Memphis operator since 2003, few hundred doors, majority Section 8, own crews and own management. I'm going to argue against something I've been recommending in this forum for weeks, because somebody with forty years in another market told me I was wrong and I think he's right.

WHAT I'VE BEEN SAYING

Pull ninety days of sold comps in a ZIP, sort by price per square foot, and look at the shape. In a lot of older working-class ZIPs you don't get one curve - you get two piles with a valley between them. Unrenovated stock trading between investors and estates down low, renovated product up high, not much in the middle. One median sitting on top of two different products. I've used that to explain why indexes mislead in these neighborhoods and why a renovated house can appraise against unrenovated comps.

I still think the underlying observation is correct. The problem is the resolution I was telling people to use it at.

THE CORRECTION

Harvey Levin has been in Indianapolis forty years - property management for most of it, advisory now - and he pushed back when I asked him to confirm the pattern there. He's letting me quote him.

"Data by zip based on SF is really hard here. I have seen house spreads of 100-500k in the same block before. Most of the buyers I have consulted with who over paid bought that way or through ripoff companies."

And his example, which is the part that should stop you: east of Sherman Drive versus west, between 10th and 16th, same ZIP code, twenty to fifty thousand dollars of difference on similar houses.

Read that middle sentence again, because it's the whole post. The buyers he's consulted with who OVERPAID got there using the ZIP-level dollars-per-foot method. Not people who did no analysis. People who did this analysis. In his market that approach doesn't just fail to help - it's the mechanism of the mistake, and it's worse than no number at all because it came out of a spreadsheet and therefore feels earned.

WHY I THINK HE'S RIGHT

The two piles are real. But they're a BLOCK-level phenomenon, and ZIP data only sometimes happens to surface them. Memphis shows it at ZIP level in some places because whole pockets moved together over the last decade, so the ZIP boundary accidentally lines up with a real economic boundary. That's luck, not method. In Indianapolis the variation is finer than the container I was telling people to measure with, so the ZIP average isn't a blurry picture of the truth - it's a number that doesn't correspond to anything.

And I'd already said the same thing without noticing. A few weeks back somebody asked me about one of our widest-spread ZIPs and I told him the two products there are physically interleaved and the penalty for being wrong by two streets is the entire spread. That is Harvey's point, in my own market, in my own words - I just hadn't followed it to its conclusion, which is that if being wrong by two streets costs you the spread, then a number computed across several hundred streets cannot be the tool.

WHAT HE DOES INSTEAD

A third of a mile radius. Then adjust for location boundaries. Then look at the actual house. That's more specific than most people's comping process, which tends to stop at the radius.

So I asked him the obvious follow-up: how do you know where the boundaries are? Is there anything you can tell somebody to look for, or is the honest answer just hire a local? His reply, and this is the part I've been chewing on since: you learn them by driving the city, for years - and even those have shifted, because of where the new houses are being built.

THE UNCOMFORTABLE IMPLICATION

Local knowledge isn't a fixed asset you acquire once. It decays. Which means the risk isn't only that an out-of-state buyer doesn't know where the lines are - it's that somebody who learned a market ten years ago and stopped driving it is carrying a map that's wrong in places he can't identify. At least the out-of-state buyer knows he doesn't know. The stale local is confident. If you're relying on a partner, an agent or a manager for ground truth, the question isn't how long they've been in the market. It's whether they're still driving it.

SO WHAT IS THE ZIP NUMBER GOOD FOR

Not nothing. It's a screen, and screens are for elimination. A ZIP-level distribution can tell you that a market has two products in it and that you'd better find out which one you're buying. What it cannot do is price a house. The moment you use it to decide what something is worth, you've converted a filter into a valuation and that's where the money goes.

The honest version of the whole thing: data tells you where to send somebody local. It never tells you what something is worth.

Questions, and I'd rather have specifics than agreement:

Name a boundary in your market and what it costs to be on the wrong side. Harvey gave Sherman Drive between 10th and 16th, twenty to fifty thousand. I want a map of these from people who actually know, because that's a thing no data product will ever contain.

For the appraisers and agents: how do you handle a subject property that sits right on one of these lines? That's a real professional problem and I'd like to hear how it's actually resolved rather than how it's supposed to be.

And if you bought out of state off a ZIP-level number and it went badly, that post would be worth more than this one. Nobody wants to write it. It's the most useful thing on this forum when someone does.