order to start a simulation, the user needs to define his/her estimated cost saving intervals. After those are completed the script needs to be reloaded. Depending on the number of simulations you would like to perform this takes from a few seconds (e.g. 10,000 scenarios) to a few minutes (e.g. 5:01 for 10,000,000 scenarios).

The script is short and relatively simple. Everybody with a basic understanding of SQL and Excel formulas should have no problems understanding it. The basic aim of the script is to create a table with all the scenario’s variables. The script has three parts:

1. Loading the variables as defined in the user interface and calculating the mean and standard deviation for each variable; four in total.
2. Generate the scenario’s variables on the basis of the in the previous part set standard deviation and mean.
3. Include the ‘business logic’ in the model and calculate the overall cost savings; this could also easily be done in the user interface and therefore does not necessarily have to be done in the script.

I have copied the script at the end of the post.

Final remarks

This is all that is needed to setup a basic Monte Carlo analsyis in QlikView. More advanced and business realistic models can be build based on those principles. QlikView has a very strong in loading external data sources (that is what a BI platform is all about after all) and has an amazing in-memory calculation engine. The only draw back I can see at this stage is the limited number of probability distributions that QlikView has build in. However, expanding this should not be a big deal for QlikTech once we identified which distributions are relevant. All in all, QlikView seems to be a powerful platform to perform Monte Carlo simulations with.

I am looking forward to hear your responses to the concepts outlined above and please leave a comment or send me an e-mail if you have any questions. Unfortunately, I am unable to upload the QlikView file that I created to the WordPress server but if you would like to have a copy please shoot me an email and I will send it to you in an e-mail.

Last but not least: The script

//Loading the variables as defined in the user interface and calculating the mean and standard deviation for each variable

LET MaintenanceMean = ($(#Maintenance savings – Lower bound) + $(#Maintenance savings – Upper Bound)) / 2;

LET MaintenanceStDev = ($(#Maintenance savings – Upper Bound) + $(#Maintenance savings – Lower bound)) / 3.29;

LET LaborMean = ($(#Labor Savings – Lower Bound) + $(#Labor Savings – Upper Bound)) / 2;

LET LaborStDev = ($(#Labor Savings – Upper Bound) + $(#Labor Savings – Lower Bound)) / 3.29;

LET RawMaterialsMean = ($(#Raw Materials Savings – Lower Bound) + $(#Raw Materials Savings – Upper Bound)) / 2;

LET RawMaterialsStDev = ($(#Raw Materials Savings – Upper Bound) + $(#Raw Materials Savings – Lower Bound)) / 3.29;

LET ProductionMean = ($(#Production Level – Lower Bound) + $(#Production Level – Upper Bound)) / 2;

LET ProductionStDev = ($(#Production Level – Upper Bound) + $(#Production Level – Lower Bound)) / 3.29;

// Generate the scenario’s variables on the basis of the in the previous part set standard deviation and mean.

MonteCarlo:

REPLACE LOAD

RecNo() AS Instance,

NORMINV(RAND(),$(MaintenanceMean),$(MaintenanceStDev)) AS “Maintenance Savings”,

NORMINV(RAND(),$(LaborMean),$(LaborStDev)) AS “Labor Savings”,

NORMINV(RAND(),$(RawMaterialsMean),$(RawMaterialsStDev)) AS “Raw Materials Savings”,

NORMINV(RAND(),$(ProductionMean),$(ProductionStDev)) AS “Production Level”

AUTOGENERATE(10000);

//Include the ‘business logic’ in the model and calculate the overall cost savings; this could also easily be done in the user interface and therefore does not necessarily have to be done in the script.

LEFT JOIN (MonteCarlo)

LOAD

Instance,

ROUND(([Maintenance Savings] + [Labor Savings] + [Raw Materials Savings]) * [Production Level]) AS “Annual Savings”,

IF( (([Maintenance Savings] + [Labor Savings] + [Raw Materials Savings]) * [Production Level])<400000,’No’,'Yes’) AS “Break-Even”

RESIDENT MonteCarlo;

//Loading the variables as defined in the user interface and calculating the mean and standard deviation for each variable

LET MaintenanceMean = ($(#Maintenance savings – Lower bound) + $(#Maintenance savings – Upper Bound)) / 2;

LET MaintenanceStDev = ($(#Maintenance savings – Upper Bound) + $(#Maintenance savings – Lower bound)) / 3.29;

LET LaborMean = ($(#Labor Savings – Lower Bound) + $(#Labor Savings – Upper Bound)) / 2;

LET LaborStDev = ($(#Labor Savings – Upper Bound) + $(#Labor Savings – Lower Bound)) / 3.29;

LET RawMaterialsMean = ($(#Raw Materials Savings – Lower Bound) + $(#Raw Materials Savings – Upper Bound)) / 2;

LET RawMaterialsStDev = ($(#Raw Materials Savings – Upper Bound) + $(#Raw Materials Savings – Lower Bound)) / 3.29;

LET ProductionMean = ($(#Production Level – Lower Bound) + $(#Production Level – Upper Bound)) / 2;

LET ProductionStDev = ($(#Production Level – Upper Bound) + $(#Production Level – Lower Bound)) / 3.29;

// Generate the scenario’s variables on the basis of the in the previous part set standard deviation and mean.

MonteCarlo:

REPLACE LOAD

RecNo() AS Instance,

NORMINV(RAND(),$(MaintenanceMean),$(MaintenanceStDev)) AS “Maintenance Savings”,

NORMINV(RAND(),$(LaborMean),$(LaborStDev)) AS “Labor Savings”,

NORMINV(RAND(),$(RawMaterialsMean),$(RawMaterialsStDev)) AS “Raw Materials Savings”,

NORMINV(RAND(),$(ProductionMean),$(ProductionStDev)) AS “Production Level”

AUTOGENERATE(10000);

//Include the ‘business logic’ in the model and calculate the overall cost savings; this could also easily be done in the user interface and therefore does not necessarily have to be done in the script.

LEFT JOIN (MonteCarlo)

LOAD

Instance,

ROUND(([Maintenance Savings] + [Labor Savings] + [Raw Materials Savings]) * [Production Level]) AS “Annual Savings”,

IF( (([Maintenance Savings] + [Labor Savings] + [Raw Materials Savings]) * [Production Level])<400000,’No’,'Yes’) AS “Break-Even”

RESIDENT MonteCarlo;

Future of AppleTV

In a recent post on Mashable.com the future of the AppleTV was discussed; following the relevant paragraph from their post:

“In the meantime, Apple TVs themselves are cheaper then ever, but Apple is strangely quiet about it. Is Apple TV the future of home entertainment or is it Apple’s blunder which will probably disappear from the market in the years to come?”  

Let us assume that the AppleTV is not a huge success, what can Apple do to remain in the living room entertainment? One option would be to upgrade the AppleTV and introduce it at MacWorld in January. But what would an upgrade look like. The functionalities of the current version of the AppleTV are limited and it is only a supplement to existing living room hardware and not a replacement. Apple could think about introducing a full living room solution, thus introduce the next version of AppleTV that includes a TV and radio tuner, a DVD player and the ability of using computer displays as a screen. This might drive the sales for Apple’s Cinema Displays and displays from other suppliers. Thinking crazy, a AppleTV solution based on the iMac could also be great solution.In other words, Apple has all (except of a TV tuner) the technology in its product portfolio to make a great product!In relation with this post, also see my earlier post about an alternative Apple living room strategy.

Google maps – not a map service for bicyclers

Without any doubt we can say that Google Maps is one of the best available maps services. Unfortunately, the suggested routes are aimed at car users and not at bike users. I often want to use Google Maps to find some directions to go somewhere. Unfortunately, I used a bike in most my commuting within the city (mainly Copenhagen in my case) and Google Maps provides routes that are optimized for car users. In other words, most of the directions I get from Google Maps are useless for my purpose. I would love to see that Google builds in an algorithm that can provide me with a bike route if I like.   

How I would have entered the living room if I was CEO of Apple Inc.

A quick Keynote play about some thoughts I have had for a while about how I would have entered the living room if I was in charge of this strategy at Apple Inc.