I’m proud and honored to type alongside you.

PS to the readers: Glenn’s too modest to tell you how he did at the conference. He outlined the Redfin business plan…to competitors (no less) and was swamped after his presentation with adulating praise and good questions.

I spend a LOT of time searching for real estate, both on Redfin and other sites (mostly for markets you guys don’t cover). It’s kind of a weird hobby of mine. Don’t laugh. Anyway, even though I am relatively specific in what I’m looking for, I don’t tend to use the advanced filters much. Most often, I search by bedrooms, bathrooms and price (both max and min). Those aren’t the only factors that matter to me, but they’re the only ones that I can easily capture in a search.

For example, I don’t want a house that was built in the 1950s *unless* it’s been nicely updated. A year search eliminates a bunch of houses I don’t want to see, but it also eliminates some I’d like. There’s no “any year as long as it’s nicely updated” checkbox (and if there were, I wouldn’t trust it because there’s no way the data could be truly reliable).

I love views, but “has view” is horribly subjective, and there are other things that could make up for the lack of a view. So I don’t want to eliminate houses that aren’t marked as having a view.

Square footage and lot size are important to me, and I do use those filters from time to time, but I don’t trust the data very much. Square footage is notoriously inaccurate in listings, and a huge lot that’s on a very steep grade is less desirable than a smaller flat lot.

Also, criteria may differ based on location, even within a single search area. For example, I may know that the ca. 1960s houses in a certain subdivision are horrible and I don’t want to look at them, but the ca. 1960s houses four blocks away have potential.

The bottom line is that it’s all about tradeoffs, and those are extremely hard to represent in search filters. So I end up putting in my absolute minimum criteria (I need X bedrooms no matter what, and there’s no way I can pay more than Y), and I manually evaluate the rest. Adding filters reduces the amount of junk I have to look at, but it also presents a significant risk of eliminating something I’d love. False positives are better than false negatives.

All that said, frankly, I think that you should offer filters for every field you have in your database. So what if only .5% of your users take advantage of it? For those users, it could be a killer feature. Just keep all the complexity hidden until people specifically go looking for it.

> like drawing your own neighborhood outlines on our map

We have this in our new MLS system — and in FBS Systems’ IDX system. Multiple, non-contiguous irregular polygons, so you can do something like within-walking-distance-of-Spellman-HS OR within-walking-distance-of-bus-routes-leading-to-Spellman-HS. Totally rocks. It’s by far the best way of isolating homes by location, since you don’t have to trust anyone’s typing (too much).

I’ll give you a feature: Intersection of multiple searches, or perhaps weighting by intersection. So: I do a series of saved searches, each one isolating for certain characteristics I want. At the end, I select the intersection of the specified searches. The resulting short list is either a list of the homes that scored on all searches, or perhaps a weighted short list with those that scored on the most searches showing up first. I kindasorta like the weighted search on Realtor.com, except that you can end up looking at homes that are missing a must-have feature. Using multiple either/or searches then weighting them post hoc might be more satisfying. And think how easy it will be to document! ;)