data.bythebay.io: Ian Lin, How Data helps UX evolves at realtor.com
Recording: data.bythebay.io: Ian Lin, How Data helps UX evolves at realtor.com
so good afternoon everyone my name is Ian Lee welcome to my presentation I'm the lead they have ego agent designer at realtor.com which is the new scope company and the fastest growing online real estate service provider and by fast growing I mean we have the industry-leading growth in the number of unique visitors compared with our main competitors like zero and Trulia and we also have number one audience engagement in terms of the number of page views in the time spent on website per visit so today I'm here to share with you a use case that's case study about how we use data to identify the potential business problems in how we solve them to make our product better in terms of user experience so this is our website for you to discover your perfect home so what whatever you no matter you want to buy or rent a home you only need to type your zip code or something like the city name and then we will bring up a list of the property in that area and you can adjust your search result here and you see there's a list of the photos of the properties and if you want to learn more about opposites fecit property you just click on the photo and we will bring you to a property page for more details and at the same time you can also express your interest in this property by submitting a lead form here which is also one of our main revenue sources so so far it sounds like a traditional online shopping thing so online shopping website like Amazon just try very hard to get you on the shopping cart and expect you can check it out and realtor.com just try hard to get you to the property page and expect you can submit the lead form here but if you take a closer look you will see the difference so for online shopping you can actually get on the online shopping website from lots of different external channels and once you get on the their homepage or product page you will be expected to put everything you can you want in the shopping cart and you you may also maybe you pay four of them immediately and maybe you just regret tomorrow next week so you can also return it immediately so you can see that this is actually a very casual in light way process I would say and you can also see that lots of things in the whole process can happen in very short amount of time however the home buying process is a different story so the first and the foremost is lots of people only do it one time in their life and another thing is the whole process usually takes about six months in average or sometimes even more than that maybe more than one year and at the same time home buyers also need to educate themselves with lots of home buying information to a source for example like the real estate market train or maybe some other financial options and maybe the neighborhood information and so on so forth and finally they all they feel confident to say yes to realtors so that will be a very long process in in that way bruiser doc I'm also want to provide the lots of home buying process home buying information to help that to facilitate the that kind of self education process and at the same time we also expect it can at least contribute some Lisa mission for us so however in this diagram you can actually see that that contribution to the Lisa mentioned doesn't really happen here because you can only see a very small amount of traffic coming from the home buying information to the search page into the property page and maybe they potentially contribute a very very little bit to the Lisa mission so that's the problem we want to look into organic search just mean that you kind of you just search something by yourself you will now triggered by some other incentives yeah so in this in today's case study I will just use one of the information page we call local page which is about the neighborhood information as an example so in this page you can see lots of information about the San Francisco you can see the snapshot of the housing market and the schools and geolocation and lifestyles and if you keep scrolling down you can even see something like the demographic data and crime data and so on so forth so all this data sounds very useful for home buying right but if we look into the number of the unique visitor you can see that this local page performs a little bit lower over maybe you can actually see it's almost all-time lowest compared with other pages and if you look into the number of visits and pageviews you can also see the same thing so in some way which we will see it like the the local page has very low exposure to our audience and we wonder why it happens this evening that users don't really need this kind of information or user do need it but they don't really know what to do with it or they don't even notice the existence of our local pages so we keep looking into other data set so for bounce rate you can see that logo page actually performs not that bad compared with other pages it only has this maybe 30 to 40 bounce rate not as high as 70 or 80 so it means that as long as our users can get on the logo page they actually try a little bit and let's try a little bit to interact with the information so that is the evidence to say that okay the information is useful to some degree but how much time they really spend on this page and where do they go right after the visit on this page that's also something we want to find out so if you go to the school to city spend time on the local page you can see maybe over eighty percent of the users just give up freezing one minute and if you go to see the top 20 local page visit pass you can also see that night of them actually start from that those external channels and ends right on the local pages and only one of them is coming from property page and then still ends on the local page so the way we see is we see a huge disconnection between the logo page and other realtor.com pages in addition to the low exposure so those are two problems we want to resolve so the first thing we we do to solve those problem is we try to reorganize the content we have on the local page because we we believe that's a very useful content but but use it don't really know what to do with it so you also know that lots of the neighborhood data is actually geo location based so we think we should be able to visualize everything on the map and if you just think like that you and you call it as a visualization you will actually get something like this disaster here so you used you can see we actually get everything on this math you have the San Francisco's boundaries highlighted and you have the heat map to showcase the crime occurrence and you have the yellow circles for the schools and you have blue dot for the neighborhood amenities sometimes actually less is more but I don't mean that you have to eliminate some of the data dimension I'm telling you that maybe sometimes you you just need to decide what you want to well what do you want the users to focus on so that they are your primary dimension and you can just subdue other things to make those primary things stand out so like this example what we did is we just we just make the background map in grayscale and at the same time we created something like a spotlight effect for the boundary of the city and we also turn the crime of currents heat map into grayscale so you can see that these three dimensions can actually coexist in harmony and you don't need to you don't actually need to add more and more colors to largely clutter the map up and you can see the blue dots and the yellow circle naturally stand out and then next things we do is we also find that lots of home buyers actually want to they actually love to compare different neighborhood before they really look into a specific property so in that case we are thinking based on the neighborhood content we have we should be able to create something like a very simple index for user to to compare all the metrics about this neighborhood so for example you can see there's 84 score out of 100 for amenity and eight out of 10 for school in 99 t 7 out of 100 for crime appearance in San Francisco which means dangerous and so I I think the in this way users can easily compare by comparing those different metrics for fighting their ideal neighborhood so the next thing is we want to solve that a low exposure problem so the way we solve it is we try to incorporate what we have in the logo page directly with the property listing so when you try to search for the properties you can see that everything about the the neighborhood also shows up along with the properties on the map so it will help users very easy to find out which part of the city maybe is also affordable and and it is safe and it is in a very good school district and it is very convenient and when user find out that property they just click on the property and then all other unrelated point we'll just collapsed and you can see the point related to that property stand out and you can also see more details at the left-hand side for that particular property yeah so this is the user experience we want to provide for our home buyers so that's actually oh I have today and the thing I try to show you is a systematical approach that rhythm takes to identify the some business problem we have and how we relate it to the user experience problem in how we can solve them to provide a better user experience for our home buyers so thank you very much I hope you enjoy it and learn something today and I would love to take questions from you guys we have we actually do it before we launch it so usually okay sorry let me repeat questions how the question was how do we how do we is that the question how do we evaluate our design right so we we actually usually do it right before we launch so the way we launched it this maybe we would create we try to identify some of the target user first and we do the a/b testing and those those target user will be the only user that can use the new features first and then after that kind of evaluation we can keep tuning the design and after that we just lounge it and of course we keep looking into those kind of data to see if that is a good design and if it really brings up a great impact to our users so we can eat iterate iteratively to do the improvement like this I was thinking if you treat your user as they integrate anything bizarre instead of you make sure you do get the desired training for yourself for the curve maybe you can also get some feedback from the instrument and to validate whether those are indeed the issue by little and could sue us were and also a petition party just to give I was not thinking so many ideas and information always presented no intro begins right and so that's why I turn solid only down maybe the not so loud however it could be skilled so many dots I didn't even know what you click but there's a lot of things you cannot edit user to specify got their preference of construct maybe crime in the school choice you start to feel how salty that's really good stop it is just like a passes the best thing is that you treat the users of hard abilities are process I'm not that I didn't do focus to study the stuff to cover to make design decisions the past and I know that's the most straightforward be back and now we are online right right right that's a very good feedback for oil process yeah we actually do a little bit user testing like that but for some of the case like you mentioned maybe you say that some of we can do that user testing to figure out why users doesn't really click on something but sometimes it's a challenge also for us to find out who are really those users don't really click through those and the also another challenge is how we contact them and how we persuade them so usually the user testing you just recruit people kind of randomly or you you you you try to focus on some of the specific group of people yeah experience with that you're looking at sort of more quantitative these measures like this it can point you in a direction to sort of restating what we're saying but you have the quality of explanation for why that is and sometimes you can you can recruit people who are not literally do I think we were representative or close to representative and then if you have a look at this screen and you do a basically usability evaluation and when I found in some cases for example like site closed and all such is that people would not click through because of something very silly like silly like they couldn't see the button they didn't and the label on the back right and yeah you only see that as the only if you don't know that maybe the quantitative that you get this recursion like you get rid of the button cuz the only it's important but it's not because of the reasons you think it's because they and see it so if it's sort of like qualitative thing that's been wonderful right so i can share solid since we did not see we'll try to get as much as possible you should be back up front so even picture myself in your business this is a we actually even just sketch that you say interface and then we actually prepare structure questions and certain people for you you are working with the size of a common recruits even just for your inquiry and then ask them to pretend you are doing this business or this house meaning of the user in interface correctly your best diamonds using the basic instead and then once we actually having a kind of workable functional prototype pre-alpha version I even asked our company's colleagues and ksenia miss Carol II guess you won't get in reading a lot of skill gears up and a lot of things they may come from this random sample I could give you more peep in sector 5 musics we can put in hardware that's my point space and knowing yeah we actually also yeah yeah thank you very much all right all right so I guess we the time is up so thank you very much and have a great weekend