Monthly Archives: November 2010

robust Christmas shopping

I usually more or less finish my Christmas shopping before Black Friday, so I usually do not worry about optimal shopping strategies. I enjoyed reading about Aurelie Thiele’s latest paper on robust timing of markdowns.  Her blog post summarizes her new paper (she provides a link to the paper), where Aurelie and her collaborators propose a method for dynamically timing markdowns over a finite horizon using robust optimization models.

This got me thinking about how to reverse-engineer the process to get the best deals.  If you’re not willing to do Black Friday shopping (the getting up at 3am and waiting in long lines do not make up for the extra savings), then shopping for a few key items is like a stopping problem (e.g., the Secretary Problem), where a series of deals are offered (some at the same time, such as in the Sunday ads) and the consumer eventually decides to purchase exactly one item with a deadline of December 24.  When should I purchase the item?  I suppose it depends on how retailers adjust the prices from one time period to the next.

These are a few rules of thumb that I’ve learned (beware: no modeling or algorithms used here).  With a bad economy, I see no drawback to a wait and see approach, since sales continually improve and store coupons become available as shoppers stay home.  I have found that in the last three years or so, the pre-Christmas sales (not just Black Friday) are often better than the post-Christmas clearance sales.  So I stock up and do some personal shopping before the holidays.  However, the clearance sales become lucrative in February and insanely good in March (for clothes, at least).

How do you do your Christmas shopping?

Related posts:


shortest paths and Thanksgiving travel

My annual Thanksgiving post is going to be a short one, mainly since I suspect that many of my readers are already traveling.  This past week, my family decided to visit family in Connecticut and New York City.  We are excited to celebrate with family and to show the kids the NYC skyline, Central Park, and the Statue of Liberty.  The biggest challenge is how to find the shortest path from point Richmond, VA to Connecticut, where the shortest path is measured in hours, not miles.  The routes take us past five metro regions: DC, Baltimore, Philadelphia, Newark, and New York City, with opportunities to be stuck in traffic throughout our trip.

We initially planned to take the train.  Amtrak (train) tickets are usually a bargain, but they were much more expensive for this holiday weekend, so we opted to drive.  Given that we are driving, we have some leeway about when we leave and the routes we take.  The biggest problem is uncertainty: we’ve never driven in this neck of the woods during normal traffic conditions, let alone holiday traffic conditions.  We’ve had some input from friends, family, coworkers, and the news.  For example, AAA predicts a 12% increase in driving this year, with today (Wednesday) being the most traveled day of the VDOT recommends avoiding Interstate 95 and traveling on Wednesday. As for the specifics of the route, we can’t simply rely on GPS, since our GPS (like most GPSs) recommend a single path based on the shortest route in miles, which follows Interstate 95 the entire path in this case.  Based on the feedback we’ve received, I printed out google map directions based on our a priori belief about what the shortest path will be.  An atlas will help us to be flexible and adjust our route in the moment.  I’ll keep my fingers crossed.

Have a great Thanksgiving!


geometry becomes relevant during the planning of a football game

I was looking forward to the NCAA football game between Illinois and Northwestern University at Wrigley field.  On Friday, a number of safety issues with the field turned this game from a sports story to a geometry story.  It was pretty cool.

This was the first football game played at Wrigley Field since 1970, and the first college football game at Wrigley Field since 1938.  As it turns out, there are a few reasons why Wrigley field is not a good venue for football.  Back in the day, the field was oriented North-South (go here for a picture).  On Saturday’s game, the field was oriented East-West, since permanent seating has been put in place since the last football game played in Wrigley Field that would make a North-South orientation infeasible.

Those in charge of planning how to safely align the football field inside of Wrigley were apparently not good with geometry (or they procrastinated).  Despite more than eight months to figure out how to orient the field, a last-minute change was announced on Friday, since the football field apparently could not be safely oriented within Wrigley Field: the teams would primarily use one end zone.  Earlier, it was announced that both teams would share one of the sidelines, which made for a few logistical issues when teams move across the field during the course of the game.  The change was announced less than a day before the game on Saturday, and it caught many off-guard.  I don’t have an aerial shot of Wrigley Field, but the image below shows how close the East end zones is to a wall (click here to see the West end zone).  NCAA regulations require a minimum of six feet between the sidelines and a wall (twelve feet in roomier stadiums).  Several good pictures of the field are posted here, which give a good sense of the cramped space within Wrigley Field.

This game is usually inconsequential (although it’s supposed to be a big rivalry), since usually both Illinois and Northwestern are not very good at football.  This year, the excitement surrounding the game was a pleasant surprise, which was caused by the fact that both teams are likely heading to bowls in addition to playing the game in legendary Wrigley Field.

In case you missed the game, Illinois (my alma mater) won by a rout.

more men attended the work-life balance session than women attended the social networking session

I am not an expert on the status of men and women in OR/MS, but I have two observations to offer.

1.  I was surprised to see that no women attended the social networking session at INFORMS (three women were on the panel).

2.  I was surprised to see that several men attended the work-life balance session at INFORMS.  Better yet, they even asked questions.

One step backward and one step forward. Certainly, there are gender disparities in blog readership (more men read blogs, particularly technical blogs), blog commenting (I’d wager than men are more likely to leave comments), twitter participation (more women tweet).  This might translate into disparities of who feels comfortable attending a social networking panel. (Note: by disparities, I literally mean differences, with no judgment attached).  I wonder about disparities because quite a few women asked me about the session during the conference (I received more inquiries from women than men) so I know that they were interested but didn’t show up. I don’t have a way to draw conclusions from my small sample of observations.

I’ll stop there before I put my foot in my mouth.  What do you think?

Related posts:

External links:

The social networking panelists at INFORMS

The social networking panelists at INFORMS courtesy of Anna Nagurney

half baked operations research

Jon Caulkins from Carnegie Mellon University gave a provocative talk at INFORMS about the policy implications of Proposition 19 in California, which would have decriminalized marijuana had it passed.  Prior to the election last week, about one third of voters opposed Proposition 19, one third supported Proposition 19, and one third were undecided.  The undecided voters could swing the election either way.  The swing voters seemed to respond to economic impacts of Proposition 19.  Since California is just about broke, Proposition 19 becomes very appealing if it reduces incarceration costs, increases tax revenues, and reduces drug cartel violence in Mexico (as its supporters have promised).

Jon Caulkins worked with RAND to unravel the implications of passing Proposition 1, and he presented his findings at INFORMS.  This talk was extremely well-attended, which wasn’t all that surprising.  I came to see an interesting talk about public policy, and I was a little curious about why others came.  (Statistic:  non-drug dependent college graduates represent 7% of the marijuana market).

Caulkins reported six key findings in the RAND report (called “Altered State”) .  There was nothing half baked in his analysis:

1.        The pre-tax price of marijuana would fall at least 80%.  This conclusion was obtained by creating would-be business plans for growing marijuana.  However, the (post tax) price that consumers would pay would be significant.

2.       Given the high tax on marijuana, tax evasion could be a huge problem, which would lower the amount of tax revenue that California could collect.

3.       Marijuana consumption would go up, but no one knows how much.  An increased consumption could have several causes.  First, many have studied the price elasticity of marijuana, but they are all based on the current, high prices.  Second, marijuana could lead to product diversification that we see with many other products.  For example, it could be bundled with, say, brownies and beer.

4.       The criminal justice expenditures on marijuana are exaggerated.

5.       Tax revenues would be lower than the $1.4B CA Board of Equalization estimate as a result of tax evasion (see #2), lower prices leading to lower taxes (see #1), and a greater potency from legal growers leading to consumers needing less marijuana for their fix.

6.       The reduction of violence in Mexico and dollars flowing to Mexico is largely biased.  California represents about 1/7 of the marijuana consumption, and while it would make some improvements, there would still be a huge demand for (illegal) marijuana in the other 49 states.

Points 3 – 5 would result in fewer economic benefits for California.  With fewer economic benefits, there was in an immediate and significant decline in voter support after the release of the RAND report (I can’t find a picture of the polling data as a function of time, but if you find it, please post it in the comments).  Caulkins great OR work likely affected the outcome of the election, which is pretty exciting.

I wish there was more good OR to help me make better decisions in the voting booth.  Have you ever changed how you were going to vote based on math?

OR and the intelligence community

The (security) intelligence community (IC) is mainly made up of people with a non-technical background (such as political science majors).  Although the IC makes tough decisions under uncertainty given limited information (one of the things that OR does so well), the IC manly uses intuition and expert judgment to make decisions.  Ed Kaplan, who gave the  Philip McCord Morse lecture at INFORMS 2010, spoke about the intelligence community and the role that OR needs to play in the IC.

There are clear needs to use advanced analytical methods in the IC.  The NSA, for example,  collects 1.7B emails per day.  How do they determine which emails to read and analyze?  Clearly, they only have the resources to read a few.  How should they allocate their resources between collecting new data and analyzing the data?  The good news is that the NSA hires OR people (they even have a summer program for graduate students in OR), and the CIA is starting to hire OR people too.

Kaplan reminds us that intelligence is not just an operations problem.  People with language and cultural skills are needed to interpret the nuances and make tough judgments that no algorithm can do.  Human agents have been highly successful at using non-analytical techniques at detection suicide bombing attempts in Israel, which resulted in a plummeting number of successful suicide attacks.  OR cannot replace the human element, but it can certainly aid it.  And often it’s routine law enforcement–not intelligence–that makes the difference.

There have been some attempts at using OR to solve intelligence problems.  Ed Kaplan introduced two obscure articles that were a lot of fun.  A 1967 article in a classified CIA journal (unclassified in 1994) applies Bayes rule to the Cuban missile crisis.  The paper uses the input from 200 agents in the field that reported that they believed that the Soviets were up to something in Cuba, even though none had visual evidence.  The odds started out as 10-to-1 against the Soviets building missiles in early 1962, and slowly increased to 3-to-1 and then 50/50 (1-to-1) based on successive applications of Bayes rule.

J. Michael Steele published a paper in Management Science entitled “Models for Managing Secrets” based on the Tom Clancy Theorem, which states that the time to detect a secret is inversely proportional to the square of the people that know it (stated without proof in the Hunt for Red October).  The paper applies Poisson processes to “prove” Tom Clancy’s theorem.  It also analyzes countermeasures according to how they would result in a secret being kept for longer.

Kaplan has written a few papers on intelligence, including his recent paper Terror Queues, which applies queuing models to determine how agents (servers) can interdict (serve) terrorists (customers) before the terrorists complete an attack (leave the queue).

Kaplan finished the lecture by encouraging researchers to continue to examine the many problems in IC from an OR perspective.  How do you think OR could be used by the IC?


The Social Network: The OR version

Social networking today is not going to look like social networking in a year from now.  What is the role for OR given the ever-evolving social network landscape?  They concluded that if you believe in operations research, you will find enough outlets to publicize OR.  This issue and many others were discussed in a panel discussion at the INFORMS Annual Meeting on social networking.  The panelists included Anna Nagurney, Wayne Winston, Aurelie Thiele and Mike Trick.  It was the most fun I have ever had at a session, and I am indebted to the panelists for their wonderful insights.

All of the panelists are bloggers, and blogging is a great outlet for both publicizing OR as well as educating the general public.  We continue to blog because it’s fun.  Aurelie also mentioned that she was motivated to start blogging in order to address the comments made anonymously during the peer review process.

Blogging is certainly constrained by social expectations (such as keeping posts reasonably short), but it allows us to write about an eclectic mix of topics in the same place while gathering feedback from our readers.  It doesn’t feel so constrained, since we could break up a long, contemplative essay into several shorter blog posts with little effort.  Blogs can adapt and change with us, and they have truly served OR well.  In the foreseeable future, blogs appear that they will be the social network du jour for operations research.

One of the major themes that we discussed was credibility.  Credibility in the blogosphere is different than credibility in the classroom.  On the blogosphere, no one cares if you are a full professor, but being able to make logical, linear arguments to make a point buys you a lot of credibility.  Wayne Winston argued that if you can step readers through your argument using simple algebra, even better (his blog post on Belichick’s decision to go for it on 4th down last year is an excellent example of a simple blog post that oozes credibility).  Credibility came up throughout the discussion as we discussed why we do not use “LOL” in tweets and how we are often asked to comment on work that has not been peer reviewed.  In the latter case, we may feel the need to educate our readers about being able to recognize more credible sources when they read the news.

Panelists had mixed feelings about twitter.  We all seemed to prefer blogging to twitter, since it is hard to limit ourselves to 140 characters (we are all professors, after all), but we noted that twitter has its advantages.  Twitter is good at building a brand, generating an audience for our blogs, and for reading other articles, but it’s not a tool that should be used in isolation of other social networking tools.  Both Aurelie and I prefer the twitter links to actual news articles to generic RSS feeds, and we both read more news using twitter.

None of panelists were too keen on LinkedIn.  Wayne Winston expressed his desire for something to come along and destroy LinkedIn, since it seems like it could do many things better.  The consensus was that LinkedIn is too transactional and doesn’t offer many social network benefits.  It reposts blog and twitter posts, where LinkedIn accepts comments from other LinkedIn users that are not fed back into the blog. This leads to disjoint sets of user comments to the same blog post.  I was relieved, since I thought that I was the only one who didn’t really get LinkedIn (but if you can teach me, join my network and show me the way!).

In the end, we all hoped that there would be a “linkedIn Killer” that would do what LinkedIn is trying to do but better.  This network would not only link CVs but blogs, white papers, essays, editorials, and paper summaries.  The network would ideally be organized according to social network (“friends”) and according to topic.  On a related note, panelists felt that a few TED talks about OR or analytics would help to publicize our field while contributing to a collection of talks that will probably outlast most of the social networking tools we are using.

There are certainly things that we can do better as OR ambassadors to the blogosphere and other social networks.  The benefits for a strong OR social network.  It’s a work in progress.  We are evolving as are the tools we are using.  In the end, we hope to better publicize our field, recruit the next generation of OR researchers, and educate the general public on timely issues using math.

External links to other bloggers on the panel (and other bloggers), in true Roshoman fashion: