In Search Of: Information Retrieval and Exploration in Spotify's Music Discovery Features

Research question: Does Spotify's Browse feature offer any benefits over its Search feature?

I designed a user study to evaluate the effectiveness of two music-searching features offered by Spotify, the Browse and Browse features. Results of the study showed that the Browse feature is not as effective for music-searching compared to Search, and that Browse may be better utilized as an adjacent feature.

Duration
Team
Method
Tools
Apparatus
- Experimenter's laptop

My role included designing the experiment, creating the Qualtrics survey, and carrying out the and performing the analysis in R. See details

Examples of an experiment instruction screen in the prototype and task.

The Problem

Spotify provides two features which overlap in functionality, begging the question as to why both need to exist. It's Browse and Search features allow users to browse through the platform's music library, but primarily through two distinct mediums. Browse offers a imagery-first context in which users navigate through Spotify's library in the form of higher-level folders. In contrast, Search allows users to navigate the platform's library primarily through text and is user-driven.

The overlapping functionality means that one must be redundant. It is very likely that the Spotify's Browse feature is underused by users, provides no real benefits to users, and its purpose could be redesigned to better utilize the imagery-first design. Thus, I investigated whether there was any benefit to Spotify's Browse feature over its Search feature for navigating Spotify's library.

Method

Two features for searching through Spotify's catalogue were evaluated.

Browse feature. — Browse
Spotify's library represented through higher level folders.

Search feature, — Search
Textual search using user-entered keywords through Spotify's library.

Participants

6 participants recruited using convenience sampling. 5 with previous Spotify experience, 4 who used Spotify as their primary music streaming platform.

Task

Create a playlist and add one songs from each of the genres provided. Each participant was provided a list of:

Four "easy" genres (Hip-Hop/Rap, Country, Pop, and R&B)
One "hard" genre (Americana, Emo, or Shoegaze) to introduce a more specific or drilled-down search

Experiment Design

Within-subjects experiment where all participants experienced both features (i.e., the two conditions).

The feature order and "hard" genre were informally randomized (three levels as stated above) between participants, but was not included as variable in the analysis.

Measures

Quantitative: Search time.
Subjective: NASA-TLX scores from a modified NASA-TLX questionnaire to measure cognitive workload and feature preference.

Results

Search time

Single seemed to produce the fastest keypress response while Stoplight seemed to produced the slowest.

Cognitive workload

Stoplight was rated with the highest mean NASA-TLX score, suggesting it might be the most workload-inducing design.

Feature preference

Accuracy was nearly equitable across designs, but Direction appeared to produce the most accurate keypresses.

Search Time

Participants completed the task nearly two-times faster using the Search feature (147.2s).

Mean search time

Was the task possible just using the Browse feature?

I anticipated that the "hard" genres would be too difficult to achieve using just the Browse feature. Thus, I asked if participants switched to using the Search feature to complete the task.

The overwhelming majority of participants (80%) had to switch to the Search feature when finding a song for the "hard" genre.

Did you need to resort to using the Search feature for the hard genre?

Modified NASA-TLX

The NASA-TLX was modified to a 7-point scale and the removal of the physical demands subscale to better fit the context.

Using the Search feature (3.96) produced less cognitive workload than using the Browse feature.

Mean NASA-TLX Score

Lower is better.

Preferences

We asked which feature participants preferred to use to explore Spotify's library both pre- and post-study.

In both cases, the Search feature was the preferred method, but post-study, preference for the Browse feature grew (20% to 40%).

(PRE-STUDY) Preferred music-discovery feature

(POST-STUDY) Preferred music-discovery feature

Key Takeaways

What was learned about Spotify's Browse and Search features.

Browse is being underutilized as as redundant feature of Search.

Browse provides no benefit over Search for user-driven navigation of Spotify's library. This is almost certainly attributed to the navigation being dictated by the catalogue hierarchy when using Browse. In contrast, Search allows for more flexible navigation as it is the user which dictates what content is retrieved and seen.

But, this is also a downside of Search, navigation requires intention and prerequisite knowledge of what is to be searched. Discovering new music with Search can thus be more difficult as new music can elude user prior knowledge. In this respect, Browse may provide more benefits for the platform if utilized as a feature for discovering music.

Browse rigidity limits its effectiveness as a library browsing feature.

Navigation of Spotify's library through Browse is constrained by the categories preset by the platform. This means users are funnelled toward specific content, or genres, which may not align with their intentions or fulfill their goals. This was demonstrated by the performance losses and the fact that most participants resorted to using Search for the hard genre when undertaking the task initially using Browse. Finding more obscure or specific content is very difficult, or impossible due to the constraints laid upon Browse.

Search is more responsive but requires prerequisite knowledge.

Search contrasts Browse primarily by retrieving content that is directly related to the user's goals or intentions, most of the time. But, this is done by user-driven search where the user provides a keyword(s) which triggers content retrieval. However, the user requires some prerequisite knowledge of the content they want to retrieve, and the exact content must be specified. This can include artist names, album titles, or song titles.

The imagery-first approach of Browse can be beneficial.

While the visual design of Browse constrains music retrieval, it may be of benefit for music discovery. One can imagine that the current approach can be utilized to visually direct users through more specific or obscure sub-genres of higher-level genres. For example, from "Electronic Music" to "Outsider House". The use of imagery and facilitate visual search and empower motivation, through interesting designs. This would allow users to visually discover new music in Spotify's catalogue in a relatively quick manner without needing much prior knowledge.

Search and Browse can serve distinct but complementary functions.

As has been demonstrated and discussed, Search primarily benefits Spotify as a music retrieval feature. Browse, which has been used a redundancy feature for Search, provides no benefits for music retrieval. Instead, its design philosophy can lend itself well to a music discovery function. By pivoting Browse to a music discovery feature, Spotify can provide a richer user experience for navigating and engaging with its music library.