Over the past few days, Redfin got into it with a bunch of other real estate websites. What else is new?
In an argument about who has the most homes for sale, which began on TechCrunch and continued on Redfin’s blog, one participant argued that what consumers really care about is advanced filtering options, not inventory.
Which got us thinking. We spend a fair bit of time on advanced filtering options. And we’ve always thought we need to spend more: every week, we get requests for filters on parking, townhouses, waterfront location (Seattle), historic designation (DC), pool (LA).
So Redfin’s Jim Lamb just analyzed 70,000 Redfin searches from Thursday, August 21 to find out which of Redfin.com’s search filters people really use. It’s an analysis we’ve done before, to figure out whether a listing gets seen more if it’s priced to be included in web searches, like at $449,500 rather than $450,100.
What we learned last night was a little demoralizing. People filter on price, beds, baths, sometimes square feet, and new (or very old) listings, but not much else:
- Price: Min 24.8%; Max 53.9%
- Beds: Min 32.8%
- Baths: Min 21.4%
- Square Feet: Min 15.1%; Max 2.4%
- Days on Redfin: 12.7% (this would include requests for new listings, listing on Redfin more than 45 days, or filters on on a specific number of days on Redfin; I suspect that almost all the volume comes from request for new listings)
On looking at this, Matt Goyer said, “Who doesn’t filter on square footage?” I could only sadly shake my head. Consumers completely skip the fancy stuff:
- Lot Size: Min 5.5%; Max 0.55%
- Year Built: Min 5.4%; Max 1.1%
- Has view: 1.1%
- New construction: 0.24%
- Fixer-uppers: 0.36%
- Open houses: 0.7%
So even as we argued that filtering options aren’t as important as inventory, we didn’t really believe it: our engineers have been hard at work on… you guessed it, more filtering options. Just now, it’s parking & townhouse filters. (Every week, I get a crazy screed from a consumer about how much people hate townhouses… which I read… from my townhouse.)
What do you think? Are we wasting our time? Confusing our consumers? As it is, we worry that the search options — pictured above — are over the top. If you leave a comment, I promise that a dozen insecure Redfin people will read it and argue over it.
PS: Greg, sorry it took me so long to post on Bloodhound. I kept waiting for something as good as the rest of this blog, and finally just decided to go with whatever was nearest at hand.
PPS: has anyone checked out what you can do with Animoto for a listing? Animoto let’s you turn your photos into a music video. It’s kind of cheesy, but easy and fun.
Eric Bramlett says:
Hey Glenn –
I think the options are nice to have, as an expandable feature from the basic options. Your stats have confirmed a gut feeling that I’ve had for a while – KISS. Most people want to look at a all the options in a certain price point, in a specific area, with at least a certain # of bedrooms. Many people just want to look at certain price points in certain areas, regardless of size.
Thanks for sharing your numbers.
August 26, 2008 — 8:04 am
Glenn Kelman says:
Hey Eric, thanks for the comment! What does KISS stand for?
August 26, 2008 — 8:16 am
Eric Bramlett says:
keep it simple stupid 🙂
August 26, 2008 — 8:17 am
Glenn Kelman says:
That’s my exact problem! I make things too complicated…
August 26, 2008 — 8:22 am
Greg Swann says:
My vote is for the most comprehensive search possible, with the arcane details perhaps hidden behind expanding category headers. People are just learning these tools, so they’re taking point-and-shoot pictures, as it were, with your high-end digital SLR. Give ’em time, they’ll catch up, especially the folks doing saved and emailed searches — IOW, the dedicated shoppers. A very robust IDX/VOW system can produce as good a short-list as a Realtor if the end-user has had time to make the software sing.
August 26, 2008 — 8:22 am
Morgan Carey says:
Honestly I think the numbers are not as bleak as they first appear. Just like in web search users will tend to try to be lazy filling in the very least first (almost every time) in order to find results.
They will do this a few times often before they realize – hey wait a second, I am having to scroll through WAY to many results – maybe I should actually “use” these little features here – So what’s frustrating me? The likely start with area drill down (I don’t want to live here, or here or here) maybe I should just click the little box that selects where I “do” live.
What else don’t I want to see – obviously now that I have seen them, I don’t want any of these cheap 100k properties, let’s set it to 200k and be realistic, we are never going to be able to afford more than 450k.
I don’t want all these condos (I want a yard) etc
So as they go through the process, their searches get more and more refined – of course because inventory numbers start dropping drastically after you actually start refining – they will often times come across a property (Out of 5-10 left) without having to get super advanced. So the “need” most of the time for uber advanced filtering is likely not always there – but that does not mean that they are not appreciated for those who are succinct about their home choice or that in implementing them, even those that don’t use them aren’t subconsciously effected by the “quality” of the search tools available.
But I digress – my whole point was that because of this “generic to specific” behavior patter (Just like in web search) the % numbers are going to be skewed based on this apparent human nature issue.
August 26, 2008 — 8:28 am
Morgan Carey says:
I guess in a nutshell what I was trying to draw a parallel between was “long tail” in search engine marketing speak, and “refined search” in inventory searching. Long tail has way less available traffic that generic terms, but it is way more targeted and desirable (From a per visitor perspective) than generic search terms. You want both 🙂
August 26, 2008 — 8:36 am
Brian Brady says:
I liked the RedTV music video (consider the source before you convert all your listings to that format)
Welcome, Glenn.
August 26, 2008 — 8:40 am
Eric Bramlett says:
Good points, Morgan. I didn’t consider the “drill down” as you’ve described it, but it definitely shows that the detailed search criteria are important.
August 26, 2008 — 8:49 am
Glenn Kelman says:
Greg, Morgan, I like the idea of thinking about search as a progression rather than a stateless series of queries. Very insightful…
August 26, 2008 — 8:49 am
Bob in San Diego says:
Like Glenn, I have a boatload of search data at my disposal. Not as much as Redfin, but more than most who are not 3rd party bots.
Going back to 2002 with over 100,000 registered users, my findings were almost the same as Glenn’s:
IMO more options confuse the consumer. We also found that those who were most likely to get frustrated or mad or scream that my search sucked because they didnt get the new listing down the street had saved to many differing advanced searches. Originally they like the advanced filters, but later forgot how they were using them.
As an agent, I also know that most buyers (and I mean the ones who actually buy) seldom ended up with what they searched for when they narrowed the options. We also know that real estate is all about trade-offs. The buyer eventually learns that, but it isnt what they want to hear in the beginning. How many times has an agent heard, “If I’m spending $X, I should be able to get what I want”?
My most effective advanced search tool was pre-VOW/IDX. When ever someone filled out the “I’m looking for a house form, they got back an email asking them the following:
Acceptable areas
Min/max price
min square footage
min bedrooms
min bathrooms
are schools important
is commute time an issue
what things are “must haves”
what are “great if I can get them”
what things are deal killers
When those questioned are answered, it allows me as the agent to step in and fill the gaps that online search can’t.
The biggest mistakes I have seen were made by those who argued that they were right because they think like a consumer.
That argument dies because people shop differently (see “Defending the Caveman”). I’m the hunter type. I ID the target, track it down, kill it, and drag it back to the cave. Others are the gatherer type, going from bush to bush looking for the ripest or sweetest berries. It takes me 10 minutes to buy a new shirt. My wife and daughter went shopping the other day and… well, better not for me to go there.
You can also glean a lot from the search experts at Google because they are experts at HOW people search.
Here is something you may find interesting that Matt Cutts wrote about Ben Gomes and Google’s search interface:
Great discussion and insight Glenn. I would love to read more about what you guys deal with.
August 26, 2008 — 8:51 am
Lou Lynch says:
I agree with Morgan and actually see that type of behavior on my site as I document every search a user does.
Recently, I did a poll of all of my users asking what I could do to improve my site. The majority of those who responded asked for enhanced search features i.e. map, water features, school district, etc.
Simple is good, but eventually they will want all the bells and whistles.
August 26, 2008 — 8:52 am
Mike Pannell says:
I get a lot of my business from long tails. Also i like the video, We have an agent in my office that uses video like that. They all rank real high on google
August 26, 2008 — 9:01 am
Ken Smith, Arlington Heights IL says:
Think about the first time someone looks at your search tool. If there are 20 options it can be very overwhelming to the average person. Plus you narrow a search down to much and you miss out on some great properties.
IMO the worst part of using to many fields to narrow the search is that most agents don’t input the data correctly so you are missing out on properties that really do meet your needs.
Just and FYI, for your pricing example “$449,500 rather than $450,100” the one that gets the most traffic is $450,000. It is a top and bottom price range so it shows up in both searches. If 24% of people use a bottom price then $449,500 has a chance of being eliminated by those searching from $450k +, if 53.9% use a max price then (based on your search tool) $450,100 gets dropped from a lot of searches that cap out at $450k.
August 26, 2008 — 9:01 am
David G from Zillow.com says:
Glen –
Great post. Two thoughts;
Your results illustrate the difference between preferences and requirements. Buying a home is often a trade-off between what we want and what we can live with (or without.) An algorithm can’t make that trade-off for us and so we have to manually weed out the options.
The most important filter remains geography and the geography where the buyer is searching often makes the other filters implicit / unnecessary.
August 26, 2008 — 9:14 am
Glenn Kelman says:
Good point Ken about $450K. Bob, LOVE the question about deal killers…
August 26, 2008 — 9:51 am
Don Reedy says:
Bob has it right; and Bob has it wrong; and the answer, dear Brutus, lies in that analysis.
WHAT questions to ask, including “deal killers”, are most important. If you ask too many questions, they won’t get answered (I think that’s where Bob is right). If you ask too few questions, you won’t uncover the cues and reasons that push or pull any particular shopper over to either your site or an individual property (Greg is right here).
Where I think Bob goes wrong is in his analogy about hunters. While it’s true that killing, dragging and devouring are intrinsic, it is likewise true that pawing, circling and sniffing (a al Bloodhounds) are methods we now embrace for deciding what game to bring home to the family. Thus, the gambit in play is to allow for the hunter, but provide clues, scents and paths for the pack.
Glenn, very interesting post. Don’t let the numbers dictate until you are sure what those numbers actually tell you.
August 26, 2008 — 10:29 am
Bob says:
Morgan hit on the problem with how most property search is done in relationship to how searchers think.
What if when you started your search, you were given a choice of possible parameters to search, and when chosen, you were given the exact number of matches for those choices?
For instance, given a broad geo match option to start with, I tick on the San Diego County box and immediately am given the number of properties that are available. A side panel now gives me all the possible fields to search on that are NOT empty. As I click on my choices, the options narrow based on the matches.
So I check San Diego and see 30k properties and a ton of possible search criteria. I want residential and only single family. If I bedrooms, I’m given the number of matches possible for each number of bedrooms. If I click min of 4 bedrooms, I’m given all matches. Each click also gives me a revised set of options that will return at least 1 property as the sidebar returns the possible matches for each option before I hit the search button.
The ability to sort on the fly like that gives the searcher the most comprehensive search experience possible with the aggravation. The first bot to implement this with the most inventory wins.
The cool thing is that you could do this with the entire tax roll, with “For Sale” being just one of the options, not the starting point.
This search tech exists, but is done primarily on the government level. It’s how some police forces track down hit and run drivers where they need to find that late model 4 door red sedan with an “M” in the license plate.
August 26, 2008 — 10:43 am
Bob says:
@Don, I wasn’t implying that there are only two types, but that there is more than one. They are all accommodated, including pawing and sniffing, with the above search model.
August 26, 2008 — 10:46 am
Bob says:
Glenn, the next time you are visiting your troops in San Diego, let me know and I’ll set up a demo using data from L.A. County.
Dave from Zillow, same for you.
August 26, 2008 — 10:51 am
John Rowles says:
Thanks for sharing those metrics.
It’s fascinating to me when users take something we’ve built and then use it differently than we thought they would. Here’s an example from our App:
We use the Google Search Appliance to make IDX feeds “googleable”. I think the most interesting thing about our approach is the ability to easily combine objective and subjective criteria to find exactly what you want quickly (ie. the 5 condos that meet the criteria “Seattle condo with space needle view and a hot tub”).
As it turns out, few people use our Search Engine that way.
Instead, we see a lot of freetext searches for addresses and street names. So why is that?
We know most moves are local, right? Something like 80%+ if I recall the latest NAR numbers. But most RE sites are primarily set up for the 20% of people who are moving into an area they don’t know, instead of the 80% who are moving locally and know the area.
I’m not sure if this is intentional or just a legacy of using basic database field searches that can only return ranges — sort of like how the modern rail gauge can be traced to the width of Roman horse butts — but the effect is that consumers are forced to sort through a list when they are, in fact, looking for one specific property.
Our users have apparently discovered that when a yard sign catches their eye, and they know what town they are in and the name of the street they are on, they can go to our App and find just the house they want in 3 seconds, w/out having to guess at the price or how many bedrooms the place might have in order to narrow down a list they didn’t want in the first place.
Its not as interesting as searching for a condo with a hot tub and a Space Needle view, but it is the task our users have decided our tool is for, and that’s fine by me.
August 26, 2008 — 11:18 am
Marlow says:
It’s a mistake to use too many filters as you may miss out on the perfect house. You can change almost anything about a home, it’s size and square footage, the finishes, the number of baths, that kind of thing. But you can’t change the location. So the most prudent thing to do if location matters to you (and it should), is to look at everything in that particular location in your price range. Don’t rule anything out, because it may be able to be altered, remodeled or remuddled to fit your needs and lifestyle. This is especially important when filtering for square footage because you don’t know if the agent put in the basement sqft or didn’t add in the finished attic space or is using the County Records before a remodel. It’s rarely accurate, at least in our state, and agents are often directed by their brokerage to just leave it out if they’re unsure.
August 26, 2008 — 11:27 am
Jim Reppond says:
Glenn,
I’m just curious – Why do you call it “Days on Redfin” instead of “Days on Market”? Is the data not that accurate? Or is it more of a MLS regulation thing? I know here in NWMLS you need to show both CDOM and DOM if you want to publish either. And I’m sure that varies from MLS to MLS.
August 26, 2008 — 11:43 am
Glenn Kelman says:
We measure how long the listing has been on our site, not how long it has been on the MLS…
August 26, 2008 — 11:49 am
Eric Bramlett says:
@Glenn –
They don’t allow us to display DOM in Austin. Is it the same in your market(s)? Have you guys caught any heat over the “days on redfin?”
August 26, 2008 — 11:51 am
Jim Reppond says:
RE: We measure how long the listing has been on our site, not how long it has been on the MLS…
Got it. My question is why?
Don’t people really want to know CDOM and DOM?
August 26, 2008 — 11:53 am
Marlow says:
It’s against NWMLS rules to show “Days on Market”.
But Redfin’s playing a little game and showing “Days on Redfin” to circumvent this rule.
August 26, 2008 — 12:21 pm
Frank LLosa- Broker FranklyRealty.com says:
Yep, I think you are wasting your time
I actually take the opposite approach and offer NO options! Just keyword searching and Max price. Not even a Minimum price.
So the goal with my http://FranklyMLS.com is SPEED not GeeWhiz, since
a) few people use it and
b) listing data is oftentimes wrong and
c) all that really matters is price, and I show them what they can afford (actually I show them 3% higher, so as to get those $504k listings).
And for all those people that say they want a 2 bedroom condo, 62% of them (a number I made up) would consider a 1 bdr and Den, so if you refine your search too much, you miss out on great stuff.
And another reason super filtering doesn’t work is the data integrity. If the listing agent puts in the parking information incorrectly, it won’t show up. And poorly marketed homes are the best “deals.”
So Max price and lightening fast searches.
Frank
Broker http://FranklyRealty.com
Owner http://FranklyMLS.com
August 26, 2008 — 12:27 pm
Glenn Kelman says:
Actually, what we show and how we describe it are the result of agreements between Redfin and several MLSs. We regularly talk to many MLSs about the features we have planned; this is cumbersome but it’s also the best way to build a website that draws on other brokers’ data.
It is undeniable that how we’ve handled this particular feature is a compromise, but consumers want the information, we want to display the information, MLSs are comfortable with this compromise, and we are working to make the feature better, too.
August 26, 2008 — 12:37 pm
Tammy at Myrtle Beach Real Estate says:
Thanks for the info.. Being here in Myrtle Beach, I was unfamiliar with Redfin.
August 26, 2008 — 12:48 pm
Brian Miller says:
Glenn,
I really appreciate your willingness to talk openly about your numbers and some struggles with search filters. Personally I like the google-based searches as John R. mentioned above. My bet is that there will be a lot more of this in the future, but for now we are programmed to search via the “standard” search tools. Yet, I wonder, how effective will the most sophisticated search tool will be if agents don’t use “space needle” view as part of the description in their MLS posting…
August 26, 2008 — 12:58 pm
Jim Reppond says:
RE: It’s against NWMLS rules to show “Days on Market”.
Marlow,
Actually the rules have changed as of July 3, 2008. NWMLS allows “Days on Market” to be published publically as long as CDOM (Cumulative Days on Market) is also displayed. (Rule 27 a.)
I’ve been advocating that our company publish this because it’s such cricial information for a Buyer in this market.
August 26, 2008 — 1:26 pm
Thomas Johnson says:
The most important filter remains geography and the geography where the buyer is searching often makes the other filters implicit / unnecessary.
Thanks, David. I guess this means that the most important thing in real estate is location, location and location? How Bloodhound is that? We are back to zestifarming nee peeing on the trees.
August 26, 2008 — 1:44 pm
Matt McGee says:
Fascinating data, Glenn — thanks for sharing that. It’s also important to know WHO is searching in this way. Can I assume your audience skews pretty young? What can you tell us about your users to help us understand this data better? Thanks in advance.
August 26, 2008 — 3:02 pm
Mike Farmer says:
I think having two different levels is effective. One for the looky-loos who have a fetish with diverse forms of layouts, granite counter tops,molding and fashionable colors, regardless of where the house is.
Advertising a serious area, where someone is more practical and just wants to filter down to the ones that meet criteria, seems the best route.
Label them “House Freak Search Form” and “A Real Damn Buyer Form”.
August 26, 2008 — 3:05 pm
Smithers says:
Glenn,
I use your search engine for the SF Bay and San Diego areas (the later sometimes gets frustrating because of the Redfin service boundaries). Redfin is one of the best of the various search engines because of the local comp and past sales data you (sometimes) supply. SDLookUp is also excellent for San Diego. I must be one of the few that uses the “max square feet” filter (are we the only ones looking for a smaller house out there?). Just going off max $ gives me too many McMansions to wade through priced based on size instead of quality/location.
I look forward to more posts by you. An EXCELLENT topic, if you are willing to take it on, is your business model. You rely on the traditional broker co-op 3% in order to have money to kick back to buyers; but the implicit reason buyers will do this is because the co-op fee does not bring them (at least the ones prefering the kick back) any value to begin with. So, if the entire market place viewed the co-op fee as a waste of the sellers/buyers money on RE agents (a premise Redfin is selling), Redfin would immediately be out of business!
There’s at least a smidgen of irony in that. A very cynical business plan, IMO.
(p.s., keep up the good search engine).
August 26, 2008 — 6:30 pm
Glenn Kelman says:
@Smithers; thanks for the kind words. I think Greg and I agree that we’d prefer that all brokers, including Redfin, simply charge a fee for their services. We started with that plan in mind, but consumers everywhere kept telling us that a buyer’s broker was free, and so then we explained the money that buyers would save as a refund.
As it is, Redfin has an enormous challenge trying to educate consumers about how our service is different than a traditional broker’s; we could also price the service in a way that upset consumer’s expectations but it just seems we should take one big change at a time…
August 26, 2008 — 6:53 pm
Smithers says:
“but consumers everywhere kept telling us that a buyer’s broker was free, and so then we explained the money that buyers would save as a refund”
We consumers are a sometimes clever, but far more often stupid breed.
August 26, 2008 — 7:50 pm
Drew Izzo from Roost.com says:
Interesting conversation. I’m a bit late to the party! I agree that filters can add complexity to the presentation of what consumers came to see – the content.
Three main points from me:
1. We also found that beds, baths and price are the most used filters so we made them immediately visible on the search results page. All other filters are cleverly hidden behind the “more filters” button.
2. Although most consumers use a few select filters, I wonder if consumers that dig deep end up registering – or finding their home at a increased rate?
3. Filters on top of inaccurate or incomplete data are wasting consumer time. Roost has comprehensive MLS content in every market we launch in.
August 26, 2008 — 8:51 pm
louis cammarosano says:
Glenn
Perhaps you need to add another level of analysis
What is the behaviour of the searchers who
Contact you and what is the behaviour of
Those that work and buy with your agents
It may be the casual searchers don’t filter too
Much an the serious homebuyers do more filtering
Or maybe not
Rather than speculate it would be interesting
To know then you might get an answer to your
Question
August 26, 2008 — 9:42 pm
Mike in Eau Claire says:
Thanks for sharing your data Glenn. I’m glad I haven’t shelled out thousands for a custom IDX searching capability yet. No disrespect intended to Morgan of course…
@Frank Llosa: very interesting, how long have you been using this stripped-down style search?
August 27, 2008 — 8:07 pm
Carolyn Gjerde-Tu says:
So it sounds like to optimize consumers response, there should be a very basic search first with an option of a more detailed/specific search later. Wondering also how map searches fit into this as it seems like most IDX map based searches are relatively simple searches.
August 27, 2008 — 9:59 pm
Malok says:
Interesting numbers – but as others have indicated, its not necessarily all about the numbers. What would be more interesting is instead of an aggregate of all searches – is if the information was produced for what the specific user ended up using as their LAST search, after refining, before leaving the site. That being said however, I think what can be taken away is that some of the most used values need to be the first things listed in your search (so that the user can get those high demand categories filled out).
August 28, 2008 — 4:40 am
Sam Dodd says:
I perfer having both a simple search (beds, bath, max/min price, area) and an advanced search (beds, baths, max/min price, area, amenities, pools, time on market, lot size, ect ect ect)
This gives the consumers the simple easy to use search they want as well as an advanced search that allows them to really vet the results.
Morgan is right in my experance. Buyers use the simple search, see the 500 results then go back and preform a detailed search. I feel it is better to give them both.
August 28, 2008 — 6:03 am
Jim Gatos says:
KISS or you’ll most likely get “0” homes found in a lot of searches, which can be just as equally frustrating as “567 Homes Found”..
I think Sam Dodd is right, let them be able to do both..
August 30, 2008 — 1:08 pm
Paul Francis, CRS agent in Las Vegas says:
Hmmm… so what are the numbers for people who actually closed a deal?
What did they do? What did they search for and how much did they refine their search? Those are the numbers that interest me because the rest of it, is all trivial information when you consider the ratio percentages of website visitors that actually close a deal.
September 2, 2008 — 12:46 am
Sam Chapman says:
I look at searches people save on my website and am amazed how many only select price as criteria. A basic $200,000 to $250,000 search returns hundreds of listings and it makes me wonder what people are doing. The only people who seem to get really specific are the very serious buyers who like to do a lot of research. My guess about the people selecting the very basics are just hopping from site to site until they find one they really like.
September 2, 2008 — 12:49 pm
Benjamin Dona says:
While the numbers are interesting, the comments from some of the savvy folks in the real estate space are invaluable reading.
I’ve been dissecting our numbers for quite awhile now and the one thing that sticks out understands your market and your clientele. For instance, in our area KISS is imperative because most of our clients are older and not all that computer literate. The vast majority of those I speak with about their user experience say the number one thing that made them choose our site over others was “ease of use.”
Another thing that we have found works well is to remember that sometimes “less is more.” Give them too much info and they have no reason to ever contact you. Those client initiated emails and phone calls looking for more information are a lot easier to convert into clients then those you simply are cold calling from a search registration.
September 2, 2008 — 2:19 pm