Buscando para Las Casitas Escondidas y Barratas en Austin

Here is a pretty good Austin map-based real estate searcher.  It takes the Austin MLS (which is
the big real-estate database used for most conventional real-estate searches) and plots all the
houses/condos/etc. on a map that's zoomable and scrollable [via separate nav window, not quite
as nice as the google maps dragging interface].

Cool part - when you mouse over properties, it promptly shows you a photo with the price.

http://www.escapesomewhere.com/realestate_searchthemls.html

Their GUI is pretty good, but there are bugs in the way they categorize listings, so for
now you pretty much have to make sure you have nothing excluded.   For example, if you turn on
"single family" to look for houses only (not condos), it misses some houses and still displays
some condos (ewwww).   Also, as with every Austin MLS I've tried, the search by "amount of land" is
very buggy and not sufficiently granular.

But we're getting there...and these are interesting times for an investor in anything, eh?  That's
because information technology is maturing and commoditizing to the point where anyone can
do fairly sophisticated domain analysis.  The next step I see is a tantalizing one, towards truly
broad and flexible evaluation models at the disposal of anyone with the savvy to use them.  That
trend will be a truly significant one economically, as well as technologically.  Why? 

Well, investment decisions are based on estimated current and future values of tradable entities. 
Those values are based on a number of factors, which it is reasonable to view as a conditional,
weighted scoring process, based on models which take into account a number of stochastic
variables.  The output score is itself a stochastic variable which we estimate with some kind
of statistical confidence model.  For example the output of a scoring process could be -
"5 years from today, that land will be worth between $195,000 and $205,000 ; 6 times out of 10".

I've given a range of the stochastic variable, and a confidence in my estimate.  If I want more
confidence, I will have to broaden the range of my estimate:
"5 years from today, that land will be worth between $192,000 and $209,000 ; 7 times out of 10".

Now, the correctness of these estimates is based on the correctness of a large set of assumptions,
the assumptions both explicit and implicit in our estimation process.  We try to assess the quality
of our assumptions in evaluating the confidence of the estimate, but we are always subject to errors
in judgement. (Improvement here is possible given honesty, experience, and attention to measurability,
among other things).   Broadly, it should be clear that with more (warranted) confidence in tighter
estimated score ranges, we can make better investment decisions and thus have an advantage in
the marketplace. 

This quest for solid estimates is the domain of financial intelligence, which involves
information gathering, analysis, and presentation to decision makers.   Computerized tools have
important roles to play in supporting these operations, of course, but the automation aspect is
something which must be kept in perspective.   For example, sifting, categorization, and note-taking
by analysts and decision makers are crucial processes which can be supported by tools, but the
tools will not soon supplant the primary human role in the intelligence process. 

Underlying the scoring process are the models.  To estimate a value V for some asset X at time T,
you must have some model of the world in which X is valued, including some model of the currency
(e.g.dollars) or commodity (e.g. gold) in which the value is estimated.    Let's call the entire set
of models which describe the world, it's currencies, the asset X, and so on the "master model for
evaluation of X":  M(X).

M is a model which represents all of the intelligence and world-view that we want to apply to the
evaluation of X.   Now, why have I brought us all this way?  Because the model M is in fact a
stochastic ontological model to which we must apply maximum conceptual leverage in order
to have an effective valuation process for X.    And now, as an excercise, let us consider the
value of Xi = X-imperative, which is the set of imperative software artifacts  (that is, .java files and
the like) in use at a particular financial intelligence operation whose real business is to evaluate
Xf = X-financial.   Now, the fundamental question is this:  Is Xi being used to define M(Xf) directly?   
Or is Xi being used to enable a process in which M(Xf) is defined by the users of the software Xi
implements?

The answer has fundamental implications for the value of Xi  (according to my M(Xi )), and hence
should be a determining factor in the way that an organization spends money to create and maintain
Xi.

Now, eventually, the decision makers in most orgs will figure out that they need to ask this
question.  It is probably being asked at some levels in many orgs today.   But the convergence
of mainstream financial consciousness to the understanding and articulation of this question
may take quite a while.  Perhaps, in fact, it will always be a highbrow kind of question.  What do
you think?

Anyhoo, enabling the creation of model M(X) is where semantic technology comes in.  Consider the
process of searching for houses.  Basically, you are looking for a house with particular attributes,
on land with particular attributes, in areas with particular attributes.  It's all about categories and
relationships, right?   That's what ontologies are for.  Crucially, they don't have to be standardized
as long as they are open and shared, because they are fundamentally and dynamically mappable.
So as long as I know you are using Sam's super duper real-estate ontology, I can map your
information into my model, which is based on the vastly superior (in my opinion) Ann's mega-bodacious
real-estate ontology.   Now, given the difference in perspective, goals and methods of Sam and Ann
there may be some discrepancies between the models which cannot be easily overcome, and work
will indeed need to be done to "suck the good juice" out of knowledge in the Sam format.   But
that challenge derives from fundamental complexity which can't be factored out by merely
technical means, and is not something to get hung up on conceptually.

With that background, there are two further opinions I'd like to render here:

    1) The juice-sucking (or "mapping" as others call it) can be performed by an analyst, not a   
    technologist, and should be treated as an explicit, repeatable, maintainable,
    metadata-supported process.

    2) If the categorization and relationship ontologies used by different information vendors are freely
    available with exemplar datasets, then it is not terribly progress-inhibiting if real-time production
    datasets are not always free, although people should always be free to produce free datasets
    if they want to do the legwork.   

This last point takes us into the interesting question of who owns the information that "Linda's house
at XXX street is on the market".  If I can read that fact in a newspaper, or find it out by driving by, can
anyone stop me from typing her address into a database, querying public records, and assembling a
profile of information that I choose to give away for free?  It would seem that the MLS people probably
have the right to say that I can't display an MLS id on it, or display links to an MLS keyed database,
without their permission.  Is that the line?    Not that I'm really interested in building such a thing; I just
think it's interesting to consider how much "ownership" of such valuable information is possible.

I'm also trying to say that I think the openness of the eventually dominant classification and
relationship description schemes may be more important to think about now than the
freeness of the eventually dominant real-time datasets.  The latter seems a clearly market-driven
kind of question.  But the former is an issue in terms of how some people will even be able to think
about the market they are in, right?

For example, if it's in someone's interest to make us people think that "condos are like houses", then
they can try to bring about the entrenchment of hidden structures that make people think that way.
We know that these tactics can work.  Look at the way people are willing to accept the idea that
airline travel should be predominantly round-trip via hub-and-spoke routes, or that they should
check out of hotels by noon, or that they should work 40 hours a week, because these are the
messages they get from their environment.  People treat these things as non-negotiable, and our
markets are less efficient as a result.  You may think that condos/houses is such an obvious
and important distinction that it could never be papered over successfully, and you will be right
to some extent.  OK, so, what's a less clear and obvious distinction that someone might want to
paper over?  Anyone?  Mueller?

But my point is not yet clear.  How should we define and measure "openness of an ontology"?  Ah,
yes.  A good question, and one we shall return to, if the weather holds.

Comments

[Craig G]

From your explaination it would also appear that this formula would work well in valuing all the earth systems assets humanity is dependent upon for life. With peace and love from the Midwest.