Zillowing the convergence: ‘Close enough is good enough’ will eventually eat every anti-Zillow argument except the ethical complaint . . .

by Greg Swann

Suburban Phoenix Real Estate Broker

Author website

When I was young, I was convinced I was going to work in either editorial or advertising. I was a teenage photo geek, a Junior Jimmy Olson with thousands of dollars worth of professional photo gear slung over my shoulder. In college, I was publisher of the student newspaper. From the time I was very young, single digits, I was producing all sorts of printed material. And because I often didn’t have a budget, I learned how to do a lot of it by myself.

The net consequence of all this is that, by the time I had to get a real job, I knew how to write, I knew how to create images — and I knew how to do many of the back-end jobs associated with producing printed words and images. I looked at my job opportunities and saw that print production paid a helluva lot better than content creation. So I went to work in Wall Street (where the very best money was found) producing 10-Ks and Blue Sky Reports and Annual Reports. I worked a boatload of IPOs, and 102 weblog posts overnight is not a very big job compared to the 100+ hours of the revision cycle on an Initial Public Offering.

All of this was happening at an epoch we might name The Dawn of the Age of Connectivity. The law firms we worked for had high-end dedicated word processing systems, and they wanted to know why they couldn’t do everything “on disk” — in the dewy-eyed lingo of the day.

It fell to me to do this, mostly because I was interested and no one else was. The “disk” problem was a bear, and there were dozens of kludgey “solutions” to this dilemma over the years. But, understood as a telecommunications problem, mere capture of keystrokes was not that big a problem.

The big problem was expressing word-processed coding as typography — and if you are not fairly well-versed in typography, you are probably already saying, “What’s the difference?” And, indeed, the difference today is much smaller than it was when I was doing this work. Typography once was an intricate art-form, practiced by very fussy craftsmen. By now, it is mostly an afterthought, handled badly, if at all, by built-in software functions — and otherwise entirely forgotten. Gutenberg weeps, but no one has time to care.

This was an iterative problem, and I worked on it more or less continuously for over 25 years. You start with eliminating all the goofy spaces typists use to line things up, and from there you go after their habituated use of the lowercase l instead of a numeral 1, a trick they learned when typewriters didn’t have a key for the numeral 1. There are hundreds of silly things typists do that can be caught and corrected by software, and there are hundreds of other things that typographers do that typists and word processors know nothing about.

By the mid-1980s, I had C software that I was using to convert word-processor keystrokes into high-end proprietary typographic systems. When I switched all my work to QuarkXPress on the Macintosh, I renamed this software XPort, and when I actually ported it to Think C on the Mac, I renamed it XP8 — the process of ridding a soul of its sins. If you prowl the anonymous FTP servers, you can still find copies of XP8. It will run fine in the Classic environment on any Motorola PPC (not Intel) Macintosh.

Part of XP8 was a quote conversion algorithm. Here again is a problem that probably means nothing at all to you, and the only time you are likely to notice a failure in quote conversion is when it smacks you right in the face. This is so often so wrong — particularly the leading apostrophe — that I know with certainty that almost nobody cares.

I care. A lot. I wrote the best quote conversion algorithm in the history of quote conversion algorithms. I can fool it, but you never will. I can fool most quote conversion algorithms with a sentence as simple as this: Judy said, “Who wrote ‘The Spirit of ’76?'” (Did WordPress get it wrong? As much as I like WP, its quote-processing is lame.) XP8’s way of doing this is truly algorithmic, so you can do nested quotes to any level you can stand — across multiple paragraphs — and still not miss. But it also has software exception processing for possible — even if rare — problem cases, along with look-up tables for aberrant constructs that nevertheless can occur in English.

What’s the point of all this? Nobody cares. At one point, I was trading email with Tim Gill, the programming god who wrote QuarkXPress, and he had no idea why anyone would worry about quotes beyond the level of processing Microsoft Word might do.

Guess what? He was right. Quote conversion was interesting to me, because it was an inherently solvable, very difficult software problem. It was interesting — to a point — to other people passionate about typography, but they were content to correct the occasional error by hand and move on. It was not interesting at all to anyone else.

Why?

Because close enough is good enough.

Given what it is, Zillow.com cannot ever accurately evaluate a house. This requires tools, techniques and personnel Zillow.com will never have. But it is entirely possible that the correspondence between a Zillow.com Zestimate and an on-the-ground appraisal will be so high that any difference will negligible. Zillow will never be able to say for sure if the house is really there, and if someone writes a loan on a burned-down house, that lender will be a lender no more. But if you think this means that AVMs — not necessarily Zillow.com — will never replace appraisers, especially for bread and butter jobs — think again.

I would much prefer an AVM with MLS access. Tax records are polluted with dubious sales — FSBOs, wrecks, distress sales, mother-in-law deals. Represented transactions are a much flatter curve and they’re supported by much more data. But all that is just more quibbling

Close enough is good enough, and as soon as an AVM vendor can get to an algorithm that predicts the ultimate sales price to the satisfaction of all the interested parties, the need for an independent umpire goes out the window.

Think of this: Hardware is cheap. Database record space is cheap. Software is iterative — never perfect, but always getting better. The ultimate AVM would simply track every available detail about every known home.

My ethical objection to Zillow.com, in particular, remains: Portraying AVM results as being the equivalent of an appraisal is deceptive. And, while I’m interested in learning how much of the Zillow mystique is nothing more than marketing mumbo-jumbo, it remains that caviling about Zillow.com’s “accuracy” is a dead end. In the first place, it cannot ever be accurate, but in the second place, it — or another, better AVM — stands an excellent chance of achieving an uncanny correspondence to real-life market results.

Technorati Tags: disintermediation, real estate, real estate marketing