Previous month:
August 2007
Next month:
October 2007

September 2007

Operations Research: a Grand Challenge in Engineering?

Past and current presidents of the Institute For Operations Research and the Management Sciences have authored a white paper on how operations research can help make things better. The INFORMS website gives a great introduction of the paper: "Operations research can contribute significantly to the resolution of many of the grand challenges in engineering that the US and the world will face in the coming century. Most, if not all, of the challenges will require inter-disciplinary teams with experts from a variety of backgrounds each addressing different aspects of the issues. OR-trained experts are particularly adept at merging different perspectives and viewpoints and will therefore play a crucial role in the development of solutions to such challenges."

Unfortunately, many issues plague "OR" (pronounced O-R and a nightmare to include in a search on Google, which treats it as a Boolean operator), not the least of which being the difficulty to define it. And for those of you who wonder what operations research is (many do), the first sentence of the paper clarifies it nicely: "Operations researchers are trained in a range of quantitative methodologies that support the transformation of data into information for improved decision-making." My former PhD advisor, Prof. Dimitris Bertsimas, was right on target when he titled his book on quantitative decision models for business students "Data, Models and Decisions"; that defines operations research quite well (the transformation of data into decisions through mathematical models). The National Science Foundation has also identified the transformation of data into decisions as a key area for operations research through its cyberinfrastructure initiative; I wrote a post a while back on the increasing importance of information in operations research-oriented departments in engineering and business schools.

The paper falters quickly after such a promising beginning, and goes on to enumerate fields where OR can help engineers and managers make better decisions (starting with the unpopular application of yield revenue management and continuing with eye-catching, public-service areas such as health care in developing countries and in the U.S.) The applications-driven paper lacks the unifying theme that a focus on information management would have provided, and instead OR comes across as an add-on to other people's expertise - certainly valuable, but not critical. I'm not sure why anyone would want to be portrayed as jack of all trades but master of none... The authors also miss the opportunity to portray operations researchers as the center of inter-disciplinary teams bringing scientists from various disciplines together, drawing from their experience in one area to help researchers in another. When I finished reading the paper I wasn't particularly excited to be working in the field, but I give the authors credit for trying - marketing OR is an uphill battle, given the aversion to math of most regular folks, and every little thing helps. (This article in the Economist did a much better job of explaining how mathematical models appear in everyday transactions and help companies, but it didn't include the save-the-world examples that supposedly help get grant money and had a decidedly business rather than engineering flavor, but we'll keep the question of whether operations research is a business instead of engineering field for another time.)

Then I returned to my research, which is on making models well-suited to the amount of information the decision-maker has rather than forcing him to make tons of unverifiable assumptions, and to the thick volume my department has prepared for ABET's visit tomorrow to get our Bachelor's degree in Information and Systems Engineering accredited. Information-driven management isn't on the radar screen of the INFORMS leadership yet, but it's definitely gaining ground in universities and in the firms that snap our graduates. Maybe in a few years we can start discussing a new name for the profession, because "operations research" really isn't catching on.


Last semester my student Kara Stauffer and I studied the Internet strategy of newspapers (I posted a summary of our conclusions on the blog: Part I, Part II, Part III). In Part I in particular, we noted that the website of the Philadelphia Inquirer lacked pictures and videos and was too crowded with words - basically, a transfer of the newspaper on a webpage, as if the medium did not deserve its own approach. Well, finally saw the light!(Of course, we take sole credit for that.) The editors discuss the site's makeover here. Toolbars, more video and multimedia... They read our mind. It feels good to be vindicated!

A notice from the National Science Foundation

Today I received an email letter from the National Science Foundation. I knew that everyone could request the grant proposal of a funded project, but still, it feels a bit odd that someone would bypass me and contact the National Science Foundation directly to get my grant proposal, on the basis of the Freedom of Information Act. I mean, if that person wants to know about my research, it is best to read my papers, and if he wants to read funded grant proposals to see what they look like and improve his own write-ups, volunteering on a review panel will provide a much more informative experience. (The funny thing is, that proposal is not nearly as good - in terms of the reviews it got - as some other proposals to the National Science Foundation that I wrote later, but did not get funded.) So I figured I would post the letter online (minus the information on the requester), in case someone wondered what these letters say. The title of the email is "NSF Proposal #0540143."

Case #07-288F

September 25, 2007

Dr. Aurelie Thiele

Lehigh University

Dear Dr. Thiele:

We have received a request for a copy of the proposal associated with your NSF funded grant "Robustness and Performance in Data-Driven Revenue Management."  Records are available to the public on request except for material that is personal, privileged or confidential, or otherwise exempt from disclosure under law. NSF will remove, before disclosure to the requester, personal information (SSN, date/place of birth, individual salaries, etc.) in the reports about yourself or other individuals under Exemption 6 of the Freedom of Information Act (FOIA) to protect personal privacy.

The FOIA generally prohibits us from withholding other parts of proposals except to the extent that disclosure of proprietary information (trade secrets or commercial or financial information) in your proposal, if any, that you assert is privileged or confidential. Therefore, I request that you carefully review your proposal, identifying by page number, line or paragraph, the parts you consider confidential. You will need to include a detailed statement explaining how disclosure of this information would harm your organization or benefit your competitors, or why it is otherwise exempt from disclosure under the FOIA. Understand that information you submit in response to this request may itself be subject to disclosure.

Your response is not an agreement between us that the information you request us to withhold will not be disclosed. Under the law, we must decide what may be withheld and be prepared to defend that withholding in court. We will, however, notify you if we cannot agree.

If you have questions regarding this request, you may wish to contact your institutions intellectual property or grants office for guidance. If you have no objection to disclosure of the proposal -- many submitters have none -- simply let me know that.   The requester is [name and institution withheld for the blog].

Please send your response to me, preferably by email at [email removed from blog to protect from spammers], within five (5) working days after receipt of this letter. Your prompt response enables us to comply with the statutory time for response to requests.


Leslie A. Jensen

Office of the General Counsel

National Science Foundation

King of the hill

When I go to the post office one block away from where I work I sometimes get comments from the tellers I chat with: "You are a professor at Lehigh? You must be making good money." I hear a similar tune when I wait for the bus back to Pennsylvania at the Port Authority Bus Terminal in New York, and while people have not been openly, aggressively envious (just longing) they sometimes make me feel uncomfortable, although my standard line is to say that I teach at Lehigh, without any details, which does not exactly sound like I wallow in money. Around here people perceive Lehigh professors as if they were Goldman Sachs vice presidents and the rest of the world was made of janitors; that never stops to puzzle me, in part because, besides the jobless former employees of Bethlehem Steel, we do have a number of well-known companies such as Air Products headquartered in the area and in part because a number of MIT graduates (I got my PhD from the place, remember) thumb their nose up at Lehigh and wouldn't even consider working there - truth be told, they wouldn't consider working at any university whose name isn't MIT, Stanford, Berkeley or maybe Illinois Urbana-Champaign. Also, I earn a decent living because I do research and am in the College of Engineering, but professors in Arts and Sciences are not paid as much and senior faculty members who joined Lehigh when it was still only a teaching institution do not receive salary increases unless they become active in research, which has a way of cutting into the purchasing power of the many who cannot suddenly put on a research cap (and sounds as if the rules were changed in the middle of the game). So I can't quite understand the big deal the residents of the Lehigh Valley make out of my occupation, but at least people aren't actively hostile to me.

This changes when Bethlehem residents talk about Lehigh undergraduates, whom they see as spoiled brats from wealthy New York and New Jersey families. For a long time this portrait of the Lehigh undergraduate population was unfortunately quite accurate, but Lehigh has tried hard over the last few years to diversify its incoming classes and admit underprivileged students. It's too bad the rest of the town still clutches to that old "they versus us" mentality, which I will grant them is difficult not to have when you come to work on Friday mornings to an ocean of red plastic cups on the street, last relic of a well-attended party. Some students do behave like spoiled children, but the attention Lehigh undergrads, as opposed to Lafayette, Moravian, Muhlenberg or de Sales undergrads, receive in the media for every stupid thing they do, points at a systematic attempt from newspapers to take advantage of the "rich kid" cliche in order to polarize the area and sell copies.

Case in point: the sad story of two Lehigh (underage) juniors, members of the varsity swim team, who were busted for holding a loud party with lots of alcohol. Nobody was injured; the police issued many (justified) citations for underage drinking.  Alas, the story doesn't end there. The two kids, who annoyed police officers by refusing to unlock the rooms where party-goers had hidden, not only are charged with corruption of minors (although they are minors themselves and I believe didn't force anyone to consume alcohol) and related offenses, which carry a possible five-year jail sentence, but they had their mug shots plastered all over the web edition of the Morning Call as if they were murderers, until readers' outrage - expressed in no uncertain terms in the online forum - convinced the paper to replace the pictures with a photo of the house where the party was held. I am not saying what the two students did was not wrong. On the contrary, I am glad the police intervened; one of the students, also underage, had a blood level of almost four times the legal limit for people above 21 (and the girl is underage). You never know what might happen when students drink so much, holding such a loud party with lots of alcohol when so many of the attendees haven't even turned 21 yet is a very, very stupid thing to do, and the two students who organized the party definitely deserve some kind of punishment.

One hopes, though, that the officers would try interrupting the kids' self-destructive pattern before they write citations that can lead to jail time just because the kids made a misguided attempt to protect their friends. What happened to an old-fashioned: "You're already in a lot of trouble, refusing to open that door for me isn't going to make your life any easier"? Kids will do stupid things in the heat of the moment, but if you have them stop, think for half a second, take a step back, they are not idiots either - they know when they've messed up. Of course the newspapers and TV stations then had a field day with the two students, which admittedly is not the police's fault. (Lehigh students in jail! A real headline-grabber in the Valley.) But should 19- or 20-year-old kids who hold a party where no alcohol-related accident actually happened be treated as murderers by the media? It is very sad that the atmosphere around Bethlehem puts Lehigh undergrads in the spotlight like that - kids end up paying the price of residents' animosity towards their supposedly rich parents.

Responsive Inventory Management

The Wisdom of Crowds does a fine job describing the operations of Zara, a Spanish fashion company, and its just-in-time strategy in a business plagued by notoriously long production lead times (Chapter 10).  The company delivers goods twice a week rather than once per season, and "comes out with more than twenty thousand [products a year]", rather than "two or three hundred" as is usually the case. Zara's commitment to speed and flexibility translates into manufacturing in Spain rather than in foreign countries with cheaper production costs but longer transportation times: "All of Zara's store managers are equipped with handheld devices that are linked directly to the company's design rooms in Spain, so that the managers can make daily reports on what customers are buying, what they're scorning, and what they're asking for but not finding." (p.193 of the paperback edition) Who would have thought simple interactions with salespeople would be so closely monitored? It's also worth mentioning that speed comes at the cost of originality: designing a dress "often means knocking off a hot new look." In that sense, Zara's strategy might well be summarized as "wait-and-see": waiting to see which designer dresses sell, and then producing something similar. (Businessworld called it "a fashion imitator.")

Admittedly, a lot of people prefer affordable clothes than truly unique ones, and Zara's success also marks the latest installment of the push-vs-pull saga, where traditional fashion companies push products on their customers through fashion shows, while Zara's customers pull products they truly want from the pipeline. The coordination of information, explained in more detail in Harvard Business School's Working Knowledge, is admirable; no wonder business magazines rave about the company. The real question here isn't about Zara vs luxury brands, it is about Zara vs other low-cost apparel companies: why aren't competitors such as Old Navy implementing the same approach? My guess is that it's easier to build a flexible company from scratch than injecting flexibility into a well-established one; this guarantees Zara its iconic status (at least in supply chain research) a while longer.

Interestingly, even Zara's approach was found lacking by a MIT LFM student, who improved its distribution and inventory policies by using an optimization model (described in MIT News as "a sophisticated optimization model using a mixed integer mathematical program", which, given that LFM students are dual MS/MBA students as opposed to PhD candidates, probably means it was sophisticated on industry standards but not for academics.) Operations research keeps making things better, one way or the other.

Updated Stats

Today I got the 2,000th visit to my blog! The average number of views per day since the blog's inception is now in the 10-11 range, although my 30-day average (from 8/20 to 9/18) is 22 visits a day; obviously at the beginning nobody knew about the blog, which makes the lifetime average much lower. Just thought I'd take a moment to say thanks to the people stopping by!

Non-traditional revenue management

Things used to be simple: you wanted something, you paid for it. These days however, there is a growing trend away from direct payment for non-traditional goods: in a few minutes (at midnight) the New York Times will make all sections of its website accessible again to non-subscribers, after two years of the "Times Select" experiment. More and more newspaper websites are following an advertising-driven business model (see my old post here), while their print publications rely on subscribers - and frankly, given the deep discounts given to subscribers, it doesn't make sense not to subscribe if you plan on reading the magazine even semi-regularly (one year of the Economist in print, including website access, is valued at about three Starbucks tall lattes a month [cheap!], to follow the example set by Nora Ephron in a piece she wrote for the New Yorker, where she popularized the Starbucks currency).

The purpose of many websites is to attract readers who would not come to the site otherwise, and the New York Times rep said it best when he explained the company had underestimated the impact of the search engines: nowadays people do not want to wait for the late night news to get updated on a story; they require the update at that very second. The New York Times is one of the few newspapers with original reporting, and can obviously benefit from it; other newspapers might well discover very soon that enticing people to their site is meaningless if they only post Associated Press dispatches that everyone has already read - the advertisers might then demand more original reporting, and by reporting I do not mean extensions of the police blotter.

In the meantime, this unfortunately also means advertisers will become more and more aggressive in placing their products due to their increasing power; already now when you download a news video on your computer you often have to put up with a quick ad first before you are able to watch the segment, and we are talking about two-minute-long newsflashes here, not feature films. Many ads also change size and appear on top of the text of the article to make it difficult for the reader to ignore them. It would be interesting to see if readers are willing to pay extra to browse a website without all this junk; maybe they aren't - after all, "an attempt last year to let readers personalize the site [USA Today] for the news they want fell flat, with few people taking advantage of it." (source: New York Times). Maybe the initiative was too early, who knows.

The newspaper industry is not the only field promoting indirect payment (access that seems free): a big controversy in France focuses on President Sarkozy's proposition "to eliminate museum entrance fees" (see the September 2007 issue of ART News); the current fees "bring in between 150 million euros and 200 million euros" at a time where, conveniently, new revenue streams are generated through recent partnerships with a museum in Abu Dhabi (paying $1.3 billion to put some of the Louvre's collections between its own walls) and a $18 million deal to loan works to Atlanta's High Museum of Art. Francoise Cachin, former head of France's museums, summarizes the situation well when she says, about the proposition to make museum admission free: "There will be an enormous gap that the government will have to make up, either by taking money from something else or by renting out artworks for money."

Someone has to pay, one way or the other.


Models and the Real World

This post is about the August 18, 2007 issue of the Economist, which for some reason contained plenty of references to computer models, not all of them in the Finance pages.

  1. The magazine mentions a paper by David Stainforth of Oxford University and his co-authors about the challenges that arise in computer models due to "Bayesian-like prior assumptions" in climate change (quotes are from the Economist's article titled "Gambling on tomorrow" in the Science and Technology section; Stainforth's paper is available for download here.) For instance, "how quickly snow crystals fall from clouds" and "for how long they reside within those clouds [...] are two ways of measuring the same thing, so whether a model uses one or the other should make no difference in its predictions." Computer models, which "are run thousands of times, with different values for the parameters, to produce a range of possible outcomes", may use parameter values evenly spaced in an interval - a common choice when the researcher does not know where in the range the true value will fall. But in the example above, "the second parameter is actually the reciprocal of the first." In other words, the most likely outcome in the first set of experiments (the value the other outcomes cluster around) won't match the most likely outcome in the second one, although both approaches are supposed to be equivalent, because they won't have been generated for the same parameters. O joy.
  2. Turn the page and, in the same section, you can read about data-driven "hot start" of climate computer models (the insert is entitled "Tomorrow and tomorrow".) The idea, which is due to Doug Smith, a climate scientist from the Hadley Center in Exeter, England, is that, instead of seeding computer models with "plausible but invented set of values", scientists should try feeding them "real starting data" - in the experiment described here, this "reproduced what had happened over the courses of the decades in question as much as 50% more accurately." Long life to data-driven models. I'll even forgive the media a much used sentence that reads: "At least half of the years between 2010 and 2014 being warmer than 1998, the hottest on record so far" in the Economist and "Writing in Science, Met Office researchers project that at least half of the years between 2009 and 2014 are likely to exceed existing records" on BBC News. Half the years between 2010 and 2014 is only three years, likely only means "with a probability exceeding 0.5", and why is a 10-year plan publicized in 2007 cut to 2014 rather than 2017? The gold medal of inaccurate reporting goes to a journalist in USA Today, who writes of the exact same data: "The projection spans 2007 to 2017. “At least half of the years after 2009 are predicted to be warmer than 1998, the warmest year currently on record,” the researchers say in their report." (It is the last line of the abstract.) Uh, no. It's half the years between 2009 and 2014. It didn't occur to the USA Today journalist, probably used to different deadlines, that it took years for papers submitted to Science to get reviewed and published - years that might change, for better or worse, some of the conclusions established in 2004. At least we are still in the planning period!
  3. Of course quantitative models also figure prominently in the Economist's Finance pages, during a month marked by hedge funds scandals. In the lead article, "The game is up," the writer explains the situation as follows: "Quantitative funds have been hardest hit [by unusual movements in debt and equity markets]. Goldman Sachs [...] said that its funds had been hit by moves that its models suggested were 25 standard deviations away from normal. In terms of probability [...], this translates into a likelihood of 0.000...0006, where there are 138 zeros before the six." A couple of pages later, in "Behind the veil", we learn that: "The complex models that drive [long-short stock-picking in hedge funds] were upended by the extreme market volatility. Four building blocks of such models [...] usually offset each other, but when they all started suffering, the models went awry. Some of the world's biggest hedge funds all began selling the same things at the same time." The article also describes the losses of one of Goldman Sachs's equity funds, "which relies on a quantitative trading model", and of Renaissance, a "quant fund founded by [...] a prize-winning mathematician and former code breaker." It'd be interesting to hear about the losses of the funds that do not rely on computers once in a while! I doubt they're faring any better.

A bit unexpectedly (given where the money is), when it comes to computer models the Met Office's Hadley Center for climate change rather than the Wall Street pundits deserves the last word - here is how the Office debunks the myth that "climate models are too complex and uncertain to provide useful projections of climate change":

"There have been major advances in the development and use of models over the last 20 years. [...] The most advanced computer models also include detailed coupling of the circulations of atmosphere and oceans, along with detailed descriptions of the feedbacks between all components of the climate system including the cryosphere and biosphere. Climate models have been used to reproduce the main features of the current climate, the temperature changes over the last hundred years and the main features of the Holocene (6,000 years ago) and Last Glacial Maximum (21,000 years ago.)

The bottom line is that current models enable us to attribute the causes of past climate change and predict the main features of the future climate with a high degree of confidence. We now need to provide more regional detail and more complete analysis of extreme events."

The last part of the last sentence says it all.

The Tale of Two Cousins

The main page of MIT's OpenCourseWare (OCW), the free online repository of MIT courses (with lectures, exams, assignments and the like), is full of sleek graphics and high-resolution pictures of happy users all over the world. While the true impact of these resources on the education of non-MIT students remains debatable (beyond the happy feeling it gives them to be connected to MIT), and while I cannot think of an occasion where I have found OCW genuinely useful, there seems to be a consensus that OCW is just marvelous. I suspect that it is most valuable to the general public when the course package posted online includes videos, and that only professors who have to teach a similar course truly benefit from the slides. Let's face it: a lot of the concepts quickly become too difficult for the layperson (even MIT students struggle if they don't show up to the lectures and only download the slides from the course website!), and there is nothing in OCW that really emphasizes cutting-edge results. This is arguably not the purpose of a course archive anyway - if a concept had not gained some legitimacy over the years, it would not be taught. In the meantime, OCW averages one million visitors a month, the most popular topics on OCW as of this writing are introductory courses in biology and physics, and if I were looking for an introduction to any of this, I would just order a DVD from the Teaching Company rather than pore over lecture slides without a professor there to help me fill in the blanks - but the popularity of OCW is a testimony to the power of the MIT brand name in the public mind.

In contrast, look at the webpage of MIT's DSpace, an online archive for MIT's research documents (theses, working papers) made possible by the HP-MIT alliance and in a way OCW's cousin: rather drab, but containing the electronic version of most doctoral theses submitted at least a year ago. If non-researchers wanted to learn about cutting-edge advances in science and engineering, they would flock to DSpace in droves. But they don't, leaving DSpace to scientists looking for specific pieces of information in the work of their colleagues. Why? None of these developments is put into context (at least not into a context that people unfamiliar with the field would understand), and reading two hundred pages of a student's dissertation is a bit much to ask: research advisors spend a lot less time editing their students' theses than their papers (with more and more dissertations ending up on the Web, this might change) and few PhD students in science and engineering are really good writers, if only because English is not the native language of many of them. As a result, DSpace's appeal is limited to a very narrow slice of the scientific community: each file is only interesting for researchers working on similar projects, and MIT has lost a significant opportunity to draw attention on the research that isn't showcased in MIT News. Nowadays the National Science Foundation insists on every grant applicant including a section on broader impacts in his proposal, and flagship scientific journals such as Management Science and Manufacturing & Service Operations Management require a paragraph (separate from the submission) aimed at practitioners and non-specialists, describing the value of the approach presented in the paper. It is only a matter of time before PhD candidates have to explain the high-level contributions of their work to a non-technical audience, and submit the statement with their thesis if they want to graduate. Then DSpace might become a much more popular website.