– Tour Your Next Apartment From Home

This piece is a memoir of friends trying to learn what is entrepreneurship about first-hand. A reflection of successes, mistakes, and what I’ve learned throughout the experience.

From a research lab to the real world...

After Brandon and I were done with Virtual Reality for Interior Design (VRID) research project, we wrote a paper and submitted it to an academic conference. 

Then we wait… for quite some time… and then we thought, “Man, was that it? is this research? you work on something cool, wrote a paper about it, and wait?” We were slightly disappointed because we would really like to see whether what we build is actually impactful for real people.

So we thought, “well, from what we’ve learned, can we actually make a useful product out of it?” so it began, the idea of

New idea, old problem

BrandonIsmael, and I thought about this idea of enabling 3D virtual tour for people to tour real estate from afar. This idea actually came from a problem I’ve faced before. 

Growing up, I moved from countries to countries and every time my family need to pick a new place, the options are either pick a place off a website, or actually fly to the city and explore the available apartments. The first is risky, because picking a place you’ll live for years based on web pictures doesn’t sound like a great idea. And the latter is expensive in time and money.

So we dug deep into how people browse apartments. We analyzed the players and reached out to them!


Getting to know the users

We talked to real estate firms around campus, and apartment residents, especially out-of-state students. Because we thought that out-of-state students would have faced the same problems we’ve seen. We talked to real estate firms around the campus because we thought that creating an entire apartment browsing platform and compete with Zillow and is not a wise idea. We thought we’d be a platform where property firms could create 3D visualization of their apartments.

After talking to these people we learned a couple key points.

From end-users:

  • Hard to visit apartments from out of state. ~33% of people wasn’t able to check their current dorm/apartment
  • They resort to FaceTime call. Which is cumbersome due to timezone difference and business hours restriction
  • More than 76% out-of-state students reach out to the community who already lives in that apartment before they move in

From real-estate firms perspective:

  • On average out-of-state and international students make up for more than 43% of their bottom line
  • Doing virtual tour using FaceTime is very cumbersome
Thus from this data we would try to design and build a system.

Designing the system

So since most people still browse apartments through the web, our application needs to support mobile phones and desktop. But, we should support VR headsets, because based on our research it makes people more conscientious about interior design and more likely to spend time and money. Also we were betting that VR would take off and we’d ride the trend. You see, in 2016, VR technology is super hyped.

Our designer friend, Madison, who’ve worked with us before helped us design the logo and the website. Later on, another friend of ours, Esther, picked up the work as Madison graduates.

Here’s a couple great design sketches that they came up with


Developing the First Prototype

Now that we got the design, Brandon, Ismael, and I were developing the first program. It was our first time working on Web VR, but there’s this new promising framework called A-Frame. The good thing is it works similarly like Unity. I remembered we were working on this project over the winter break. Day and night. Ismael even slept on my couch for some days. But after a couple weeks, it was done! well… sort of… it was very buggy and missing lots of parts, and definitely didn’t look like the sketched design, but it works to some extent

Champaign Housing Fair

After we built our first prototype we want to test how the real estate firms reacts to it. And we happen to learn that there’s a housing fair, which is kind of like a mini trade show for property firms to market their units to potential tenants.

So we got a table at the housing fair. I was able to negotiate a great price by highlighting our student entrepreneurship story. And I thought this event would be a great place to test 2 things:

  • Does end-users (out-of-state students) see value in our prototype?
  • Would real estate firms use our platform in their websites?

Thus, we did do an informal user tests, we had dozens of people fill out surveys, while also do sales pitches during the fair

By the end of the fair we learned a lot:

  • >90% get a sense of depth using virtision
  • No one really has a VR headset…
  • ~80% of end-users would buy furniture when viewing apartments
  • >90% of people want to visualize furniture placement before deciding an apartment
  • A lot of international students would love to use our tool since they can’t check the apartments before hand
  • End-users would not want to pay for the service
  • 5 property management firms would like to try our product, but would not willing to pay if such service exist

It’s an interesting mix of information. At the time, we were very enthusiastic about the fact that end-users really like what we’ve built. And some property management companies would like to actually try it. But looking back, this really spells doom, because our key feature was the VR tour which is accessible by enabling VR headsets, but no one really has them. Also, the people who would enjoy the benefit are end-users, but the key economic buyer in this case is actually real estate firms. Yet, real estate firms still find it hard to spend any money on it. But regardless we went forward

Cozad New Venture Competition

During that semester, we also have Cozad New Venture Competition. A competition held by U of I to guide and grow student startups. So we participated! I have an entire story about it written in this medium article

We learned a lot from Cozad NVC. I personally learned to be a better pitcher!

A lot of the stuff learned during Cozad also aligns with the things I’ve learned through reading books and sources like YCombinator Startup School. Like emphasis on understanding the problem, and users need. Using lean startup. Using Business Model Generation as a tool etc. 

I definitely recommend UIUC students to check out Cozad New Venture Competition. Even just getting to know the founders community around campus is really helpful in the long run. I met many amazing business friends during this period.

I think the Technology Entrepreneurship Center (TEC) at UIUC has done a great job providing resources to student entrepreneurs

University of Illinois Engineering Open House

Another event that happened during that semester was Engineering Open House. It’s an event hosted by Grainger College of Engineering to let students showcase their projects. We thought it would be a good time to test the general population to try using Virtision and get feedback.

A couple things we’ve learned during EOH are:

  • Anecdotally, the general public feels that 3D VR or 3D desktop is more appealing than 2D videos
  • But also, quantitatively, we noticed that people spent more time because of interactive our system is
  • But still, only a handful people has VR enabled devices

New Team, New Plan!

Our project started with 4 people from our previous VRID research team: Ismael, Brandon, Madison, and I. However, at this point, Ismael and Madison had graduated and pursued separate careers. Then we’re helped by two talented friends Esther, and Ignatius, but soon they also graduated.

At this time Brandon and I were seeking new members to join the team since we need expertise in design, sales, and architecture. Interestingly, I remembered there’s an awesome designer at company I interned before, and she was very into AR/VR design. That designer was Dawn! I reached out to her and had a chat about Virtision, and we were excited to work together. She mentioned that she worked on a VR application that showcased U of I’s newly designed Siebel Center for Design. She also mentioned that she had a team of three who would like to join our venture. Thus, Dawn, Kshithij (K), and Brandon Dang (Bdang) joined the Virtision team.

From there, I came up with a new plan that’s based on our previous findings. And here were the key points:

  • We no longer focus on VR devices as a main platform. We focus on the web and mobile experience. This is due to the lessons we learned after most people don’t have VR headsets yet.
  • Change our strategy to get real estate firms on-board by letting them try Virtision for free. And once they notice the benefit of our 3D scenes, we can upsell them. Also, we want beta testers to see the affects of Virtision on their business.
  • To get real estate managers on-board we want to make it easy for them to try our product. After talking to a lot of them, none of them have 3D models, but they do have a lot of 2D floor plan,
  • We decided maybe having a tool to convert 2D floor plan to 3D model would be a great addition.

AI-powered Automatic 2D Floor Plan to 3D Scene Algorithm

That summer, I worked on the deep learning system that could create 3D models out of 2D floor plans. I remembered working day and night at my parents’ apartment in Dubai. It was a tough engineering problem. But I was determined. The rest of the team was on campus. K was trying to get more real estate firms to try our product, Dawn was designing some of the pages, Bdang was making sure we can generate and show 3D models properly, and Brandon work on the system architecture.

As for me, I worked on Vulkan, the codename of the 2D -> 3D system. What it does is this:

  1. It takes a 2D floor plan
  2. An instance segmentation algorithm is used to recognize structural parts of the house like walls, doors, windows, etc.
  3. In the future, we would also recognize furnitures and contextual information like location of bedroom size of living room. I thought in the future, we can leverage AI to recommend users what furniture to buy. Thus could be a potential revenue stream.
  4. Then from these information, the system should create 3D .obj model by creating walls, doors, and windows.
  5. I anticipate that the model will not be 100% accurate, thus we think we should let user (real estate manager) to edit it simply by clicking and selecting.
Below is the training of the instance segmentation algorithm. I found datasets of floor plans with corresponding 3D models. And I used a Mask-RCNN model and retrained the last layer using the dataset. It actually achieved pretty good result. Below is some training screenshot that I found

Somehow I couldn’t find much images about the system. But below is some of the images that I still have. The example below is the system taking an input image of a floorplan and have wall as ground truth then the predicted wall. There’s also other models that predicts doors and windows but I couldn’t find it anymore.

Pivoting into New-building Market

Once the summer was over, Vulkan was almost done. However, I heard a very bad news from K, real estate firms just don’t really want our product. Beside the 5 clients that was trying our product, none really willing to pay for anything. They all said they could are happy with pictures of or videos of their units. This really reminds me of a lot of the advice we got from previous entrepreneurs “everyone will say your product is cool up until you ask them to pay for it”. So we definitely learned that first-hand.

However, we heard that there’s a couple new buildings coming up on campus, and I thought maybe for buildings who are already built, there isn’t much incentives to use our product, but what about new buildings? They can’t take a video of an unfinished building. Then we talked to CPM, one of the property management firm who’s having a new building and we learned interesting informations:

  1. They want to market their unit as soon as possible
  2. The architects have blueprints and CAD of their building that’s in progress
  3. They already spent thousands of dollars for architects
  4. Typically the construction takes 1-2 years
  5. They want the unit to be leased as soon as it’s done

So we thought, okay, this could work. We could position Virtision as the tool to market new buildings even before it’s done. Sadly this means that Vulkan was not even needed, all we need is to work with the architects and polish their 3D model into Web VR-ready format. RIP Vulkan…

Based on our calculations we believe that this market is fairly big as there are 350,000 new apartments built in the USA (as of 2017). Thus we pivoted our plan into this new market. And thankfully K were able to talk to CPM and they end up using our product for their new apartment building.

Learning Marketing and Advertisement

I know very little of marketing and advertisement before doing Virtision. But I learned a great deal during this period. In fact, I feel like you’d learn things faster since you really can’t waste precious time while doing a startup. I learned how to use Google Ad platform, Google Keyword Planner, Google Analytics, Facebook Ad platform, etc. 

This is actually the first time I realize how big and efficient Google and Facebook are as advertising platforms. Facebook gave us free credits, I think it was around $20, and we we tested making a simple advertisement video, and surprisingly we were able to reach a couple thousand views! It’s ridiculous. Conversion was not good though, because we just sort-of patch things together. But at least I now understand that marketing is really a numbers game. And it actually pays off to identify which channels works best to reach out to your audience. 

We learned that since the economic buyer is real estate managers, social media marketing is actually have a very low conversion. At the same time going to trade shows puts us right with our audience but it doesn’t scale really well. Looking back, this is another red flag. Typically if you’re average sales for a business is upward of $100,000 it makes sense to hire a sales team. If it’s below $100 it makes sense to do an online social media ad. But Virtision was placed around $1,500 per scene and $50/month, it’s pretty tough to target our audience and the right channels were unclear.


We were broke college kids trying spin our side project into a real business. We have time and energy, but no money. We know that, at some point, our team is going to graduate and need cash infusion to keep iterating in our business model and product. Since I don’t think we could achieve product/market fit anytime soon, we need to raise money. And boy I did apply to a lot of incubators, fellowships, and VC applications out there. I learned a lot about pitching though. I learn to pick the right metrics and tailor the content based on the audience. And potential investors, all they care about is the unit economics. What’s the market size? what’s the current growth? what’s the revenue? These are what they really care about because at the end of the day they need to make their money back. I also learned a lot of rejections. But I think at this point, I already get used to it.

Growth and Plateau

Startups is all about growth growth and growth. In our case the important metrics were:

  1. Numbers of clients (real estate firms)
  2. Number of scenes
  3. Number of visitors
  4. Time spent in our scene
  5. Conversion rate (how many people end up leasing)

These metrics matters a lot since #1 and #2 would define the revenue of the business, #3 and #4 would result how much value our system is bringing to our clients. To them Virtision is a marketing tool to increase engagement and conversion of their apartment listing.

The first couples of months were pretty good for Virtision. We got 4 clients trying our product, users and views are going up. Sessions are in the minutes.

At the peak, Virtision has over 2,000 users, 4,000 visits. And people spent about 5 and a half minute in our scenes, compared to 1 minute 20 second average of scenes with just pictures. How did we know? We did A/B testing with our clients and figure out this number. We also notice higher conversion rate, around 34% more. This is great news.

These numbers are surprisingly general for AR/VR tools, as we later went to Augmented World Expo 2019 and learned that other companies sees very high engagement metrics and conversion rate

However, many months later, growth stagnates…. It’s still very hard to sell these tools to more real estate managers. We reached out to firms in Chicago and it’s a tough sell. This is another thing I learned, B2B sales is hard if you don’t have the right connections and understand the industry channels. Just reaching out to the target audience is not an easy task. 

And when we present our benefits (higher engagement and time-spent) they don’t really care about it since most of their unit is selling fine with just an easy picture and video taking. This shows that the problem we’re trying to solve is not big enough for them to pay the cost. The target market of new-building is also doesn’t last long. Our main client who used our tool, were able to lease their units half a year before they finish the building! They were surprised with this and said that Virtision is very valuable during this stage of the development. However, they ditched our platform once the unit is done. They said, once it’s done they could just, again, take photos and videos, and they’re fine with the engagement and conversion numbers they had.


Moving on...

After about 1.5 years, a lot of us had other responsibility. Paired it up with stagnant growth and our lack of funding. I thought it’s time to move on. I’ve learned many things while doing the Virtision project, and I’m excited to use that knowledge in my next ventures. I learned that, you could have a great team and a great product, but if you’re solving a problem no one cares about, that’d make a bad business. 

That’s why I’m taking time off an try to reflect, explore the world, observe, and polish my problem-finding skill. I think I have to agree with Marc Andreesen on his point about the importance of innovating in a thriving market. As I reflect, I also realize a lot of the things I’ve done was about making cool products, but not necessarily solves what people need. Thus, In the future, I want to correct my course, and try to spend more time identifying the problems worth solving even before building the solutions…

On to the next chapter… because I believe this is just the beginning of my entrepreneurial journey 🙂

Leave a Reply

Your email address will not be published.