How might we design an adaptive UI of an existing app that provides value to both the company and the end-users?
Presenting LinkedIn: In the Room, an adaptive UI concept.
By collapsing the existing “People you may know” and the “Find Nearby” features. In the room leverages users’ behavior including search history & applied jobs while also incorporating proximity to ultimately provide recommendations of “People you should talk to” in order to get one step closer to your dream job.
My Role | wireframing, competitive analysis
Team | 4 UX designers
Duration | 3 weeks
Have you ever been at a career fair, not knowing which lines were worth the wait?
How might we incorporate adaptive data like context-awareness, time-awareness, behavior patterns, and the alike to collapse the number of steps or refine the output to the users? We explored and analyzed ten existing mobile apps to identify opportunities for an adaptive UI by looking at their current user interaction flow for a specific task and proposed an idea on how we might adapt the current flow.
Currently, users have to first open the app, then go to previous or favorites in order to place a regular order, even if they order the same item every day. This takes 5 clicks. People are creatures of habit, so why doesn’t the Starbucks app encourage and suggest your go-to order?
Idea: Automatically set up a preorder if it’s part of your usual weekday routine. Users receive a popup at the beginning of the app that suggests a favorite order over time.
Currently, Spotify offers several custom playlists to users upon entry to the dashboard based on time of day or other contextual features. However, for those who use Spotify for specific environments (the gym, studying, dance, etc), a lot of search is involved to find the perfect playlist.
Idea: Based on time of day, location data and previous activity, have Spotify suggest your workout playlist if you’re headed to the gym, your study playlist if you’re at the library, and so on.
Currently, Messenger is designed with a bright white interface that makes it difficult for users to interact with it at night. Messenger recently released a dark mode as an easter egg, but only to some of its users. To encourage wellness and healthy sleep habits, the application should adapt to the user’s comfort depending on time of day.
Idea: Use local time, user’s sleep schedule, current light level to help create a Messenger app that slowly adapts into night mode for the user. This will help reduce user screen time in order to promote a healthier lifestyle, or at the very least a healthier sleep schedule.
Currently, Uber and other ride-sharing applications are commonly used when the user is out on the town and needs a ride back. When the user’s battery is low, however, sometimes the process of calling for an Uber can drain what’s left of the battery, leaving the user stranded at 2am in an unfamiliar part of town.
Idea: Have Uber include a low-battery mode that helps users quickly land a ride to their chosen destination. Destination options, number of passengers, and additional ride factors are heavily simplified to allow users to quickly order a ride home while using the minimal amount of battery.
Currently, users check PNC’s mobile app to stay on top of their checking and savings accounts. To maintain a good savings balance, users have to manually save certain amounts from their checking or deposit straight to savings. What if there were a way for the savings account to know when to withdraw and save certain amounts?
Idea: Based on current income flow and spending habits, have PNC autosave to savings a variable amount that helps users beef up their savings without completely depleting their checking account.
Coordinating concert attendance can be a logistical nightmare for some. Currently, the Ticketmaster app does a great job of getting the user through the checkout process, but doesn’t help much with coordinating who you’re going with.
Idea: Our redesign of Ticketmaster would involve adding a “suggest a friend” feature, which helps users quickly tack on extra friend tickets at time of purchase. Often times people like to go to concerts with the same people so this would help make the ticket-buying process easier for groups.
AirBnB is currently great at coordinating trips easily and painlessly for travelers. You can pay for lodging, arrange for experiences, and add or remove dates at the user’s convenience. If you travel with the same people for most of your trips however, the application currently doesn’t auto-suggest or recommend additional lodging for your friends or colleagues.
Idea: With our update to AirBnb, we would have an auto-recommend feature that suggests people to add to your booking. People are creatures of habit and like traveling with their go-to travel buddies - why not let AirBnB help with the logistics?
Venmo is a great application for sending and requesting payments from your friends or roommates. More recently, it has turned into an easy form of bill collection for housemates. While the app is convenient, roomies find themselves sending the same bill requests every month and have to go through the process of typing things in every time. Can we simplify this interaction?
Idea: Using transaction data, Venmo would send the user a reminder to either make a payment to a given user or request money from a user every month as part of their billing. When a user goes to fill out the bill request, it can autosuggest people or payment names given the user’s history.
Currently, Transit doesn’t offer any information or features for parking. As a transportation app, there seems to be room for opportunity to expand on its offerings.
Idea: Based on context-awareness, have Transit show predictive data on times when parking spots may be open. For example, alert users “Spots usually clear up around 6:15pm, want to come back then?”
Currently, LinkedIn shows you a list of “People you may know,” based on shared education, work experience, social media, etc. But are you really using LinkedIn to connect with people you already know? There is also a “Find Nearby” feature that is intended to help you find people near you, but it has low discoverability and usability.
Idea: Adapt proximity, previous search history, and applied positions to show you a more personalized list of “People you should talk to” to help advance your career goals.
Narrowing the scope
Among the ten ideas, we began narrowing down the scope by evaluating various factors including the size of value (users and stakeholders), ease of ML development, and the risk of errors, to rank the ideas on a spreadsheet.
The top four apps with the highest estimated values were AirBNB, Venmo, Transit, and LinkedIn.
For AirBnB, we propose a personalized feature that automatically recommends adding your travel partner to a booking.
The business travel market is worth 1 trillion dollars globally. Currently, AirBnB’s market is 15% business travelers and they hope to increase it to 30% by 2020.
The service provider would automatically know the other traveler and be able to push recommendations automatically to their AirBnB account or email.
The development based on past user-experience should be easy as it pulls from the current database.
Since it is an auto recommendation feature, risk is relatively low but value is high because it takes the thinking out of the planning.
For Linkedln, we propose a personalize and context-aware experience that is equivalent to a digital handshake. In this automation the app would connect people based on proximity and interest when you are networking an event in your area.
Linkedln currently has over 500 million users on its platform with 40% of them actively networking in any given day.
This would help Linkedln outpace Facebook for event promotion and given Linkedln more active users which would increase ad revenue.
Proximity data matched with interest in a field should be relatively easy to map off the Linkedln database.
There are risk of casting too wide a net, but people tend to ignore things that are not relevant and are not bothered by solicitations of this kind.
Using data from email and texts, Venmo would send the user a reminder to either make a payment to a given user or request money from a user.
Venmo was the most popular P2P payment app in 2017 with over 20 million users.
It will increase referrals and add users to the platform. It will also offer something competitors do not.
It will need to integrate with email and text and be adaptive to the language the user uses to signal a financial transaction, so it will need to potentially learn words.
Risk of error could be relatively high misreading payee and payers, as well as mismatching names to users.
For GoMobile PGH, we propose a feature that would give users predictive data for planning their outings around parking at certain times of the day.
Parking around Pittsburgh isn’t easy. Users rely on the app currently to pay for parking. This added feature would help people know when to come into downtown to do business and find spots with greater ease.
The value for the service would be more revenue since more people would find spots and rely less on private parking experiences or parking illegally.
This would take time to adopt since it would need to track data at a very sensitive scale and also be adaptive to holidays, weekends, and special events.
Error risk is relatively high as there are many factors that determine whether spots are available.
Final Choice: LinkedIn
LinkedIn was tied with AirBnB for first in terms of estimated value, and we also thought this adaptation was something that we wished existed today. Because it felt so relevant to university students, we felt the most strongly about it and was excited to make help users find the right connections to advance their career.
LinkedIn: In The Room
RECAP OF CURRENT INTERACTION
LinkedIn recommends based on commonalities between you and other LinkedIn members
School, work, other experiences
Contacts you’ve imported from email and mobile address books
Currently hosts a “Find Nearby” feature that connects with bluetooth, but generally goes underused
INITAL ROUGH SKETCHES
From the current screens, we took notes on where we could collapse information, focusing on increasing discoverability and providing adapted, additional information of “People you should talk to” based on a user’s past history of search, applied jobs, and work experience.
We used this as a basis to create our first iteration of wireframes for the proposed interaction presented during our interim critique.
Wireframes for Interim Critique
Improve discoverability of feature, making users aware of this feature → push notifications, pop-up?
Proactive vs reactive
Consider making the next-steps adaptive (e.g. “you should send your resume to this new connection”)
Most job fairs have lists of employers beforehand (like on Handshake)-- maybe this would be helpful for users to leverage when figuring out who to talk to
Think about specific use case scenario— Consider situations where people use this while waiting in line during job fairs (waiting process)
Consider prompting the user who hasn’t input recent data to confirm what types of jobs they are looking for to avoid erroneous recommendations (ex: if users are switching career paths)
Option to report error and better calibrate suggestions
Bottom line: there would absolutely be value for this feature, we just need to make it visible and flexible to user needs
Adaptive Interaction Flows
In the new adaptive interactive flow, a user clicks on “In the Room” and is face-to-face with six potential links. In the old design, users click on “Nearby” feature to find other users in the vicinity and a list of people pops up on the screen.
Our design is adaptive in that it prioritizes who is shown based on recent user’s activity to generate better matches between people. Furthermore, the user has the option to drill down for more information on a link by clicking “More info.”
Our design is primarily reactive, tailoring the recommendations based on user activity. There is also a proactive element whereby the user is recommended to talk to certain people who may be nearby, even if they weren’t initially part of their suggested.
In addition, our design takes an intelligent approach over an if-then approach as its logic follows a more complex line of thinking than that of a linear decision tree.
To address error recovery, if the person is not a good match, the user can X out the person and the system learns from or adapts to user input and tries to generate a better pick in the next round. If the match is right, the user can connect, and the data is used for future matches.
With our design, we intend to leverage the benefits of the People You May Know feature and the Find Nearby feature to help autosuggest people nearby to you.
The job hunt can be exhausting and time-consuming — with our new feature In the Room, you can efficiently identify people worth talking to and immediately begin the networking process.
You only have one opportunity to make a first impression:
let In the Room get you there quicker.
Limitations and next steps
How would Bluetooth work at long distances? In large career fairs, how would the accuracy change?
Phones typically have Bluetooth range of up to 10 meters, so for these cases, rather than mapping location based on Bluetooth, we may consider combining this with GPS location
We think we could cross-reference invitation / RSVP lists with the In the Room feature, possibly expanding it to include spatial views beyond just in the room.
Consider all the parties involved: job hunters, recruiters, event coordinators. How to motivate all parties to turn on this feature?
Our team suggested that the event organizers would encourage use of the feature across all parties.
Would recruiters want users to get tips on how to talk to employees - would this be borderline creepy?
We plan to aggregate these tips from the user’s profile, which would already be publically available.
Job hunting is a messy problem space and this application would be useful from a suggestion standpoint
There’s value in a In the Room feature but the hardware / implementation side may be beyond the scope of this project.
Our team worked great together and this was a fun project! We wish this feature existed for future job hunts.