by Joe Francica
Last week was the 45th International Council of Shopping Center (ICSC) Conference and Exhibition in Las Vegas, a show of about 25,000 attendees that come to deal properties and get free a latte at the Starbuck's booth.It is also where some GIS technology companies go to get their best shot at pitching their solutions to large retail customers. Solution providers such as Thompson Associates, geoVue, MPSI, Vectiv, Claritas , and Market Insight Group are representing the modeling and site selection contingent of the GIS world.ESRI's Business Information Solutions group, the recent acquistion from CACI, is also exhibiting.
But all is not well in the retail modeling world.Great stides in modeling and sales forecasting techniques have been made, and new products that make it easier to implement and visualize a retail trade area are being deployed.But as Bryan Gross of MPSI points out, "I perceive the number one pitfall in retail modeling to be the misuse and misunderstanding of methods versus reality.As an example, the religious use of market share, origin-destination models in environments that are transient and dynamic in nature may provide a sound classical approach, but lacks realism.Our area of focus these days is on the element of time."
From the owner/operator's point of view, there must be both a sound modeling approach as well as people who have years of experience in site selection.Hal Reid, a Direction's Magazine columnist and Manager of Development Systems at Allied Domecq, parent company of Baskin Robbin's, says, " The real use of a model is to keep the Site Selector honest.If the model comes back at a sales forecast that is way above or below what the site selector thinks the site will do, it makes him rethink his evaluation."
And so there is debate on using and calibrating models in addition to setting the proper expectations of the end result and how they will be applied to multi-million dollar site selection decisions.Clouding the issue is not only the choice of models but a basic understanding of the sound methodology.Gross goes on to say, "I've witnessed the complete misuse of statistical methods, GIS, Segmentation systems, Neural Nets, and other powerful techniques as well.What has tainted the field for the professionals are the many amateurs claiming levels of expertise, capabilities and accuracies that simply cannot be achieved. Miss-claims of accuracy, for example, create erroneous benchmarks against which all those interested in true modeling must henceforth compete."
At the heart of some of this controversy is the use of the "Huff Model" that was first introduced in 1964 by a young economics professor named Dr.David Huff, now teaching at the University of Texas-Austin's McCombs School of Business.At the time, Huff was trying to model probabilistic human behavior and the ability to predict the frequency by visitors to shopping areas.Huff, today, is not sure his model is being used properly given the necessary variables needed to calibrate his model.Huff would argue that in order to correctly use his model, actual sales data from the subject store as well as competing stores is necessary.Many consultants who attempt to use the Huff model today do not use these data and instead rely on patronage information captured at the point of sale from the subject store only. Tony Lea of MapInfo Canada states that, "There are huge risks associated with making up parameters for a particular retail type and market.If they (the consultants) are not calibrating it with market and context specific household retail patronage data then it is a reasonable claim that it is not really a 'Huff model'."
Reid sees it this way: "I think the
issue for most consultants is that the Huff model is too difficult to explain
and implement.So, the consultants take a more simplistic approach using
elements of the Huff model to give their work legitimacy." From the consultants
point of view, Gross concludes, "There is always a healthy tension
that exists between basic and applied methods.Basic research looks at
the world in perfect form, conceptually designing hypotheses to describe
the structure and function of that world in utopia form.Applied methods
are just that; they have to deal with the troubles and imperfections of
reality.I don't think David (Huff) is looking at it too narrowly.I believe
he is approaching the issue from the basic (theoretical) perspective.Consultants
have to make things work, even in an imperfect world.They are required
to take theory and fit it into an empirical framework.The challenge is
to know how much to sacrifice when applying theoretical methods to emprirical
problems and communicating this openly and honestly."