AirbnbShare is a data visualization project for insights of sharing economy. By visualizing a rich dataset from Airbnb, it allows both travelers and investors to explore the trends of Airbnb easily, and helps them to make better traveling or investment decisions.↧Click to explore the interactive data viz↧
Airbnb is best known for connecting people and culture through its sharing platform of residential-space. However, during our initial data exploration, we noticed an interesting, yet surprising fact that entire room listings dominate the market. Sharing space is not popular on Airbnb, which contradicts with our common sense. We believe peple would like to know facts like this.
"How might we help travelers and investors to make better traveling or investment decisions by revealing the facts behind data?"
We interviewed 6 participants to learn about their Airbnb experience, and what they are curious about Airbnb dataset. The interview findings lead us to 2 types of potential users: travellers who wants to live like locals, and general investors. And both types of potential users show interest in a data visualization project revealing insights of Airbnb and sharing economy.
Following the research, we began sketching and designing our first iteration following Five Design Sheets guidelines.
The initial paper prototype concept contained a dashboard of 3 different kinds of visualizations: a map of all listings by Los Angeles neighbourhoods, a circle of the distribution of different type of room by each neighbourhood, a chart of the comparison of sale price and Airbnb price by neighbourhoods.
User testing findings
After this in-class evaluation, we found that a several of these visualizations were difficult for viewers to understand and explain. In addition, too many filters actually confused the users, and prevented users from exploring these visualizations.
1. The map design is straight-forward, and visually appealing
2. The position & length attributes used in the bar chart is effective
1. Too many information & filters
2. Colors among graphs are coflicting that cause confusion
3. The data may be distorted after being normalized
*Low-fi Tableau prototype
We incorporated the map, bar chart and parallel coordinate into our low-fidelity prototype, and these three interactive visualizations provided users with more information, including the Airbnb’s price vs long-term lease.
User testing findings
We conducted in-person usability tests for our second prototype. We gave 3 participants (2 male, 1 female) a contextual scenario and asked them to think aloud while they are completing the four tasks through the low-fidelity prototype.
1. The map listing distribution is easily perceived
2. The colors used on the map are pleasant
1. Too many listings and region data that is difficult to read
2. The Brushing filter overall is confusing
3. Filters are not connected to control the whole dashboard
When user hovering over a dot on the map, the tooltip will show the name, type and price of the listing, providing richer information to the user. Meanwhile, all the listings of the same type within the same region will be highlighted, and all the corresponding data will also be highlighted in other views. For example, the pie charts will display what percentage the highlighted listings make up the total listings in Los angeles area.
When you hover over the name of the region, the bar representing that area will be highlighted. When you hover over each segment of the bar, each segment representing the type of listings in that area will also be highlighted. As stated above, all the corresponding data in the coordinated views will be highlighted at the same time, and the pie chart of Review and Listing Number Breakdown will be updated to show the data of that specific area and/or type. A tooltip will appear to provide more detailed information about the data.
We found ourselves restricted by the ability of Tableau. We gave up many ideas generated during the ideation process simply because the technical limitation. We will consider to implement our design in a more powerful and flexible platform, such as D3, which would make our final product stay as close to our original design ideas as possible.
Initially we were also looking to discover the pattern and trend of the different variables’ change over time, but fell short to find the right data source. Similarly, we couldn’t realize our idea of conducting a comparison between the crime data of each region and the Airbnb data, because only a small part of the crime data in the Los Angeles area are available, which omits the data of many regions with the most Airbnb listings in Los Angeles area. We hope to incorporate these features into our future design once the data sets are available.