{"id":5068,"date":"2017-04-05T15:53:50","date_gmt":"2017-04-05T19:53:50","guid":{"rendered":"https:\/\/digital.hbs.edu\/platform-digit\/submission\/airbnb\/"},"modified":"2017-04-07T17:04:38","modified_gmt":"2017-04-07T21:04:38","slug":"airbnb-data-informed-decision-making","status":"publish","type":"hck-submission","link":"https:\/\/d3.harvard.edu\/platform-digit\/submission\/airbnb-data-informed-decision-making\/","title":{"rendered":"Airbnb: Data Informed Decision Making"},"content":{"rendered":"

Airbnb: Data informed not data driven<\/strong><\/span><\/h2>\n

The highly successful startup believes that using data in highly complex environments is challenging. The company believes that it is common for data scientists to \u201ctoss the results of an analysis \u2018over the wall\u2019 and then move on to the next problem\u201d, and when decision makers do not understand data they may not act on it and lose out on the value of insights. Airbnb emphasizes proactive partnership between its data scientists and decision makers within the organization, and so the solution was integrating data scientists with decision makers for better decision making. Through this the company has managed to improve on its features and services in ways that created value for users while allowed Airbnb to maximize booking and capture customers.<\/p>\n

<\/p>\n

Price Tips:<\/strong><\/p>\n

Based on data from millions of stays Airbnb has developed a tool to help hosts price in order to increase the likelihood of bookings. Using algorithms Airbnb provides suggested pricing for every day of the year. Making this a user friendly feature hosts can easily look at the calendar and see dates in green indicating a high likelihood of booking and red to indicate that the host should revisit pricing. This feature both creates value for hosts and allows the company to capture value from higher booking rates.<\/p>\n

 <\/p>\n

Improving 性视界 Results:<\/strong><\/p>\n

In a continuous process to improve search results for guests, Airbnb dove into data to create solutions of this algorithmically challenging issue. The main challenge was factoring in user unique preferences alongside rankings and geographical proximity. Through a series of complex tests and experiments Airbnb final algorithm combined user behavior data using statistical models and data visualization, to produce more location relevant results for users. Leveraging community insights the search results show guests locations within a destination where they are more likely to have great experiences. This model has also made it easier to the company to apply to cities across the worth rather than map out individual cities purely based on proximity from city centers which is not always where guests would have the best travel experience.<\/p>\n

 <\/p>\n

Understating Host Acceptance Decisions:<\/strong><\/p>\n

In order to maximize the guest user experience Airbnb needed to maximize the likelihood that a stay request would be accepted by a host, therefore requiring Airbnb to create better matching algorithms. Airbnb used the results of the experiments to influence guests search results by ranking host properties with a higher likelihood of acceptance higher. To simplify this model (shown in the flowchart below) took into considered accommodation requests and the host preferences based on historic data, this produced a preference coefficient. \u201cThe weight the preference of each trip characteristic has on the acceptance decision is the coefficient that comes out of the logistic regression\u201d. This yielded a 4% improvement in acceptance creating value for both hosts and guests and capturing value for Airbnb through improved booking rates.<\/p>\n

<\/p>\n

Diversity at Airbnb:<\/strong><\/p>\n

The data science team at Airbnb found that only 10% of new hires were women, by digging deeper into the data the team realized that while the percent of female applicants was much higher than 10% the interview process seemed to be\u00a0unfair to female candidates. While the exact details of what was uncovered were not clear, the company made changes to the process starting by making the initial evaluation stage blind which yielded a higher percent of women making it to the next stage of the interview process. Through small data-informed changes, Airbnb eventually doubled the ratio of female hires at the company.<\/p>\n

 <\/p>\n

While it’s clear that Airbnb is data conscious, its approach to using data to inform decisions is key to its success. The tight integration between decision makers and data scientists has allowed the company to leverage insights to further business goals. When in many organizations valuable data insights are often lost due to managers’ inability to interpret the data or further explore it, at Airbnb\u00a0the data-informed and integrated culture allows employees to avoid these pitfalls.<\/p>\n

 <\/p>\n

How Airbnb’s Data Science Team Doubled The Ratio Of Female Employees Last Year<\/a><\/p><\/blockquote>\n