Amit Dixit brings with him over 21 years of experience of product and software development, large-scale platform management, technology roadmap development, Cloud migration and Cloud strategy across business domains such as banking, retail, ERP and e-commerce. At Zolo, Dixit leads the teams in design, product development, technology strategy, and programme management.
In shared accommodation, finding the right roommate can make someone’s life very easy. For co-living companies like ZoloStays, the perfect roommate helps in increasing the length of stay, superior customer experience and promoting a shared economy of living. Roommate personalization or matching is more important for student housing as multiple sharing is a very common phenomenon during college days. There are many portals for finding the right roommate in many countries.
We believe that technology can play a very important role to solve this problem because only technology can help us uncover hidden trends and make cohorts or groups based on data which is not easily available and visible to us. We at ZoloStays have a special focus on finding a perfect roommate or finding a roommate of your choice. Our solution not only offers a perfect roommate but also offers a roommate of your choice. In order to personalize, preferences are collected at the time of booking or at point in their journey with ZoloStays. Preferences data is one of the many data points which are used in matching algorithms.
Resident’s cohorts are created from the data by using various techniques such as clustering and classification by using K-means and KNN (k-nearest neighbors). Both Supervised and unsupervised clustering techniques are used to make the resident cohorts. These cohorts are created on similarities or dissimilarities. Clustering is a way of finding groups of homogeneous residents. On a very high level, grouping is mainly based on three factors: interpersonal skills, ability to interact and situational awareness specific to profession or locality. Machine learning helps us identify the weightages for every feature. Weightages helps us identify the importance of specific features. In few cases, communication is considered as the most important criteria where people come from different linguistic backgrounds while in other cases where there is a common means of communication, other factors take more importance. It is very important to provide the suggestions in near real time and which is ensured by our technology solution by using offline and online computation. There are REST APIs available to provide personalized roommate suggestions in real time.
As customers provide their preferences, roommate personalization algorithms predict the possible matches from current residents in real time at the time of pre booking. Customers can choose their preferences such as mother tongue and meal preferences. ZoloStays room allotment engine takes care of their preferences and allots sharing rooms according to matches. There are feedback loops to test and further refine the accuracy of our model. It is relatively simpler to provide roommate personalization in large properties with a large number of beds as the number of options are also large. It is hard to provide roommate personalization in smaller properties with less inventory.
We are witnessing huge demand for such a solution in Student housing and our algorithms provide more accurate results in such a large sample size. Every hostel is looking for ways to allot rooms to students in the beginning of the session. Currently, only 5-10% allotment is driven by personal preferences of students and their parents. Roommate personalization engine provides data backed suggestions and students can make informed decisions. This also helps in optimizing the room/bed allocation. Currently, options are given to request for a change in case of negative feedback, but we are hopeful that we will be able to achieve 90% accuracy in providing best roommates to the students and residents in properties managed by ZoloStays.
There are many other use cases to increase engagement based on these cohorts. In colleges, It helps in overall personality development of the student when we help people with common interests come together.