Power Search: If you build it, will they come?

Let’s talk about power search. Redfin currently offers over thirty ways to filter your search, but it seems that no matter how many options we provide, there will always be someone out there who wants more.

Here’s a sampling of the kinds of emails we regularly receive in our Tech Support inbox:

  • “I would like to suggest adding a filter for places that except dogs.”
  • “A check box for homes with and without a pool would make the tool that much better.”
  • “Can you please add a selection for number of Stories with a Minimum & Maximum.”
  • “Please add a Master on Main option.”
  • “It would be great if you could automatically filter based on a minimum school score.”
  • “Please add an option to search by dollars per square foot ($/sq.ft.).”
  • “It would be greatly beneficial for me to have an option to filter by accepted financing options.”
Dang. That's a lot of search options.
Dang. That’s a lot of search options.

These are all great ideas, and we would love to be able to add every one of them to our site. Of course, adding a new search feature is never as easy as simply putting a new checkbox or dropdown in the search options dialog. For every new search option we add, we need to normalize the relevant data across over two dozen MLSes, re-layout the search options dialog to fit the new option, add the relevant data to the optimized search tables, and actually code the new search option user interface.

Even so, isn’t it worthwhile to spend that time if it makes our site better and gives the users what they want? Absolutely it would be… But is “more search options” really what most users want?

While many users write in to say that they want this or that new search filter, I wanted to know just how often users are actually using the current options we provide. To that end, I asked my esteemed colleague Dane Brandon to engage his mad Hive / Hadoop skillz to pull a million rows of search query data from Apache logs of user activity on Redfin.com stored in the Cassandra file system.

After a few hours of cleaning, formatting, and pivot-tabling the resulting csv in Excel, I had my answer.

The results were clear: despite the daily requests we get for new features, the vast majority of users on our site rarely use the advanced search features that we already provide. In fact, only one search option was used in more than half of searches: maximum price at 61.2% of searches.

Here’s what the full breakdown looks like for all of the property feature filters we offer (note that our recently-released HOA dues search option is missing because I did this analysis before we added that option):

How Often are Search Filters Used on Redfin?

Notice that the only other options that appeared in more than 25% of searches were minimum beds (34.7%), and minimum price (25.1%)—all options that appear “above the fold” in the search box. Minimum baths (21.4%) and minimum square footage (15.8%) each also had decent showings, while everything else fell well below 10%, and over half of the filter options didn’t even crack five percent of searches.

Even more surprising was that just over half of user searches aren’t even limited to a specific region (e.g. city, neighborhood, zip code, or county):

How Often are Region Types Searched on Redfin?

One could conceivably argue that the low usage of our more advanced features is a UI problem, and that if the search dialog were better designed, perhaps more users would discover and use these options, but from my own personal experience and what I’ve heard from other serious home shoppers, I believe that most people just want to see all the homes for sale, and avoid using filters because they’d rather not risk missing out on a home that might be close enough to what they’re looking for.

So does this mean that we’ll be removing power search options and turning Redfin into a Fisher-Price®-inspired simplistic real estate search? Not a chance. However, hopefully it does explain why we decide to prioritize other new features with broader impact like Instant Updates, Home Price Tool, and Open Book above adding new search options.


  • Sean H

    I certainly fall into the camp that wanted to see nearly everything, in case there's something “close enough.” But we also started out looking at a huge geographic area, so it felt pretty daunting. I actually ended up downloading the data and writing a script to score houses based on our criteria, so ones that met the most of them would bubble up to the top, but ones that weren't perfect matches weren't completely eliminated either. This not only helped us sift through the existing inventory much more efficiently, but by looking at the clusters of high scores we were able to narrow down our search area a bit (though most of the narrowing happened after looking at tons of neighborhoods in person).

    It's very challenging from a UI standpoint, but I think most people approach a house search with a short list of “must haves” and a long list of “nice to haves.” And some things are tradeoffs. For example, are super low HOA dues really such a great deal if the property tax is far higher than average? Is a 7000sf lot on a corner equivalent to a non-corner 7000sf lot? If someone can come up with a search that would let people enter all their criteria, along with its importance, I think that would be pretty powerful (though I'm not willing to go out on a limb and say it would become more popular than the simple search).

    The other issue is that search criteria are only as useful as the data behind them. I don't completely trust the MLS data, or Redfin's parsing of it in some cases (not a knock against Redfin — it's just a very tricky problem). There are often errors in the data, or the prose description contradicts something in the bullet lists, etc. So I still tend to err on the side of manually looking at listings to make sure I'm not missing out on one where the agent made a mistake entering the features or something.

    Regarding the lack of region-specific searches, I found that often times the regions Redfin defines aren't really sufficient. If I want to search just a specific group of blocks, or some neighborhood Redfin doesn't know about (which seemed to be the case more often than not in my experience), it's easier to just zoom the map in to the area I want to look at and then show everything on the map. Once we were focused in on a few pockets, county, city and even zip code were far too large. In many cases (at least in the CRMLS listings), the “builder's tract name” field was the granularity we wanted, but those aren't searchable as far as I can tell (hey, another search criteria request!).

  • http://www.facebook.com/dfabulich Dan Fabulich

    I just ran a follow-up analysis, restricting my investigation to saved searches only. (Note that people with saved searches are arguably already power users.) It lines up pretty well with the numbers when we look at searches overall.

    max_price 63.0%
    region_id 59.8%
    min_beds 48.9%
    min_approx_sq_ft 46.2%
    min_price 35.6%
    min_baths 35.3%
    max_approx_sq_ft 27.7%
    sold_within_days 11.6%
    max_days_on_market 9.4%
    min_year_built 8.9%
    min_lot_sq_ft 8.3%
    max_beds 8.0%
    min_parking_spots 6.6%
    only_with_open_house 2.6%
    full_text 2.3%
    max_year_built 1.9%
    max_days_since_last_price_reduction 1.8%
    only_with_view 1.6%
    max_lot_sq_ft 1.2%
    only_new_construction 0.8%
    only_fixer 0.8%
    only_on_waterfront 0.7%
    min_days_on_market 0.7%
    min_days_since_last_price_reduction 0.4%
    max_hoa_dues 0.4%

  • Pingback: Homes Listed with a Redfin Agent are Viewed Twice as Often | Redfin Corporate Blog

  • Pingback: Seattle Bubble • What Features are Real Estate Sites Missing?

  • Emily

    I like using the free field of “remarks” myself. I usually enter “modern”. Surprised that users don’t narrow their searches more. Interesting data, thanks.

  • Chris F

    To your point about “Even more surprising was that just over half of user searches aren’t even limited to a specific region (e.g. city, neighbourhood, zip code, or county)”… the reason this has such a high percentage is that you do not allow users to search with multiple cities. Many times a family will be searching in “an area”… I may want to find a home in one of 3 adjoining cities and I want to see all results together in one search rather than having to create 3 separate searches which then don’t allow me to sort all 3 lists together, etc. “No search region” is not because people don’t care where they move to. So your users (myself included) leave the region field blank and then we need to use the map to show us everything… of course that casts a wider net than we want and that 4 adjoining city that we *do not* want in the search results is in there as well unfortunately. I completely agree that you can’t provide every possible search parameter… but the lack of “multi-region” is a pretty big disadvantage compared to other search tools I resort to. But I do love RedFin so I really hope you hear all the other feedback in other forum posts on the topic. Thanks!