Making NCR's Kiosks accessible for wheelchair users
Design guidelines for the next update of the NCR Kiosks
UX Research
Benton Humphreys, Varnit Jain, Kevin Key, Yunfei Wang
Qualtrics, Mural, Figma, AdobeXD
Timeline (Aug '19 - Dec '19)

6 weeks
Conducted observations, interviews, surveys, task analysis to understand pain points
4 weeks
Analysed research results using Affinity Mapping and addition concept testing
4 weeks
Designed a new accessibility mode for wheelchair users
Usability Testing
2 weeks
Conducted usability tests using Hi-Fi prototypes
The NCR Corporation, previously known as National Cash Register, is a leading kiosk manufacturer whose products include different kinds of self service-kiosks, such as the XK32, the SK47, the SS75, and the SelfServ 90.
However, these Kiosk systems are not usable by a variety of users due to both hardware and software accessibility issues.
At the same time, NCR has not been oblivious of accessibility and has implemented a uNav interaction panel to its food-ordering kiosk model- the XK32. It includes an audio jack (voice instructions) and a 5-button keypad.
1. Literature Review
Since neither of us had any prior experience working with Kiosks, the first step in our research process was to go through existing literature. While there is an ever-growing field of research to suggest best practices for accessibility within digital design, we found out that there seems to be a lesser focus on integrating accessibility into kiosks. We sourced, reviewed, and synthesized several research papers that either touch on kiosk accessibility or associated applicable field. Some of the key takeaways were:
2. Task Analysis
To better understand how users interact with the current system, we decided to observe users interacting with the kiosk at peak usage - the lunch hour. Our goal with observation was mostly exploratory, but with some research objectives. We intended to build a task heirarchy of the existing system and determine if users experienced any usability issues with the ordering process. Additionally, we wanted to know how a novice user interacts with the system compared to a user that has been familiar with the system for some time to determine learnability.
3. Semi-Structure Interview
Since we were dealing with a vulnerable population, we had access to a select people for in-person studies. We wanted to use surveys to gather information about user needs and pain points. However, before creating a survey without much domain knowledge, we decided to conduct a semi-structured interview as our first research method to understand the problem space better. We were able to contact someone who advocates rights for people with physical disabilities across Atlanta. Our main information goal from this research method was to weed out obvious questions and get in-depth knowledge about the community's perception of Kiosks. We also wanted to know more about the technologies they found helpful in public spaces.
4. Survey
We chose an online survey as a research method to collect data primarily because of its ability to collect data from a large set of users in a short time. We wanted to learn what users with disabilities like and dislike about a point of sale kiosk. We wanted to know how users’ disabilities hinder specific interactions with a kiosk. Based on our previous research methods, another goal for us was to learn about common frustrations at each stage of the process of completing checkout. We also wanted to determine if certain users prefer tactile interaction versus touch screen and for completing what tasks.
1. Personas and Empathy Maps
After going through our data, we felt the need to develop personas to refer back to understand the user's needs and motivations. The personality types, preferences, and frustrations were all informed insights from our interview and survey results, which were generalized to form a cohesive persona. We created empathy maps for both our personas to further elaborate our understanding.
3. Affinity Map
After collecting all the data from the observations, the semi-structured interview, and the survey results, we aggregated it in the form of an affinity map. We were able to distill the information user needs and design opportunities. We were also able to find recurring feedback that could be grouped into common themes.
From the task analysis, observation study, interview, and survey, we had a general idea of where users were likely to experience problems. A few of the most straight-forward problems, and likely the most easily remedied, were:

We started designing our solution by brainstorming different design concepts with users’ limitations, needs, and design implications we analyzed from previous research. We did this through an informed brainstorm session where we generated ideas relevant to user needs and pain points.
We explored kiosk features across a range of design considerations, ranging from Universal Design to User-Group Specific Design to User-Specific Design.
1. Design Concepts
We finalized various features and ideas from brainstorm session into three abstract design concepts that covers different ranges of potential users of kiosks mentioned above. Sketches and descriptions were created for each concepts for assessment by experts and users for feedback.
After fully defining our concepts and developing sketches to represent them, we began developing our first feedback session. At this point, we wanted to present our concepts at a relatively low fidelity to allow for feedback with original and creative ideas for improvement. The goal of this feedback session was to provide usability experts with a broad understanding of the three concepts we created to obtain feedback on the idea as a whole, and for certain features within the idea. For each concept we wanted to understand the following:

After speaking with usability experts from NCR, we decided to pursue concept 2 (uNav Compatibility Mode) as a basis, and incorporate features from other concepts. From a user’s perspective, concept 2 most significantly addresses the accessibility issues presented during the initial observation. From a stakeholder’s perspective, the proposed concept is a companion to the already existing uNav control and will be considered as an evolution of the existing system.
2. Wireframes
Following the feedback from the concepts, we chose to take a deeper look at the uNav compatibility concept by generating mid-fidelity wireframes. These wireframes were constructed with the intent to inform a user of how they might navigate through the ordering process from a high level, what the information structure would look like (e.g. information hierarchy and element size and placement), and how certain features might influence their experience (e.g. tab indexing and highlighting selected elements). Here are a few wireframes out of the many that were used for testing.
We tested the above wireframes with our primary stakeholder, NCR, in the form of an open focus group discussion. We took notes from this session and captured valuable feedback and related suggestions. We can make certain improvements to the current design:

Our group analyzed the feedback and incorporated the results in the high fidelity design.

3. High Fidelity Design
4. Task Based Interactive Prototypes
Usability Testing
1. Expert Based Testing
We chose to do a heuristic evaluation walkthrough to evaluate our prototype and weed out the common mistakes in the interface. We used three cohorts from the HCI program as usability experts and asked them to perform certain tasks using our prototype. Using this approach, we were able to capture the overall experience of a new user as well as point out accessibility and usability issues on individual screens and interactions. We had the following information goals:

Summary of Qualitative Feedback:

Quantitative Feedback:
2. User Based Testing
We chose to do moderated task-based usability testing with users to determine how well the interface is integrating with the uNav keypad as an accessibility feature for input during the food ordering and checking out process. Also, we want to collect data on users’ performance for examining the logic flow of tab indexing and get feedback on content, functionality, and visual layout. We conducted 9 in-person single-participant sessions. We had the following information goals:

Summary of Qualitative Feedback:

Quantitative Feedback:
3. Results
Apart from the issues mentioned above, we found that the level of usability and visual representation was satisfactory and the task flow is easy to follow. We found participant reception to our prototype to be positive, and most issues could be solved with some UI changes and building a detailed information heirarchy. NCR is satisfied with the overall design recommendations and plans to keep them in mind while moving forward in the developement of the XK32 Kiosk.
If you would like to know more about this project, reach out to me at [] or [LinkedIn] .