Spotify Autotunes
Erik Bergqvist
Master of Science Thesis MMK 2016:174 IDE 192 KTH Industrial Engineering and Management
Machine Design SE-100 44 STOCKHOLM
Examensarbete MMK 2016:174 IDE 192 Spotify Autotunes
Erik Bergqvist
Godkänt
2016-11-16
Examinator
Claes Tisell
Handledare
Teo Enlund
Uppdragsgivare
Spotify
Kontaktperson
Lawrence Kennedy
Sammanfattning
Music consumption genom tjänster som Spotify ökar. Samma trend kan observeras för musik konsumtion i bilar. Många biltillverkare har börjat integrera digitala strömningstjänster som Spotify i nya bilar. För äldre bilar där en integration inte är tillgänglig förlitar sig användare på att koppla en smartphone till bilens ljudsystem. Smartphone applikationer är dock inte optimerade för detta användande och i vissa länder olagligt. Att strömma musik genom en smartphone konsumerar även mycket mobildata och batteri.
Den här rapporten täcker en konceptstudie för en retro integrerad lösning som kan strömma musik via Spotify. En fysisk design, program strategi och interactions modell har utvecklats och kan ses på sidan 27 , 13 och 19.
Koncept studien resulterade i ett konceptförslag för en självständig produkt kapabel att leverera en musikupplevelse specialanpassad för bilen. Arbetet är baserat på design thinking metodologi och människocentrerad innovation. Produkt designen och interaction modellen har ett högt fokus på bredd och innovation för att ge ett underlag för framtida utveckling.
Detta är ett examensarbete och inte ett uttalande för Spotify's framtida strategier eller planer.
Master of Science Thesis MMK 2016:174 IDE 192
Spotify Autotunes
Erik Bergqvist
Approved
2016-11-16
Examiner
Claes Tisell
Supervisor
Teo Enlund
Commissioner
Spotify
Contact person
Lawrence Kennedy
Abstract
Music consumption through digital streaming services like Spotify is increasing. This trend can also be seen in music consumption in cars. Many new auto manufacturers are starting to integrate streamed music services to new cars. For older cars where a integration is not available users rely on connecting their phone to the car's sound system. Smartphone applications are generally not designed for this purpose and in some countries illegal. Using a smartphone to stream music also consumes a lot of battery and mobile data.
This report covers the concept study of a stand alone retro fitted car integration capable of streaming Spotify. A physical design, content strategy and interaction model has been developed and can be seen on page 27, 13 and 19 respectively. The concept study resulted in a stand alone product proposal capable of providing a music experience designed for driving conditions. The work is mainly based on design thinking methodology that focuses on human centered innovation. The interaction model and physical design has a high focus on innovation to provide a broad foundation for further development.
This is an independent exploration for Spotify and not a statement on Spotify's future plans or strategy.
1 Introduction
1.1 Spotify in the car (Spotify’s own research) 1.1.1 Design brief
1.2 Goals and Limitations
2 Method
2.1 Market research 2.2 Design thinking
2.2.1 Design workshops 2.3 Design by analogy
3 Design framework 3.1 ISO standards
3.2 Minimum viable product user experience, Scenarios 3.2.1 Multi user support
3.2.2 Information feedback and input 3.2.2.1 Navigation
3.2.3 Installation of playback device 3.2.3.1 Device to car sound system 3.2.3.2 Power supply
3.2.3.3 Remote control or other UI 3.2.4 Placement, installation and Contact map 3.3 Content strategy
3.3.1 Fm radio and Pandora
3.3.2 Assets in Spotify: trilogy of content 3.3.3 Intent
3.3.4 Playlist, Browse, Pivot and passive discovery 3.3.4.1 Meta playlist
3.3.5 Middle intent strategy 3.4 Developing the interaction model
3.4.1 Voice interaction
3.4.2 Sound only model user testing 3.4.3 Expanding interaction model pivot 3.4.4 Colour codes
3.5 Designing the remote control 3.5.1 Functions
3.5.1.1 Geometrical shapes and type of interaction 3.5.2 Mood Board and general aesthetic direction 3.5.3 Rapid prototyping
4 Result
4.1 Interaction model 4.2 Hardware
5 Discussion
5.1 Navigation, content and features 5.2 Hardware design
5.3 Emphasizing voice more in the design.
References
Appendix 1
Mobile application UI Remote only
Bluetooth dongle to car
1 Introduction
With music becoming more and more accessible, often through mobile applications and streaming services music has become a common phenomena in everyday lives. Music serves as a background noise that divides some of your attention away from monotonous tasks. For this reason music is for many considered an essential part of life and a preferable addition to everyday tasks like cleaning, exercise, work and driving. (Frith, 2002)
Music can also be a very active enjoyment and is capable of inducing powerful emotions.
Music is widely considered as highly pleasurable and can produce powerful mood immersion. (Salimpoor, et al., 2009)
Consequences of not being able to maintain focus on the driving task can be dramatic.
Driving accidents are one of the biggest risk factors in everyday lives and thousands of accidents occur every year. Tired, intoxicated or distracted drivers are the biggest risk factors. Accidents caused by distraction from mobile phones have increased (WHO, 2016) Using mobile phones while driving increases fluctuation in speed, distance to other cars and acceleration behaviors. (Saifuzzaman, Haque, Zheng, & Washington. (2015)).
The increased risks of car accidents caused by mobile phone usage have resulted in a full or partial ban in 49 countries and 22 states within the USA (Cellular-News, (2007)).
Cars newer than 2006 typically have an AUX cable or bluetooth connection to the sound system but the standard interface in the car does not allow the user to change playlists and access deeper functions inside the Spotify application. Instead this must be done using the application on the phone. To counter this, certain car manufacturers integrated music streaming services to their newer cars and this has proven to increase interaction with the music services. The majority of cars still do not have these type of integrations and Spotify wants to explore the possibility of developing a retro fitted integration that can replace or reduce reliance on the mobile phone and increase availability and interaction in the car.
1.1 Spotify in the car (Spotify’s own research)
There is a great interest for streamed music services in cars, however the user experience through Spotify’s mobile application is considered poor. This has caused many auto
manufacturers such as Tesla, Ford and Volvo to integrate streaming services into new cars.
For cars lacking an integration a bluetooth or AUX connection can be used to connect to a smartphone and achieve external playback. Mobile user interfaces are generally poorly optimized for driving conditions. Common critiques are small text, small buttons, complexity and multilevel menu hierarchies.
Spotify launched a initiative to investigate the user experience of Spotify usage in cars. The initiative was composed of contextual interviews where real users were interviewed and
observed in driving situations. Social media posts were analysed for car and music related quotes and trends. The study consisted of :
● Social media data analysis of 188 000 posts related to car audio
● 14 hours of ride alongs with 6 drivers who used audio of some form during drives.
● 4 in depth interviews about needs and problems related to car audio.
The target group was 18-34 year old drivers.
The research yielded 6 key insights to problems and opportunities that can arise with music in cars. These 6 insights was expanded to provide guidelines for this study specifically.
1. One step setup: since most cars have ready to use audio sources like FM radio or CD player. It is important to have a friction free setup. Otherwise users are likely to rely on whatever is available since there are too many other distractions demanding attention.
2. En route customization: Users might not know what they want to listen to before they start their drive or they simply get bored and want some variation. For these reasons it is important to be able to make adjustments as you go along.
3. People pleasing: Car audio can be a source of conflict if the music is mismatched between the people in the car. It can however also be a bounding experience if a match is made. It is important to be able to find content that can suit several persons at once and too much personal customisation of content can be a hindrance in this regard.
4. Transitioning: A car ride is often just part of a journey and at some point it is likely that users will want to switch audio to a different source like headphones. Or they want to wrap up their session with a finishing song or listen to the end of an interview.
5. Mood immersion: Music while driving can be a strong source of mood immersion.
The sense of isolation can be compared to singing in the shower. Believing that no one is watching or hearing you can make you lose inhibitions and do things you would not otherwise do like sing along or listen to artists or songs that you normally would be embarrassed of.
6. Anytime versus specific: FM radio especially can be a source of passive music discovery, new music is introduced regularly that might fall outside of a person's normal listening habits. There is a value in listening to music that you are already familiar with and being introduced to new music. Finding a balanced ratio between the two is going to be a key to great users experience.
1.1.1 Design brief
The retro fitted solution should be a stand-alone physical product that is capable of streaming music through a build in 4G connection to the car's audio system. The product concept should strive towards low cost.
1.2 Goals and Limitations
The goals of this project is to visualize a minimum viable product concept that could solve the usability problems and be used for further product development work.
● To develop, validate and user test an interaction model that provides a stand alone music experience that allows the product to work without aid of secondary products like a smartphone or similar. This is a concept study with the focus on exploring divergent ideas in order to give a broad foundation for future development.
Delimitations:
● Exact specifications of electronic components are not available or such calculation falls outside of the scope of this thesis work. Rather existing products will be used as estimations using a design by analogy method. (Moreno, D. P.(2014))
● No software will be developed to user test interaction models.
This thesis work has been done in cooperation with Spotify at their office in stockholm but does not represent any statement on spotify’s behalf, nor does it reflect any future strategic decisions from Spotify.
2 Method
This section will cover some of the tools, method and theory used throughout the process of designing the car integration.
2.1 Market research
In order to understand the competition, finding existing solutions, potential for improvement and mistakes to avoid. Extensive market research was performed using google search, tech magazines, internet forums and visits to resellers (Long S.J 2013). Existing products were analysed by physical samples when available and trough images when not. Relevant products was compiled in a state board, See fig 1
fig 1. State board of existing product on the market. None of these products are intended specifically for the use case of this report but are intended for similar use.
From the market research certain trends and problems were detected. Many devices that have been designed for in car use and are often cluttered with buttons and small low
resolution screens. Typically they appear to be designed to be small, cost efficient and to deliver a large/broad function. Because of the limitations of low resolution screens buttons often need to be designated to one function and some sort of explanatory graphics are printed on them.
Insights from the state of art research led to certain realizations and principles.
● Having phone app for advanced settings is a huge design benefit. There is no need to design for full functionality on the device UI, only for the use case
● Too much reliance on visual interaction, buttons usually have small printed text or symbols.
● Many buttons are used for multiple functions depending on which mode the device is in.
● Devices are designed to be placed in primarily one place in the car and can not easily be moved around. This makes them dependent on car model to be compatible
● There is no obvious standard placement which implies that there is no satisfactory solution to placement on the market.
● Bad polymer choices makes devices appear cheap and does not match the higher standard of materials that are common in a car's cockpit.
To determine potential placements of the playback device and remote control the automotive environment had to be understood. For this a field visit to a used car lot combined with examining images of car interiors was used. See fig 2,3.
fig 2. Car sales lot in täby, stockholm
fig 3. Map of potential attachment points for external products.
NO Name Comment
1 Sun shield/ back view mirror Great variation of size and shape. Risk of obstructing functionality.
2 Steering wheel Large variation of size and shapes. Risk of interfering with steering.
3 Ventilation outlets Some variation of placements and size. Fairly standardized. Shape of outlet consistent.
4 Handbrake / gear shift Great variation of size and placement. Missing in some car models
5 Center console Great variation in size and layout.
6 Open storage places Great variation in size and layout.
table1. Different attachment points for the remote control.
No standardized layout or set of features could be found. The cockpits of cars varied depending on manufacturer, model, year and type. Some components existed in all
examined cars but where these was positioned could also vary greatly. Ventilation outlets for the car's climate control was present in all cars. This has been exploited by many auto accessories producers and the ventilation outlets can be used as an attachment point. See table 1.
The main unit will be bound to the power source and its position varies depending on the location of the 12v outlet.
2.2 Design thinking
Design thinking is an iterative process of research, ideation, prototyping and validation.
Unlike an analytical process design thinking relies on divergent and convergent thinking.
Research is done in order to understand the underlying problems that surface in user testing and dissatisfaction with current products. Solutions are explored in a divergent manner through creative ideation tools like brainstorming and prototyping. These ideas are then user tested against principles found in the research and are then validated or dismissed. These cycles of research, ideation and testing are iterated until a satisfactory solution has been found. (Brown, T (2009)) (Buchanan, Richard. (1992))
2.2.1 Design workshops
The workshops consisted of 5-8 participants from varying departments within spotify like engineering, management and design, see fig 4. The diversity of participants was
considered important to get input from as many different stakeholders as possible. Having a diverse design team allows ideas to be improved from a wider base of knowledge (Plattner, H., Meinel, C., & Leifer, L. (2014))
fig 4. Picture from one of the workshops. Ideas are described on post it notes and clustered in themes or similarities.
A total of 8 workshops was held focusing on different user groups.
From these workshops the user needs and behaviors could be arranged in scenarios that the product would need to handle. Three main scenarios was established unboxing and installation, multi user support, navigation/information feedback/input.
The workshop participants also served as a focus group for expected control options. The most essential controls were considered to be : Skip song, Play/pause and change playlist.
2.3 Design by analogy
There is no determined electronic components or electronic design that the product will use.
Because of this quantities like size, volume, proportions and certain minimum lengths have been estimated by design by analogy. Design by analogy is a method of associating properties of one domain to another, in this case existing products with part functionality to the concept proposal (Moreno, D. P.(2014)). From the design brief and goals there is no
indication that any new technology development will be needed and therefore analogous estimations can be made for all components.
By using the design by analogy method, a volume prototype of the main unit was produced see fig 5. This prototype was used in cars to determine the viability of its structure and size, being one piece and not for example connected to the 12v outlet by a wire. There is no standard available for the 12v outlet in regards to placement and volume allowance. The model in fig 9 was used to get a visual estimate of the space that could be expected around a 12v outlet. Typically the amount of free space from the outlet varied by angle. In all
examined car's 180 degrees of the outlet offered considerably more space then the remaining 180 degrees. The 12v outlet allows for rotation and by having the majority of volume in one direction the user can position the device to fit even tight installation points.
The volume is estimated by using a Huawei carfi model E8377 and samsung gear S
smartwatch. Two products that together provides wireless internet connection (Huawei carfi) and a playback device (smart watch)
fig 5. Volume mockup of base unit featuring main electronic components. The mockup is attached to a 12v outlet located inside a car's cockpit.
3 Design framework
This section explains and motivates the principles, guidelines and constraints that the final product had to conform to. This section covers the principles for the stand alone solution and alternative setups for solving the user problems and pain points are described in appendix 1
3.1 ISO standards
There are several iso standards regulating the design of in car accessories and systems.
The iso standard’s cover the areas of installation, operation and information feedback.
For a product to be compatible with iso standards it may not interfere with the primary controls of the car and must be compatible with the car manufacturers standards. Products intended to be used while driving must be placed so that they can be operated without reducing the driver's ability to handle the vehicle. This means that products may not be placed in a way that they risk obstructing the use of controls like turn signals or backview mirror. (ISO 4040 (2001))
Controls may not impede the driver's ability to handle traffic situations or demand attention in a way that can affect safety. The controls should be placed as close to the driver's normal line of sight as possible to minimize the time where visual attention is not on the road.
Displays should not distract the driver from accessing other important information (ISO 4513 (2003)) . Drivers tend to divide visual attention between interaction elements in the cockpit and the road. The driver's main attention should be on the road and systems should be designed to allow interaction by fast glance looking, minimizing the amount of time the driver spends looking for and at a UI element . (ISO 15007-1 (2002) ) (ISO TS 15007-2 (2001) ) ( ISO 15008 (2003) ) ( ISO FDIS 16673). Considerations for glance looking become extra important when dealing with mobile phone UI. Some people are performing ill advised actions with their phones like texting and driving and during this behavior the driver's attention is constantly shifting from the phone to the road.
For this concept study compliance with ISO standards are not necessary but they should be observed. The established design requirements can also be used to guide ideation and market research.
3.2 Minimum viable product user experience, Scenarios
A minimum viable product (MVP) is defined as a product or prototype featuring only the minimum set of features to successfully perform its purpose and be successful on the market.
In order to determine the minimum viable product several scenarios were described. These scenarios was determined from the design workshops as well as from contextual research observations. From these, three essential scenarios were formed as a boundary for the minimum viable product.
3.2.1 Multi user support
Since the device is intended to be a permanent addition to the car it must allow more than one user to use it. This can involve passengers and other drivers such as family members borrowing the car. To successfully handle this scenario the device must be able to handle account switching but also to handle multi users that are not changing accounts. Account switching will be handled by the native client either by connecting to the device directly or by a remote link such as 4G or wifi. For the device to handle multi users that are not changing accounts the standard user interface must be able to find content that is diverging from the main user profile. This can be done by providing generic listening options as well as specific options through a search feature like voice search or manual typing through a smartphone.
Which of these options that will be used is dependent on the level of intent by the user.
3.2.2 Information feedback and input
Since there is no accurate way of predicting the user's mode or music desire, there is a need to navigate content and to maintain a orientation. This requires information input through a user interface but also information feedback. Information feedback can be done by visual, haptic or audible signals. Since visual attention is the most severely limited one ISO TS 15007-2 (2001) a graphical interface is not suitable. A graphical interface would also be an expensive solution and much of the biggest advantages of a screen can be substituted by temporarily connecting a smartphone with a suitable application or feature.
A haptic feedback system would require an extra component and, compared to sound cues, consumes a lot of energy. This feedback system could be problematic with battery usage.
Relying on sound cues only would be the best alternative.
3.2.2.1 Navigation
When users know what they want they will need a way to navigate to that content and when they are browsing for new music they may need a way to return to a previous point. This can be solved in a high intent way by having a voice feedback, that tells the user what he is listening to. Or a voice search that allows the user to navigate to a known point directly. In a low intent setting this can however be tedious or obtrusive to the music experience. In a low intent session it is better to rely on information feedback or input that is not audible like colour coding or numbered presets.
3.2.3 Installation of playback device
Installation must be simple and designed so that any user can install their device without any prior knowledge or skills. Preferably no instructions should be needed. Measuring how easy something is to install is not a simple task and simplicity must be determined on a
comparative basis. That is of two options which is considered to be the easiest. Number of steps, with few being preferable, can be a guidance but not a absolute determining rule.
There are also factors like price and sound quality to consider.
Installation is in this case defined as connecting the device to the car’s sound system and power outage, as well as connecting the device to a Spotify account.
3.2.3.1 Device to car sound system
There are three widely available ways of connecting a playback device to a car's sound system Fm transmitter, Bluetooth wireless connection and auxiliary cable.
FM transmitter is the only available option when there is no AUX port or bluetooth connection available. This option does increase the total obtainable market but requires additional hardware, hardware that is most likely not going to be used if the car has a AUX or bluetooth connection due to poor audio quality, caused by interference and size restrictions of antenna, and a less intuitive setup. FM-transmitters can be purchased as a add on product, several product that would be suitable to this use are available on the market.
Bluetooth is as well as Auxiliary generally available on cars made 2006 or later. Both options provide a good sound quality option but bluetooth can be trickier to connect, mostly due to many different protocols used by different car manufacturers and a dependency on a digital interface that may or may not be intuitive. Many users also use their bluetooth connection to connect their phones for hands free phone conversations, this option should preferably not be impaired by the playback device.
A auxiliary cable enables users to connect their phone as well as the playback device and the installation is to be considered the simplest of the three options. Having an Auxiliary port and detachable cable will allow multiple cable lengths as well as enabling the connection of an external FM-transmitter device. A auxiliary connection is recommended.
3.2.3.2 Power supply
The playback device will need energy to function and there are two options available for connecting to the car's electric power system. External power sources like battery, solar panels or windmills on the roof are dismissed due to practicality and any rewiring options are
dismissed in favor of simple installation. It is possible to use a car's OBD port or 12V outlet to power the device, these options are similar in terms of connection but the 12v outlet is
typically available in the cockpit and is intended for similar use. Many users rely on charging their phone through a USB port adapter to the 12v outlet and for this reason it would be preferable to include a usb port to the device in order to accommodate this use.
3.2.3.3 Remote control or other UI
The main ui for the device is likely going to be placed in the center console of the dashboard.
If the remote control is connected to the device by wires this will require some management of these wires so that it is not tangled or caught in clothes or luggage. Since the playback device itself can be located on many different positions in the cockpit or possibly even be located in the trunk the length of any wire must be adjustable or long enough as standard, this adds more complexity.
Having a wireless connection from the remote to the device is preferable in this regard and will require batteries in the controller.
3.2.4 Placement, installation and Contact map
From the design framework and market research a contact map of how the different parts of the product relate or communicate to each other was drafted, see fig 6. This is to be
considered the ideal setup from a usability aspect.
fig 6. Contact map of the products components.
1. Magnetic dashboard ventilation mount. Hold the remote in place by magnetism and friction. These are available on the market today fits most ventilation outlets. Other options like glue, tape or suction are available. Glue and tape may leave residue and users might object to damaging the interior of their car. Suction cups requires a suitable surface, which can not be guaranteed across all car models. The magnetic vent option was chosen due to its simplicity and proven concept.
2. The remote control is powered by batteries and due to the long expected life span of the product (preferably over 3 years) AAA batteries were chosen to facilitate simpler battery changes. Rechargeable or special batteries would have been preferable for the physical design of the remote control but would have resulted in a bigger pain points over the lifespan of the product.
3. Remote control where the principal interaction with content is intended. This interface could have been placed on the main unit itself but due to potential ergonomic
problems from poor placement of the 12v outlet these design concepts were discarded. Microphone for voice commands are placed on remote to ensure microphone is in range.
4. Wireless bluetooth connection to main unit. A wire between the main unit and remote would pose several design problem.
5. Main unit containing 4G connection storage and playback device.
6. Main unit is powered by 12v outlet.
7. Auxiliary cable to car's sound system.
8. Communication with Spotify by 4G cellular connection 9. Spotify’s main client. Account and content handling.
10. Spotify Connect protocol allows users to remotely control playback device like any other sound system connected to wifi or 4g.
11. External client like smartphone, desktop or tablet. UI for advanced settings and high intent engagement.
12. Cars own sound system.
3.3 Content strategy
Before a interaction model or physical design can be developed a content strategy must be considered. The content strategy will address the user's needs and how content can satisfy these. The content strategy will be tailored to the specific conditions related to driving. This means a high focus on simplicity, easy interaction and navigation (availability) and the ability to change context of what is being listened to. Context switching requires a pivot option which is the ability to influence the content in a less direct way. Pivoting can be seen as a reaction to an event like a favorable playlist or song and deciding to play moore of that.
The user will engage a service with different intents, sometimes he will only desire some background noise or audible distraction from the monotonous task of driving (Dibben, N., Williamson, V.J., (2007))With this intent it can be said that the music must not be bad rather than good since the user isn't actively listening anyway. It may also be that the service is engaged with a highly specific intent and a specific song or playlist is desired. There is also a middle ground where users have a relatively high criteria of what content will be considered acceptable but does not know what that is. This use case will be referred to as middle intent.
3.3.1 Fm radio and Pandora
Fm radio is a popular content format and has been present in cars for decades. Fm radio however suffers from a inherent customization flaw since each station is niched towards a group of people rather than a personal profile. Despite this people are satisfied enough with this model to continue using it. This indicates that full personalization is not required for a
successful content strategy. Whether or not the Fm model will be successful depends on the current intent of the user. It will not suffice in a high intent session when users desire to choose each song, an intent where Spotify typically succeeds but is unable in the driving environment. It will however excel in a low intent session when the general theme is the only criteria for the user. The gap where a new content strategy can be developed lies in the middle intent.
The limitations of Fm radio also adds value in that songs the user would not normally listen to, yet aligns with the general theme of the station, will be feed into the content stream. This contributes to the phenomena of passive discovery. Since radio stations provide content predetermined by a radio host or DJ there is also a human connectivity aspect to this discovery. Finding new music selected by someone else is a sublime form of social
interaction and future reasons for listening to a station can be influenced by this kinship. By analysing how radio stations are marketed this can be further observed in that radio hosts and talk content providers are often a key marketing aspects, see fig 7.
fig7 Rix FM add in stockholm, Rix FM is a popular radio station in sweden
Aside from Fm radio there are two other fairly successful content formats provided by
Pandora and Sirius XM. Both services can be described as a personal radio and stations are tuned towards choices made by the user. They still suffer from lack of choice and a high intent options, commercials between songs also contribute negatively. These services provides more personalization than Fm and enters the middle intent from the low intent perspective. Spotify needs to enter it from the high intent side, and preferably from the low intent side as well. A strategy for this is shown in fig 8
3.3.2 Assets in Spotify: trilogy of content
Spotify is more than just a music library and being able to discover music is one of the core features of the application. Spotify allows users to find content by 3 essential methods, automation by algorithms, curated playlists arranged and updated by Spotify employees and by manual searches by user. These three aspects often relate to each other, for example a
manual search can yield curated playlists or content chosen by algorithms. And algorithms can choose curated lists or personal library as suggested content. Neither of these three components can deliver complete success on its own and a optimal user experience is achieved by maximizing the contribution of each aspect. For this project the algorithms and curated lists can not be altered directly but by maximizing the user input the benefit from Spotify’s internal assets can also be better utilized.
3.3.3 Intent
In fig 8 different intents are displayed on a scale of low to high or from a user that knows what he wants (High intent) to a user that does not know what he wants (Low intent). For low intent users an automatic selection of music based on profile preferences might suffice otherwise a suggestion feature is needed. Allowing the users to browse a library based on broader terms like genre or moods assuming that the user in this case does not know exactly what he wants it is better to make suggestions until a match is made rather than putting any cognitive strain, in order to make a direct choice, on the user. In a higher intent session it can be assume that the user has a good idea about what he needs. In these sessions it is better to allow the user to make a direct voice search or connect the graphical interface available in the smartphone application. The general principle for the model is to provide a low intent option as default and let the user increase engagement from there.
fig 8. Model of intents and how features relate to current intent. Intents are scaled from low to high or users that know what they want to users that don’t.
3.3.4 Playlist, Browse, Pivot and passive discovery
From research findings and observations done throughout the project it appears that a high intent session can move to a low intent session passively and low intent to high actively. A highly niched playlist can seem repetitive if the user is not in a high intent mood and a high intent listener can find a diverse playlist to feature unpreferred content. Users are likely not going to be in a pure low or high intent but rather exist in a gradient between. The cognitive stresses of driving and lack of a graphical interface makes this use case an even greater challenge.
This challenge is the main motivator behind the search for a pivot option. A “pivot” option means being able to find new content by choices like “more of this type of music”. Users
should not always have to make a direct choice in listening since this can be considered a high intent action. Not doing anything can be considered the low intent option and a pivot function is a middle intent action.
3.3.4.1 Meta playlist
Music listening is highly diversified and there are numerous ways of classifying content and listening habits. Perhaps because they are not rooted in music taste or similar but rather human preferences and social constraints. Because of this there are infinite ways of sorting music depending on what criteria you use. There are however some that occur more frequently like genres, mood and decade. In these general themes there are several
common sub divisions. A song can belong to several different clusters of music, and themes like genre and mood can have several overlaps in content. Since there is no clear hierarchy or established sorting norm, any playlist can be considered a subset of another and can be further divided into smaller categories. Rather than having many large playlists that overlap frequently it is better to have playlists that consist of a narrow theme that can be clustered to larger themes as a meta playlist. Arranging playlists like this gives greater freedom in
combining them into a customized session. Depending on intent there is also shifting needs for customization and having a meta playlist option matches the general interaction model of low intent content first and allow interaction towards higher intent content.
A high intent session has greater allowances for interactions than a low intent session and therefor you first get to the low intent point then interact further to get to the higher intent content.
In the middle intent area it makes further sense to select clusters of music rather than specific songs to minimize amount of interactions.
3.3.5 Middle intent strategy
In the intent model shown in fig 11 music is divided into two sections that are based on intent rather than music theme. The user either knows what he wants or he doesn't. When the user knows what he wants he will need a way to navigate to it and when he doesn't know, he will need help to find it. The content strategy is to divide these use cases into two features:
known content and suggested content. The own library shelf allows user to access known content through presets or through voice search. When the user isn't sure about what music he wants to listen to he can try the suggested content feature that suggest music until a match is made. These two features are visualized in fig 9.
fig 9. Content deliver by the two suggested features. Own library on the left and playlists curated by spotify to the right.
The amount of features was limited to two to avoid feature creep. The division of known versus unknown was considered the most universal and polarizing options. In the intent model shown in fig 8 these features can be placed on either side if the middle intent gap.
Browse is a session initiated when the user don't really know what he wants and is open for suggestions. It can also be that a passenger has joined and a need for a common choice is now there. Once a suggestion that suits the current mood is presented it may trigger a need to go higher in intent and a pivot option is needed.
The browse tab should start of making guesses for the user then diverging from that. Finding new music for main user or something for a passenger.
If the thumbs function isn't there, then the navigation model will require something else to keep it from becoming too deep and complex. I believe that a flat menu is extremely important and should be a requirement. A pivot function is a way to keep the menu flat but with depth.
If the “navigation” isn't based on “just play me some music and i'll see if i like it”. The browse tab needs to be static so that you can navigate back to a playlist or keep some kind of navigation. That also requires the user to somehow know what he wants.
Playlist feature is about moments when the user knows what he wants, starting from a higher intent. Once that initial need has been satisfied there may be a need to go broader, zoom out to go to a less specific session with some passive discovery. Same goes for voice searches. There is a thin line between giving the user what he wants and what he needs.
Finding that balance is going to be one of the main challenges. As a principle i believe that you should deliver the low intent option first and then provide a way to move on from there, especially since visual and cognitive attention is low.
3.4 Developing the interaction model
The remote control is essentially a physical UI and should emphasize the functionality and the interaction model well. Before a remote control design could be determined a interaction model had to be developed.
Screens and graphical interfaces are undesirable due to economic and safety reasons.
Graphical interfaces require visual attention which severely limits the user's ability to drive safely ISO TS 15007-2 (2001). A interaction model based on audio only was considered preferable.
3.4.1 Voice interaction
Many products and applications have tackled the problem of driver distraction by relying on voice interaction. Voice interaction brings the advantage of being able to perform the desired action without taking your hands of the steering wheel or demanding visual attention.
Voice interaction does however face certain problems in user experience, interaction time and performing complex interactions. Currently there is no product found on the market that can provide a satisfying example of a fully functional voice interface that does not provide cognitive limitations and risks lowering engagement with a service.
Voice does however excel in direct searches and compared to typing your query it brings significant advantages in the context of driving. Having voice available also enables searches of song titles with no visual guidance of a screen or other graphical interface.
Direct searches requires the user to know what song or genre they are looking for and because of this it can not completely replace any other interface.
3.4.2 Sound only model user testing
In order to test the interaction model without having functional software, a Wizard of Oz test (Bella, M. & Hanington, B., (2012)) was performed on 8 participants. A live version of Spotify was used on a desktop computer together with a dummy remote control, see fig 10.
fig 10. 3D printed mockup of a remote control.
Test participants was allowed to freely interact with the remote control in order to perform certain tasks. The test administrator used the desktop client to simulate the user's
interactions. The “skip” button changed songs within the current playlist and the “playlist” or
“browse” function changed playlists. The users did not have access to any graphical
interface but the interaction model can be visualized as in fig 11. This interaction model is referred to as single shelf and each card represents a playlist. Two playlists of similar colour do also feature similar music as each other for example all red playlists feature rock music.
To simulate the distraction users face in driving situations a virtual driving simulator was used. Test participants were required to stay inside a lane of a virtual test course.
fig 11. Visualisation of the Single shelf interaction model. Content is navigated from left to right and the axis is unspecified in regards to genre or theme.
Users were allowed to interact with the dummy remote control, seen in fig 10, the test leader changed playlist or song when users simulated a button press. Test participants were asked to perform two tasks one for their own library and one for Spotify’s curated library. Users were required to login to their own account in order to benefit from Spotify’s profile settings.
A total of 8 tests was performed on users aged 22-46. All users had previous experience with spotify as a service and held a driving licence.
Task 1: Personal library
By using the remote control, find a playlist that you recognize and find suitable for your current mood.
Task 2: Spotify playlists, passive discovery
By using the remote find a playlist that suits your current mode.
All test participants were successful in completing their tasks and the average number or interaction for task 1 was: 2-3
and task 2: 3-4 .
During the test a contextual interview was performed and several users expressed desire to be able to find one of their own specific playlists faster and to have a pivot option during task 2. For example to be able to choose similar genres from the one they found in the browse tab.
“I can hear that this is a rock playlist but i prefer more classical rock or metal”
-Test participant
The sound only model showed promise but it lacks depth. As can be seen in fig 16 several playlists of the same genre or theme comes after each other in the content shelf. It is hard to determine which playlist comes first and makes it hard for the user to do any navigation. A single shelf only interaction would rely too much on algorithms to provide music and lacks a pivot option.
3.4.3 Expanding interaction model pivot
Based on desires from the user testing, three more models were developed.
Single undetermined axis grid navigation, two axis undetermined grid navigation and single shelf with pivot option.
In fig 12, the two undetermined axis grid navigation model is visualized. This model can be said to mimic the home menu in the spotify application. In the home menu algorithms selects playlists most likely to match the user's current mood. The home menu provides shelfs with playlists selected based on a criteria like “recently listened to” or “workout”. Both shelfs and playlists can change depending on when you access the home menu. This makes any mental model for the home shelf rely on having a graphical interface to guide the user. In appendix 1 a mobile app interface using the same strategy as for the remote can be seen.
For the sound only model a new way of navigating the grid would also be needed.
fig 12. Two undetermined axis grid. Playlists are not sorted in rows or columns.
Since there is no definitive way of arranging music in columns or rows sorting the axis in the grid navigation models was problematic. You can't successfully, for example, sort music on axis like higher tempo or more bass and expect it to work for all users. For the two
undetermined axis model users would not be able to determine any difference between moving up/down or left/right. For this reason it had no major practical difference from a single shelf navigation and creates confusion and disorientation.
fig 13. Single axis grid, playlists are sorted in columns by genre or theme. Rows are unsorted. Users can interact freely in the grid and similar playlists can be accessed by moving vertically. Moving horizontally will change genre or theme.
The two axis model needed more structure than the two undetermined axis model but must still have some dynamic aspect in order to not be permanently fixated.
For the Single undetermined axis model the determined axis was sorted based on themes like genre, decade or moods similar to how the single undetermined axis model is sorted see fig 13. This would allow users to find a general theme like “Rock” and then navigate between similar lists.
From the wizard of OZ user test it could generally be said that users recognized the general theme of the current column. It was however uncommon that users could determine sub genres accurately between rows. The sound only model required users to guess the entire theme of a playlist on a song by song basis which they were unable to do confidently. The rows served little purpose and playlists in the column could just aswell be mixed since users were unable to determine differences between them as a whole playlist.
fig 14, Dynamic single row model with pivot option. Users can navigate in columns like the original single row model and interaction in the rows are done indirectly by skipping songs.
Based on realizations from the two axis models and to allow a pivot option a new dynamic single row model was developed see fig 14. By having several rows but allowing only direct interaction in the columns users would be able to navigate directly between broad genres or stations and tune in to a playlist by liking the songs belonging to it. Each column was still a mix of the corresponding rows but the like/dislike option would change the composition or priority of the sublist in the column, see fig 15.
fig 15. Left: Playlists from one column start with the same proportions. Right thumbs upp/down function changes proportions and priority in the playback queue.
The single undetermined axis grid model also features a pivot option since similar playlist can be accessed by moving in the rows. In this model the thumbs function would be
replaced by vertical arrows. Both options feature a viable pivot option but the dynamic single row model was more in tune with the intent model defined in fig 11, and the general principle of providing a low intent option first and a higher intent option by further interactions.
3.4.4 Colour codes
The interaction model is mainly based around guiding the user to content. But if the user knows what he wants but doesn't want or is able to do a voice search, a colour code system can be used as a middle intent option for this use case. Knowing for example that your workout playlist is under “red” or that the “purple” playlist contains music suitable for commuting allows for a quick navigation to this content. Colour codes can work as an non interfering addition to the main interaction model.
3.5 Designing the remote control
The remote control needs a physical interaction that matches the content interaction model.
The development of the physical user interface is dependent on the interaction models and most hardware choices are bound to the interaction model and content strategy.
3.5.1 Functions
From the workshops the functions skip song forward and backward as well as play/pause was considered to be minimum for basic functionality. The content strategy added the own library feature and browse feature. To be able to make direct searches voice search was also added, see table 2. The interaction model added them pivot option represented by thumbs up/down.
Function Description
Skip song Ability to change song to next or previous in playback queue.
Play/pause Start and stop music playback
Thumbs up/down Pivot option, to get more or less of similar genres and themes Browse Playlists provided by Spotify with high chance of discovery Playlists Own library, content known and curated by user
Table 2. List and description of functions.
All of the functions in table 2 had to be accessible from the remote while maintaining a flat menu structure.
3.5.1.1 Geometrical shapes and type of interaction
Since visual attention while driving is limited the remote needs to be handled with minimal or no visual demands. One way to allow this is to utilize the geometrical layout of the remote by varying the size, angle and tactility of buttons. Several concepts of different geometrical properties was made in a CAD software and was 3D printed to get a physical model to test interaction with no visual attention. See fig 22.
fig 22. CAD, 3d printed and clay models exploring angles and geometrical properties.
The 3d printed models proved that you could find the right buttons without looking and that the geometrical layout was less important than expected. Few features and to avoid things like switches between moods etc was more important factors. A simple interaction model is prefered and to keep the menu system in this as flat as possible. A flat menu system means that no button should change function based on previous interactions. Ideally each button should only hold one function and each function should have a dedicated button.
There are many different components available for input on a remote control. For example buttons, joysticks, knobs and scroll wheels was all found in the market research. All of these have their own advantages and disadvantages when compared to the intended interaction model. Screenless touch interfaces was not found in any existing products. However a
gesture based interaction utilizing touch would suit the slower interaction behavior that comes naturally with listening your way through content. There is for example little use in being able to skip twice in quick succession if you don't know what the next songs in the playback queue are anyway.
3.5.2 Mood Board and general aesthetic direction
Spotify as a brand does not exist in any physical product and all design guidelines available are intended for use in digital user interfaces or graphical art. Physical language aspect must still be considered and defined. To help in this endeavour a mood board was composed which focused on a sublime language that mainly tried to fit into an automotive environment rather than express Spotify as a brand, see fig 16. Branding is already present in the music and personalized content.
fig 16. Moodboard of aesthetic direction. The three key words automotive, intelligent and simplicity are described in top left corner.
The main themes for the product became Simplicity, intelligence and automotive. These aspects are to be conveyed both aesthetically and through interaction.
Simplicity is expressed by avoiding any unnecessary functions, aesthetic elements or features. Keeping the menu system flatt and not tying in several functions to the same buttons. Promoting functions that directly relate to music content. A generally rounded and soft form in order to promote interaction and preferably to grip the remote fully.
Intelligence is about conveying the many features that is behind the functions like playlists , personal and curated by Spotify. The interaction model should guide users and allow them to influence their music session in a less direct way. A play more or less of this feature (pivot option) that expresses the intelligent functions provided by the Spotify service. Users should be able to quickly access a specific playlists or song but also be able to find content by low intent choices.
Automotive is the biggest aesthetic guideline and is about making the device and remote fit well into an automotive environment. Important factors here are colour, material and
placement.
3.5.3 Rapid prototyping
For the visual, functional and volume prototypes a combination of sketching, clay modeling, CAD visualization and FDM 3D printing was used. The purpose of these prototypes was to quickly get a sense of physical and aesthetic properties of the design concepts. The process is iterative and can start with sketching designs in a two dimensional environment, see fig 17.
fig 17. Collection of sketches for different solutions and ideas.
If the design shows promise in two dimensions you can start exploring with a third
dimension. For this purpose clay is used to quickly get a sense of volume, If needed further processing is made to examine functional or ergonomic properties. See fig 18.
fig 18. Clay models in different stages of refinement.
To get a sense of aesthetic promise and to allow quicker variations in volume and details, the design is made into a full 3d model using CAD software. These models can then be used to render images that simulate the material and color of a final design. See fig 19.
fig 19. Rendered images of 3D models.
The 3d model can also be used to 3d print models of the design using a FDM printer. This is done mainly to verify and variate volumes. See fig 20.
fig 20. 3d printed physical models of the computer aided designs.
In fig 21 a clay model representing the final design can be seen. This model have AAA batteries incorporated in the model to validate volume requirements. This model was used as a reference for further modeling using CAD software.
fig 21. Functional clay model featuring AAA batteries to validate volumes and sizes
4 Result
By combining the conclusions from the hardware setup with design principles of the remote control and interaction model a final concept was produced.
4.1 Interaction model
The interaction model is based on audio only and is described in a flowchart, fig 23.
fig 23. Interaction model described in a flowchart. Decision making focuses on simplicity and choices comes down to liking the current song or not.
4.2 Hardware
This chapter presents and visualizes the final concept of the stand alone retrofitted spotify integration.
fig 24. Complete product. Main unit, remote control, magnetic vent mount and auxiliary cable.
The complete product consist of four parts, the playback device and remote control. But also a changeable auxiliary cable and a magnetic vent mount. See fig 24
fig 25. Remote control for device. Most interaction related directly to music selection is done through the remote.
The remote control is powered by AAA batteries and is buttonless, see fig 25. Gestures are interpreted by the capacitive touch surface. The ridges help guide the user's finger and gestures are confirmed by audio feedback and by having the icon on the remote light up, see fig 26. The remote is connected to the main unit via bluetooth see fig 6.
fig 26. Gestures for remote control.
Current shelf of content is lit up and the different playlists are assigned to a colour code, see fig 27. The playlist and browse functions are described in fig 9.
fig 27. Colour coded presets are shown by highlighting the current shelf.
fig 28. Main unit featuring most electronic components. Playback unit and memory.
The playback unit features a 4g connectivity and storage capabilities. It is one piece and connects directly to a 12v outlet. The backside features two USB outlets that allows users to connect a phone charger see fig 28. The 3,5 mm auxiliary outlet connects to the car's sound system. The product concept is visualized in an automotive environment in fig 29.
fig 29. Visualization of product mounted in a car.
5 Discussion
I believe strongly in the design thinking methodology and the main principle is to start with user needs (derived from observed problems/pain points) and choose and develop
technology accordingly. Backtracking until you find the initial needs that drives users towards a product/service and develop solutions based on that rather than the superficial problems you encounter in user tests. For me it's rather about human centered innovation, industrial design and interaction design are subsets of design thinking. In design thinking it is common that you change focus from the original problem formulation and instead step backwards to explore a new option to move forward. In the case of this study the focus was shifted towards designing a stand alone music experience or service that could work in a driving situation. For a viable stand alone experience the scope of this project had to be
backtracked to developing a content strategy and interaction model before any physical concept could be developed. Taking this broader scope made the thesis a considerably larger project than initially intended but this was necessary to achieve a credible end result.
Diverting attention to the interaction model and content strategy came at the expense of the physical design and because of this most attention has been put on the remote control since it was considered the most essential component.
5.1 Navigation, content and features
The questions regarding navigation has been, “if you don't have a screen what do you navigate to?” and “If you don't know what you want how do you navigate to it?” Browse is a feature in its own that gives suggestions and thumbs is a tool for passive discovery/curation.
The user need that can be satisfied with this approach is the need for suggestions and passive discovery, and the ability to navigate is not a user need as far as i can tell from the research, rather it is quite the opposite. Any form of navigation implies a cognitive load.
Users that don't know what they want need a simple way to get suggestions.
I believe most users don't know what the different shelves in the app really means and therefore relying on a pre existing mental model is not ideal.
Having the browse content easily available rather than in one of the many tabs in the app will also be a great opportunity to increase usage of this content. Users that don't know what they want are likely to go to the only known point of content, presumably own library.
Before the user decides to take action there is a window to present options instead. Since music is such a natural phenomena in cars there is a greater allowance for instant playback which should be exploited.
From the user tests people were surprised by how good the curated lists and general UX of that was. And i believe the car can be the primary place for curated content. Even if it's from a driving mode in the application.
If Spotify really wants to push the curated playlist content I think a simple UI/UX like the remote is the place to do it.
I also don't think that curated vs radio, cards vs stations or some other form of content is essential here, rather it should be whatever delivers on the core need of suggestions. “Just play me some music, i decide if i like it”. Familiar tracks makes you feel comfortable with the
list, new songs keeps you updated and gives you passive discovery. It doesn't have to be all new or perfect customization like what Spotify’s radio feature tries to deliver.
I think this fundamental need will always be there and Spotify will most likely always provide some feature that addresses this, (cards, radio, other user playlists etc) reverting back to own playlists only seems unlikely.
Currently i believe that lists curated by a person delivers better on the passive discovery needs, the human connection and uniqueness of Spotify.
5.2 Hardware design
There are several functional considerations i think are important.
Wired concepts was dismissed because it is hard to make it look good attached to the dashboard, you can't as easily grab it and use it as a handheld device, you can't as easily just put it away when done using it, or give to a passenger. A wired remote also risk getting caught in seat belts and jackets or other luggage grocery bags etc and snap the wire, or just get tangled, having pens/toys etc caught in it. Depending on placements in car wire length is going to vary greatly.
A wireless solution gives more freedom and i don't see any real disadvantages to it, assuming battery pain point is solved by for example AAA batteries.
Joystick options are just as risky as touch when it comes to wrong input, dropping it, grabbing it wrong or feeling your way to the buttons or placing some other object on it etc.
Having the UI on the plug directly was dismissed since the placement of the 12v outlet varies greatly. Having the plug wired to the 12v on a usb charger wasn't possible due to business case concerns (Using the data plan outside the car). And it would require an area close to the 12v outlet that could hold the device.
5.3 Emphasizing voice more in the design.
Voice interaction is still a long way off for being the only means of navigating content or performing actions. However the promise of voice search is great. Having voice search capabilities states that any content is available including secondary content like podcasts or news reports. Voice interaction also feels high tech. Voice has a big marketing advantage and emphasizing it more in the design could be preferable. Some use cases will be best solved by voice interaction. The remote and voice must work in symbios to cover all use cases.
References
(Bella, M. & Hanington, B., (2012)). Universal Methods of Design, Beverly, MA: Rockport Publishers. pp 204
(Brown, T (2009)) Change by design, HarperBusiness
(Buchanan, Richard.(1992)) Buchanan, Richard. "Wicked Problems in Design Thinking."
Design Issues
8, no. 2 (1992): pp 5-21.
(Dibben, N., Williamson, V.J., (2007)). An exploratory survey of in vehicle music listening.
Psychology of Music, pp.571-589.
(Frith, S., (2002)). Music and everyday life. Critical Quarterly, pp.35-48.
ISO 4040 (2001) Road vehicles — Location of hand controls, indicators and tell-tales.
ISO 4513 (2003) Road Vehicle — Visibility, method for establishment of eyellipse for driver's eye location
ISO 15007-1 (2002): Road vehicles — Measurement of driver's visual behaviour with respect to transport information and control systems — Part 1: Definitions & parameters.
ISO TS 15007-2 (2001): Road vehicles — Measurement of driver visual behaviour with respect to transport information and control systems — Part 2: Equipment and procedures.
ISO 15008 (2003): Road vehicles — ergonomic aspects of transport information and control systems — specification and compliance procedures for in-vehicle visual presentation.
ISO FDIS 16673: Road vehicles — ergonomic aspects of transport information and control systems — Occlusion method to assess visual distraction
(Moreno, D. P.(2014)) Moreno, D. P., Hernandez, A., Yang, M. C., Otto, K. N., Holtta-Otto, K., Linsey, J. S., Wood, K. L., and Linden, A. (2014) “Fundamental Studies in
Design-By-Analogy: A Focus on Domain-Knowledge Experts and Applications to Transactional Design Problems,”
(Salimpoor, et al., 2009)Salimpoor, V.N., Benovoy, M., Longo, G., Cooperstock, J.R., Zatorre, R.J., 2009. The Rewarding Aspects of Music
Listening Are Related to Degree of Emotional Arousal.
(WHO, 2016) Unit 2. Risk factors for road traffic injuries. World health organization, pp.
23-35
(Long S.J 2013) Discussion on Market Research System for Product Development and Design, Chongqing Technology and Business University, China pp. 647-650
(Plattner, H., Meinel, C., & Leifer, L. (2014)). Design Thinking Research Building Innovators (Understanding Innovation). Cham: Springer.
(Saifuzzaman, Haque, Zheng, & Washington. (2015)). Impact of mobile phone use on car-following behaviour of young drivers. Accident Analysis and Prevention,
pp 10-19.
(Cellular-News, (2007)). Countries that ban cell phone use while driving.
http://www.cellular-news.com/car_bans/. Accessed september 24, 2016
Appendix 1
This appendix features alternative solution to the user problematics of using Spotify in the car.
Mobile application UI
The same interaction model shown in fig 8 was used to design an interface for driving inside the existing Spotify mobile application, see fig A. High intent is sorted out just like on the remote. But instead of relying on voice or a smartphone, the driving mode UI relies on the fact that it is already in the high intent environment of the native application. Instead a lower intent interface is present in the now playing view and high intent options are accessed by closing this view.
fig A Driving mode for the application. Main view
Unlike the remote control it only has one shelf of content. This is because the graphical interface adds another element of information. Users can see if the playlist is made by himself or by Spotify. The pivot option works similarly in that the initial menu shows a playlist that is a mixture of several other playlists. By double tapping on a playlist card a pivot menu or shelf is opened. This shelf contains the sub playlists and users can select one of these specifically. The cross button in the lower part of the screen allows the user to return, see fig A.
fig B Driving mode for the application. Third party app integration
fig B Driving mode for the application. Third party app pivot view
The same single shelf layout is used to allow integration with other applications se fig B. This single shelf layout is minimized when not in use and retains a uniformity across the different views of the mode. Same layout is for example used for the locked screen view, see fig C.
fig C. Driving mode for the application. Lock screen view
Remote only
A remote control for spotify could provide a physical UI in the car and relieve some of the pain points. It would however not address the data and battery consumption pain points. It could be a low cost option and reach high distribution.
Bluetooth dongle to car
A device that primarily simplifies connecting to the car and allows the smartphone
application to know that it should launch a driving mode. Main interaction with content would be from the phone. Data could be bundled with dongle and reduce battery and data
consumption from the phone.
