Background
Enhance-X is a downloadable editor that offers unique and experimental features to Samsung users. It functions primarily as a testbed for testing these features in the market.
Feature proposals
Enhance-X Management
Samsung ORCs
Filtration
Positive feedback
Negative feedback
Opportunity area
The desired flow
Different overseas research centres (ORCs) would propose editing features that they have developed to Enhance-X management. Based on stakeholder decisions, we would implement the features within the app. After a certain period of usage, the feedback for the feature is analysed, and if it is positive enough, the feature is moved to the native editor. More commonly, the feedback highlights issues with the features, which is then relayed to the original developers at the respective ORC, for them to fix.
Problem statement
Design a comprehensive feedback collection framework which gathers actionable data to enable feature improvement & migration
UX Process
Initial Research
We started the project with broad research understanding the kinds of feedback collected across different web and mobile apps (particularly image editors), and the ways in which it is collected. We looked in-depth at the variety of feedback collection mechanisms, and the advantages of one over the other. We also looked at Samsung’s own mechanism for feedback collection, a “Contact us” present in the Settings of most Samsung apps, the Samsung Members community for active discussions, and also what kind of back-end data is collected, and how it is analysed.
In parallel, we worked with stakeholders to understand the exact business requirements from the feedback, understanding technical limitations from app development, as well as storage and legal issues at the organisational end. We also worked with the PM team to understand the storage and anaylsis of currently collected data, to identify opportunities for improvement.
Explicit feedback in consumer facing apps
“Contact Us” in Samsung apps
Samsung Members Community pages
In-app mechanisms are preferred by editors for immediate feedback, and community blogs for reflective feedback
Specific feedback is best collected through targeted questions, whether in-app or in surveys
Closure of the feedback loop can make customers feel heard, and encourage furhter participation
Implicit data helps collect information of user behaviour that the users themselves may not be able to provide
1
2
3
4
Approach
Implicitly collected data
Redesigning back-end data collection & analysis to meet product needs, within the scope of Samsung’s internal tools
In-app explicit feedback
Introducing a quick, unobtrusive way to get raw, immediate feedback from users on individual features
Based on the research and stakeholder discussions, we finalised on a two-part approach to the feedback collection:
From the breadth of research, we ultimately boiled down to four key takeaways:
Part 1/2
Implicitly Collected Data
We defined implicit feedback as insights inferred from usage data of the app - as opposed to explicit feedback, where the app would outright ask the user for feedback like ratings. From our research, we had identified certain kinds of implicit data collection and analysis methods that could be introduced within the app. Further, in our study of the current data collection in Enhance-X, we noticed significant gaps where far more meaningful data could be collected - certain instances of these are shown below.
Initially, logging had been very simply defined, where the app logged taps on every button, ie, which button was pressed and how many times. Basic user demographics and phone model details were also captured. We identified the opportunity to log far more detail with the taps, detail which would be essential later to understand the success/failure of a feature.
Post collection as well, certain issues were identified in the interpretation of data. This led to further improvements in how data points were defined for logging - in a way that made later interpretation easier and more efficient. Overall, a number of proposals were implemented. Examples of some of these are shown below.
How many times “Undo” was pressed
How many times a standard “Feature could no be applied” error message was shown
How many images were “Saved” vs “Discarded”
Which feature was removed with Undo?
Which features are commonly removed after applying?
Which feature caused the error?
Why was it not applied? What was the issue?
What impact did this have on the user journey?
Which features combinations are typically Saved/Discarded?
Which features generally lead to a Discard?
Example 1:
Logging details of what is Saved
Example 2:
Improving the ‘Magic’ algorithm
User 2
Save
Brighten +10
Contrast +10
Bg blur +30
+
+
The data saved in the array (a, b, c, d) can later be computed to find data for individual features as well as combinations of features. It helps identify not just which features are applied, but which are saved. Later comparisons with other data points also helps understand the Save to Apply ratio of each feature, the combinations of features which work well together, etc.
Magic is an auto-correct feature which applies an algorithm consisting of multiple values of multiple features across an input image. Magic needs to cater to what users believe looks best. If without the auto-correct, user behaviour shows that they tend to frequently apply two features together, over a large dataset, understanding correlation between these features can help improve the Magic algorithm.
User 1
Save
Contrast +25
Fix grain +20
Fix blur +20
Brighten +10
+
+
+
User 1
Save
Portrait +20
Sharpen +10
Contrast +20
Brighten +10
+
+
+
Part 2/2
In-app Explicit Feedback
Upon request from senior stakeholders, as well as a need for certain qualitative feedback, we worked on designing a method for explicitly asking for feedback. To start, we collated the different kinds of explicit feedback mechanisms from our desk research. These included, but weren’t limited to feedback icons, options within Settings/More menus, pop-ups, surveys, and community blogs.
Weighing the pros and cons of these different methods, in parallel with product requirements, we decided on a binary Like/Dislike, and a more specific survey/questionnaire on feature issues. The questionnaire, we realised, was necessary to get more specialised, qualitative feedback on how a feature could be improved.
With the kind of mechanism decided, we worked on the touchpoint of the user journey at which we could introduce the mechanism, as well as deciding the frequency of asking for feedback. While we wanted as much feedback as possible, it was essential to avoid frustrating the user with frequent prompts, to design a method that is unobtrusive and does not hamper the user journey, and to always provide an easy and clear exit from the feedback screen/form.
Upon evaluation and refinement, the final mechanism introduced was of the “Thumbs up/Thumbs down” at each feature’s second-depth, along with a “What went wrong?” pop-up in case of a Thumbs down reaction. This would help gather actionable inputs on how the feature could be improved. The options under “What went wrong?” were also customisable per feature, depending on the kind of feedback the feature owners preferred. It also offered an input textbox for users to enter any issue they faced, that was not provided in the options.
Feedback icons in Google Translate and Netflix
Feedback options within menus in Splitwise and Google
Feedback pop-ups in Photos, StarTracker, and Rapido
Possible touchpoints:
Last photo you edited,
rate your experience!
After 3 times of
cancelling a feature, ask
“what went wrong?”
When exiting the app,
ask for details
Conclusion
A version of Enhance-X was released in the Galaxy App Store with both of these mechanisms implemented, post rigorous testing and checks
The implicit feedback analysis is highly successful and being used to define success criteria for features
Thumbs up/thumbs down currently has infrequent usage - improvements in this are in the pipeline
Reception
Currently being collected:
What isn’t being collected:
User 1
Save
Portrait
Fix blur
Sharpen
Brighten
+
+
+
Existing logging
Feature ID a
Feature ID b
Feature ID c
Feature ID d
Event ID X
Feature ID a
Feature ID b
Feature ID c
Feature ID d
Event ID X (a, b, c, d)
Improved logging
This implementation was highly appreciated by the then MD of SRIB, Dipesh Shah, for its innovation and the potential it created for deeper data analysis
Brightness
Contrast
Upon opening Enhance-X, a user can see different available features
In the 2nd depth, they see the applied feature, along with the feedback icons of thumbs up/thumbs down
Pressing “Thumbs up” leads to a small animation acknowledging the feedback
“Thumbs down” triggers a pop-up asking the user why they disliked the feature
Feedback Collection
in Galaxy Enhance-X
Building the feedback collection mechanisms for Samsung’s downloadable AI Editor
In collaboration with PM Team and our Developers
ROLE
TEAM
TIMELINE
Design lead
UX
4 weeks
Oops.
This page isn't yet optimised for this viewing ratio. Check out my site on a laptop or PC to see! Else, you can be redirected to a PDF version of the project, through the button below:
