Monday, 27 February 2012

Operational Research and the behaviour of our fellow humans

In my blog as editor of the International Abstracts in Operations Research, I referred to psychology several times.  The postgraduate training that I followed in O.R. began with the reminder that the best O.R. teams are interdisciplinary.  It is not enough to have excellent mathematical models or spreadsheets.  The results need to be implemented.  And therefore, if the results involve people, and not inanimate objects like money, machinery or raw materials, then the implementation must reflect the psychology of those involved.  So, either the O.R. scientist needs to take human psychology into account, or, work alongside someone who can do so.

I was reminded of this today, when asked to explain what O.R. was about to a researcher whose specialism is the psychology of gender.  I spoke about two student projects that included aspects of human psychology.  Both were about call centres.

In the first case, the call centre regularly was one of the contractors for telephone orders from the supplements inserted with the Saturday editions of newspapers.  These could be for clothes, gifts, or books.  The call centre needed to be advised well in advance for the dates when "their" supplements would be produced.  So, the student asked, when do customers place orders from those supplements?
The answer is below, with an explanation.  But before you cheat, what do you think?  Was it Saturday morning?  Saturday afternoon? Saturday evening?  Sometime on Sunday?  Or Monday?

In the second case, the call centre was not a public one.  It simply dealt with calls to a financial institution from professional financial advisers making enquiries about financial products, or purchasing them on behalf of clients.  The company let us have the records of every call to the centre for six to eight weeks, letting us know the time of the call and the duration (among other things).  The student was supposed to look at the shifts and see whether there were ways that the call centre could improve its handling of enquiries and sales.  She was looking to see if queue models which assumed a Poisson process of arrivals could be used.  (Such models are ubiquitous, well studied in theory and practice.)  In the process, something strange emerged.  The average length of the calls from about 8:30am to 4:30 pm dis not vary with time of day.  But this average was less than the average length of those made after 4:30pm up to the time of closure at 7:00pm.  The difference (the latter were 20-25% longer, a statistically significant difference) was consistent over all days and all workers on the centre.  So it seemed unlikely that the variation was due to people chatting about trivia unrelated to financial products, though the student did ask about this, discreetly.  The company contact had to delve into details of what the calls were about before a good psychological explanation emerged.  Well?

Case 1: the peak demand from a supplement that went out on Saturday morning was late in the evening on the following day, Sunday, roughly between 8:00pm and midnight.  Why?  Families would tidy the house at that time, recall the supplement, and ask "Do we want to order from this, or not?"  If not, it went into the bin.  If so, then the order was placed as the parents of the family relaxed in a tidier home over a glass of wine.  By the following Wednesday, the call centre was receiving a negligible number of calls from the supplement form the previous Saturday.   Its life was over.

Case 2: most of the calls during the day were about one item of business, an enquiry or purchase.  At the end of the working day, there were often calls which covered several items.  For the latter, the financial advisors had accumulated several items of business to be dealt with in one extended call.  And once this was realised, then the staffing could be scheduled to help the callers -- and also to have at least one expert available for the call centre operators to refer awkward questions to.

Having related these two stories, I have been wondering about O.R. studies which have involved the psychology of gender.

Wednesday, 15 February 2012

Help for Stroke Victims

Operational Research made our local radio this morning -- except that those two words were not used.  There was an interview about some research work that has been carried out by the Peninsula Medical School in Exeter concerned with the rapid treatment of stroke victims.  I hadn't heard about this research (even though it is being done a few hundred yards from my home) and my attention was caught for two reasons; first, because my late father was a stroke victim, and spent the last twenty years of his life with lateral disability; second, because the speaker on the radio talked about "computer simulation".  A little tracking down on the web led to the website of PenCHORD.  There the work is described in more detail.

In essence, the story is that the chances of survival for stroke victims are much better if the victim can be given a particular treatment (clot-busting) quickly.  BUT, the treatment concerned is not suitable for all cases of stroke, and should not be given unless the victim has received tests which need to be in hospital.  So the aim of the "computer simulation" was to look at the effects of removing bottlenecks in the flow of the patient from ambulance to ward.  Our local hospital has implemented the results with great success. 

I admit that one area of O.R. that I never got into during my working life was projects with the health services.  But I am going to follow the work of this local group.  Watch this space!

Saturday, 4 February 2012

Messing with food chains

One of the less-travelled roads in an O.R. education is biological models.  Perhaps the best-known family of models is that developed by Volterra.  At their simplest, these models describe the interaction of two species, one of which is a predator on the other.  (Hence the name "Predator-prey models")  There is a nice student paper about them here.

This week, two stories related to predator-prey models came my way.  The first is a tragic story from the Everglades in Florida.  Apparently, Burmese pythons are breeding successfully in the region and eating the local mammals, fish and birds.  The newspaper story didn't mention anything that might be a predator on these snakes, and the evidence is that human intervention so far has made little impression on the population.

Looking at the picture, it is hard to imagine what might be a predator on such a large snake.  But if the biological models are to be believed, the ecosystem is heading for a new stable solution - and sadly, in that solution, there won't be as many mammals in the Everglades as there were pre-pythons.

The second story is local and much happier.  About a mile from here, the Devon Wildlife Trust (DWT) has its headquarters in an old water-mill just outside the city walls.  The mill dates from the time when the area between the city walls and the river Exe was given over to the wool industry.  Exe Island, as the area was known, was criss-crossed with mill-leats powering the mills, and wool was hung out to dry on numerous racks.  The DWT has a night-vision camera watching the mill-wheel, and recently this recorded a mother otter and young otter playfully learning the way through the water.  Otters are gradually returning to the river Exe, after years of absence.  There hasn't been human intervention aimed at the otters, simply changes to other parts of their food chain, making the fish that they need more abundant in the river.

 An otter seen by day by DWT photographer)

What happens next?  It is extremely unlikely that we will see growth in numbers of otters at the rates seen among pythons in the Everglades.  Mammals breed less profusely than snakes.  So the move to a new stable state will be slow.   And that state is almost certainly sustainable, as it will be reverting to one which was stable several decades ago before pollution affected the food chain that provides for otters.

So two stories of the food chain.  One where messing about has led to serious and undesirable ecological problems; the other, much more benign and generally desirable.

(Written as a contribution to the INFORMS challenge "O R and food")

Thursday, 2 February 2012

Cycle safety

The Times Cities fit for cycling What has this to do with O.R.? Not much, but as someone who has determined that for local journeys, cycling is the optimal method of transport, I thought that I would include the logo!