以混合研究法發展出音樂串流服務使用者的人物誌 ─以Spotify為例

Abstract

音樂是我們生活中不可或缺的娛樂之一。拜科技的進步的所賜，現在聽音樂不用花較多的成本和時間就能自由的聽到想聽的音樂。 音樂串流服務的市場日益茁壯，當使用者面對越來越龐大的音樂庫，一個好的推薦系統能幫助使用者更快的找到想聽的音樂。然而，在推薦系統的演進和使用者經驗的相關研究卻仍存在著落差。目前，音樂領域的使用者研究有發現了一些現象。有學者發現了人們在選擇音樂時，傾向於選擇熟悉的而非喜好的音樂。近來，由於越來越多的軟體服務有開放API供開發者抓取資料，更多的研究者開始利用此方式擷取使用者的聽歌資料，這些資料相較於傳統的問卷或是焦點訪談更為貼近真實情況。 此研究採用了混合研究法，首先在BBS和臉書上發問卷，篩選合適的53位受訪者，利用Spotify的API搜集受訪者自2017/3/28 至 2017/4/18的聽歌資料。經過主成份分析與集群分析法後，將受測者分群，從中選出訪談者做進一步的深度訪談。最後，由量化和質化的數據歸納出最後的人物誌。

Music is an indispensable thing in our lives. With the advances in technology, we can play the music whenever we want than before.

Users of music streaming services grow larger day by day. However, a gap between the advances recommender system of music streaming services and the user experience. Recently, user research of music field discovered some phenomenon. Ward, Goodman, and Irwin proved that familiarity is an important factor when people make music choice(2014). Moreover, researchers tend to use API to collect data and explains user's behavior in the music field, temporal dynamics also proved an important factor as well.

In this study, mixed methods were used to develop personas from 53 users recruited from Facebook and BBS, all of them use Spotify to listen to music mainly. After getting authentication from all the participants, their listening data had been collected from 2017/3/28 to 2017/4/18 through Spotify Web API. Principle Component Analysis and Cluster analysis was used to separate users to the different group, then representative users were selected for in-depth interview to gain insights for each cluster.