Roles & Contributions

Concept Statement (Synopsis)

LuxCopia is an interactive supply room assistant and predictive inventory management system designed to eliminate the logistical frustration of "hunting and gathering" medical supplies and mitigate chronic administrative overload for frontline clinicians. The system translates complex patient orders into an intuitive physical map by syncing directly with electronic health records (EPIC) to generate procedure-specific supply lists, instantly illuminating the corresponding physical bins with personalized, color-coded light beacons and auditory cues. 

Unlike traditional, static supply closets that rely on tedious manual barcode scanning or intuition-based searching, the system allows nurses to interact with localized tablets or mobile devices using multi-modal search (voice, text, image) to locate items, while integrated weight scales and spring-loaded trackers automatically update the hospital's inventory database the moment an item is removed. 

By substituting manual data entry and blind searching with automated tracking and intuitive physical wayfinding, LuxCopia empowers clinical staff to reclaim valuable shift time for direct patient care while shifting hospital supply management from reactive to proactively optimized.

Tailoring of Scope

For this project, we quickly realized we would have to simplify and make a prototype testing some of our core features we wanted to bring to our users. For this reason, we chose 3 specific tasks across our two work roles to conduct evaluations on. As you will read later on, while we had a small physical prototype representing the entire supply room with 6 cardboard boxes, we quickly saw in our first two user interviews that it was not an accurate representation of the vast messiness of a regular supply room. This initial scale failed to test the "findability" issues inherent in a cluttered clinical environment. Quickly and creatively, we had to pivot in Clem Library looking to see how we could represent a vast amount of boxes with similar Wizard of Oz lighting functionality. As you’ll read in the report, we were able to transform a TV using Google Sheets, some duct tape, and cardboard to represent a larger room with the same system we had going. This pivot was essential to simulate "visual noise," allowing us to see if our lighting system could actually cut through the chaos of a high-density storage area.

The two work roles we focused on with this was a frontline clinician quickly finds supplies for a patient or one-off need, and a supply manager identifying urgent restock issues and updating item status. This important choice lets us test the parts of the design most directly tied to the project goals: speed, clarity, convenience, and trust. It also kept the evaluation focused on realistic benchmark tasks instead of trying to simulate the entire hospital infrastructure. We justified this "middle-ground" prototype, using a digital screen to represent physical space, because it provided the necessary fidelity to test user reaction times and mental models without the prohibitive cost of a full-scale physical build. However, this approach absolutely had limits since the prototype does not include a real badge scanner, real EPIC integration, actual voice recognition, or real inventory sensors. Additionally, our evaluation was limited by the environment and participants since we were testing in a busy library with premed students unable to perfectly replicate the high-stress, interrupt-heavy environment of a chaotic hospital ward where clinicians are often multi-tasking. Instead, those behaviors are simulated through the interactive, Wizard-Of-Oz prototype setup requiring the active monitoring of our entire team to keep the system up and running. The reliance on human teammates being the operators or “wizards” introduced a slight latency that would not exist in a fully automated system, which very very minimally impacted our data on speed and trust. For these reasons, our evaluation is best understood as a usability study of the interface and workflow rather than a final performance validation.

UX Targets

UX Goals:

UX Goal 1 Rationale (Quickly get a full patient specific supply set):

This goal matters because one of LuxCopia’s main purposes is helping nurses quickly collect the supplies needed for a specific patient. In the current workflow, a nurse may need to search through the supply room, remember where items are located, read several labels, or ask another staff member for help. This can slow down patient care and add stress during an already busy workflow.

We used 4 minutes as the baseline because it represents a reasonable estimate for manually finding a full set of supplies without LuxCopia’s guidance. We also checked with our client Jackson if this baseline made sense, which he then affirmed. We set the target at 1 minute 30 seconds because the system should make the task clearly faster than manual searching, with 1.5 minutes feeling like a realistic goal. This target shows that LuxCopia is not just adding another screen, but actually reducing the time and effort needed to prepare for patient care.

We also focused on nurses who are new to LuxCopia because the system needs to be easy to use without much training. If a new nurse can quickly understand the patient supply list and follow the highlighted bins, then the interface is doing its job.

UX Goal 2 Rationale (Quickly find a single item from the supply room):

Clinicians do not always need a full supply set. Sometimes they only need one item, such as gauze, and they need to find it quickly. This task tests whether LuxCopia can support that common one-off supply search.

We used 1 minute as the baseline because manually finding a single item may still involve scanning shelves, reading bin labels, or relying on memory. We set the target at 30 seconds because we wanted our system to make finding an item feel almost immediate. Since this kind of task can happen many times during a shift, even small delays can become frustrating over time.

The goal is for LuxCopia to make single-item search feel simple: search for the item, see where it is, and go directly to the correct bin. If the system takes too many steps, users may ignore it and go back to searching manually.

UX Goal 3 Rationale (Nurse Trust):

This goal matters because speed is not enough by itself. Nurses also need to feel that LuxCopia is easy to use, trustworthy, and convenient. If they do not trust the system, they may hesitate, double-check everything manually, or create workarounds that avoid using LuxCopia.

We used baseline ratings of 4/10 for ease, 5/10 for trust, and 5/10 for convenience because the current supply workflow is usable, but still frustrating for many clinicians. Supplies can usually be found, but the process often depends on memory, searching, or help from other people. From our previous interviews, these ratings felt like a good initial estimate of how nurses feel about the current situation regarding supply rooms.

We set the targets at 7/10 for ease and 8/10 for trust and 8/10 for convenience because LuxCopia does not need to feel perfect during prototype testing, but it does need to feel clearly better than the current process. A 7/10 for ease means users can understand the system without major confusion. An 8/10 for trust and convenience means users see the system is seen as reliable enough to actually use during clinical work. This is important because LuxCopia only works if nurses are willing to follow its guidance instead of falling back on old habits.

UX Goal 4 Rationale (Complete the highest need restock/update workflow correctly):

This goal matters because LuxCopia is designed to support supply managers as well as clinicians. Supply managers need to identify low-stock items, locate the correct bins, restock them, and keep inventory information up to date. If that process is slow or inaccurate, the supply room would still be frustrating to use because items might be missing or outdated.

We used 4 minutes as the baseline because restocking items usually takes the same amount of time as finding those items, and we modeled this task to be similar to Benchmark Task 1 in UX Goal 1. For the same reason, we set the target at 1 minute 30 seconds.

This target matters because restocking is a repeated task. Saving time on one restock may seem small, but it adds up across many bins in one shift. For example, if a supply manager saves about 2.5 minutes per restock across 40 restocks, that is about 100 minutes saved.

UX Goal 5 Rationale (Supply Manager Trust):

Supply managers need to trust that LuxCopia matches the real state of the supply room. If they do not trust the system, they may still check shelves manually, keep separate notes, or rely on their own memory. That would reduce the value of LuxCopia because the system is supposed to make inventory work more accurate and less reactive.

We used baseline ratings of 4/10 for ease, 5/10 for trust, and 5/10 for convenience because these ratings seem to be the average impression from the supply staff we talked to during our interviews. This identifies the process works, but it is not as efficient or dependable as it could be.

We set the targets at 7/10 for ease and 8/10 for trust and 8/10 for convenience for the same reasons stated in the UX Goal 3 Rationale. Because supply managers use this workflow repeatedly, small doubts or annoyances can become major problems when the system is used many times during a shift. For LuxCopia to be useful, supply managers need to feel that checking the system is faster and more reliable than walking around and checking bins manually.

We based this selection and analysis process off of Chapter 10 in the Textbook:

1) Selection of Benchmark Tasks (BTs)

• We selected our BTs based on a mix of Criticality vs. Frequency since BT1 & BT3 (Patient Order/Restocking) are high-criticality tasks because if they fail, the primary function of the hospital supply chain breaks down. And then for the one-off BT2, it was a high-frequency task and represents the daily tasks that if we could optimize, could save some of the highest time savings for clinical staff.

2) Establishing Baseline Levels

• Baselines were established through our user interviews with the nursing staff and supply managers for the contextual inquiry phase, the last phase when we pivoted, and client feedback.

3) Setting Target Levels

• Our targets were set using criteria like substantial time savings, qualitative measurement improvements outside of the quantitative metrics we looked at, and also feasibility. Obviously it would not be the same as a nurse or supply manager walking across the room with the cart, but we tried to adjust for those here with our own dry runs. Luckily, even after part of our prototype changed, it was still pretty centralized together in a close space and we felt we did not need to necessarily update these timings because of that

4) Team Test Run Verification

• We then had to verify that the prototype fully supports these BTs. During internal testing, we confirmed all of these features and even made changes to address Figma routing issues and bugs with the supply room prototype.

Design Refinement

Client feedback:

After looking at our wireframes, our client Jackson remarked that our overall direction is strong, but raised skepticism about how a clinician would use the system under pressure. In a real supply room, a nurse may be tired, stressed out, or busy interacting with colleagues. During these situations, taking time to navigate the Luxcopia system may add unneeded mental strain for the user. Because of these comments, we decided to focus our refinement less on adding new features and more on making each screen easier and effortless to use. We adjusted our wireframe design to emphasize and direct the user's next steps to encourage a smoother workflow. For example, on the patient order screen, the most important information is not a long description of the items, but whether each item is available, which bins are lighting up, and what the user should do next. So we focused on improving the visibility of those attributes rather than trying to fit many niche features onto a single screen.

A second piece of client feedback was that the connection between the tablet and the physical supply room should be more obvious. In our system, selecting a patient’s order or searching for an item causes the correct bins to light up, but a first-time user might not immediately understand that the tablet is controlling what happens in the room. We realized this connection should not be something users have to figure out on their own, especially when they are in a rush. To address this, we refined the design so the tablet uses the same color cues as the physical bin lights, making it clear that the color shown on the tablet is linked to the bins the user will interact with. We also made the location information more direct: instead of only saying that an item is available, the screen points the user toward the lit bins or any alternate locations. These additions made LuxCopia feel less like another system the clinician has to operate and more like a guide that helps users move through the supply room with less confusion and fewer interruptions.

Instructor/TA input:

The Phase 3 feedback pushed us to make the workflow more complete and easier to follow in a real shift. A major point was that our storyboard did not fully show how the system fit into the user’s day, and that some key tasks were still unclear, such as how the system recognizes the user, when the room lights turn on, when they turn off, and how shared lookup would work. In response, we modified the wireframes to make the badge tap emphasize its role as the clear entry into the personalized workflow. We also made it clear on our screens that the room only illuminates after the user selects a patient order or a one-off item. When designing our prototype, we also made clear through Wizard-Of-Oz demonstration when the lights turn off to signal the end of a task, as well as how shared lookups will work. These considerations directly respond to the concerns raised by the instructor and TA as we worked out the transition from the tablet interactions to the physical feel of the system seamless.

The instruction team also pointed out that our wireframes felt unresponsive. It was sometimes unclear what parts of the screen were clickable, how users would know when an item had been collected, and how inventory counts changed when supplies were taken out or restocked. To address these issues, modified our wireframes to better visually reflect the changes that should occur during each stage of our workflow. We also carried this feedback into our physical prototype. For example, we thought more carefully about how the supply bins should be labeled and how users would connect the labels to the tablet. We also simplified the digital navigation because our earlier version tried to show too many possible paths at once, which made the workflow harder to follow. Overall, these refinements made the design clearer, more responsive, and more realistic for the work conditions LuxCopia is meant to support.

Prototyping

Scaled Prototype (Old)

To implement the physical portion, we, naturally, built a physical prototype. Using standard craft supplies, a laser cutter, and an Arduino kit with LEDs, we built a scaled version of the supply room with six bins, each having an LED attached to it to simulate the illuminating. Additionally, we had a scaled 3D printed figure to represent the user in the scaled space. We decided on having a scaled supply room over just having bins because we wanted to simulate the time and movement of being in the room and interacting with the system. Since the physical portion of the system is so vital to the functioning of our system as the sole indicator of where items are, it was extremely important to make sure this portion of the prototype accurately simulated our envisioned design. As you will find out in the evaluation section, there were significant drawbacks to this design.

This design utilized an Arduino by having each LED macroed to a specific key on one of the team members computers. When a user picked an item on the digital system, stocked an item, or took one out, this team member utilized the Wizard of Oz concept to illuminate/deilluminate relevant bins, giving feedback to the user. We chose an arduino to leverage our team members previous experience and knowledge of the system. Additionally, due to the availability of free parts,  we were also contrained by the types of components we could use; a large reason as to why we strove away from systems like Raspberry Pi. As such, we settled for a simple arduino and LED system controlled by the computer running the microcontroller.

Each bin we stocked with wads of paper, representative of the various supply items. Rather than having the user pretend, we wanted to capture the physical and cognitive process that takes place when grabbing a set number of  items from a large container: grabbing too many and putting the rest back, grabbing too little and having to grab more, and counting as they grab. This time and a possibility of a critical incident could be lost without this feature. 

Furthermore, for both roles, we had a basket mounted to the front of the printed figure to simulate their inventory.  Similar to the above, we wanted to capture the natural time and process of gathering items, walking around the room, possibly dropping items, etc. 

If you would like to view the code of the arduino, please click here

Below are pictures of the system:

TV Prototype (New)

For the updated "TV" prototype, we utilized the TVs in the study rooms in the University libraries to simulate a shelf of 40 bins. This idea came serendipitously from the location where we decided to pivot ideas. Having been in Clemmons Library floor 2 (familiarly known as "Clem 2") at the time of the prototype shift, we were lucky enough to be there when study rooms were open to discuss new prototype ideas. Happenstance, our room had a TV and this is were we realized it's potential.  Using the TV, we could accurately capture the cognitive overload that comes with looking at so many bins, unlike the original protoype. Additionally, we also considered the entire study room part of the prototype. Having the study room  be about the same size as a real supply room, we could actually simulate the physical movement of the user.  To simulate the bins illuminating, one of our team members would share their screen to the TV through an HDMI cord. The user would interact with the TV version and when items were selected, picked up, or stocked, the team member would change to different pages of the google sheets to simulate bins being illuminated and de-illuminated and another team member would edit the figma screens to reflect those actions, utilizing the Wizard of Oz prototyping for the "physical" portion. Furthermore, we decided on using google sheets due to the teams familiarity with the program and the need for a quick prototype due to our sudden pivot late in the phase. 

If you would like to see the different screens of the bins, more than what's displayed above, please click this link. 

Here are a few figures to better explain this prototype's set up:

Evaluation

The goal of this evaluation was to understand how well the LuxCopia system supports efficient, accurate, and low-stress workflows in supply rooms for frontline clinicians and supply managers. We hoped to learn whether users could quickly locate and retrieve supplies with LuxCopia, how accurately they interpret what the system shows about inventory, and generally where users experience confusion, hesitation, or breakdowns in the interaction flow. 

The UX target table (that was seen earlier) directly informed the design of our empirical evaluation by organizing each UX goal in specific benchmark tasks and measurable performance outcomes. Through observation, timing, and post-task questionnaires, we evaluated system performance using task completion time and subjective usability ratings. The usability test followed a structured protocol combining task-based evaluation, think-aloud feedback, and controlled facilitator intervention to ensure consistent data collection across participants.

Our evaluation was conducted in-person on the second floor of Clemons Library in a conference room. We recruited 12 participants, all aged approximately 18-23, with most having little to no prior experience with inventory systems or the hospital setting. 7 of the participants acted as our frontline clinician users and the other 5 acted as our supply manager users. 

We initially began our evaluation using the scaled physical prototype and tested it with two participants. However, we soon realized that this prototype had many issues. We found that the user relied on the labels on the supply boxes rather than the intended LuxCopia system. The participants also mentioned that they hadn't noticed the lights we had implemented, which were important for our intended design. When the labels were removed, the user then became confused about the contents of the bins, creating usability breakdowns. A fully physical implementation with more space, boxes, and lights would have required a significantly larger and more complex setup, which was not feasible within the time constraints of the project. As a result, we switched to the TV prototype (as mentioned in the prototypes section). This shift allowed us to observe participants using the LuxCopia system as intended during the tasks, resulting in more meaningful and valid evaluation data. 

The participant recruitment was opportunistic, as group members went throughout the library to find individuals willing to take part in the study. We decided to use student participants due to the difficulty of finding nurses, nursing students, or supply managers available in the time frame that we needed. Since we switched our prototype, we relied on having a TV and a decent amount of set up time, which made it harder to schedule meeting times, preferring opportunistic interviews in order to get a sufficient amount of data points. Once someone agreed to be a participant, they were given a scripted introduction outside the room that explained their assigned role and the tasks they would be expected to complete. Depending on their user class, participants were instructed to complete a set of benchmark tasks that required the use of the LuxCopia system to navigate through the supply room and complete the assigned tasks. For frontline clinicians, they were told to "gather all supplies needed for Natasha Ferry’s dressing order" and "locate and retrieve gauze pads for immediate and urgent use." For the supply managers, they were told to " identify and restock the item of highest priority." The participant was also instructed to consistently verbalize their thoughts and actions throughout the completion of the task, following the think-aloud protocol, which allowed us to stay on track with our Wizard-of-Oz setup as well as capture the reasoning of their actions. 

We created an intervention policy that was strictly enforced to preserve ecological validity. If the participant paused for a long time, neutral prompts were used (ex. “What are you thinking right now?”, “Please continue in the way that feels most natural to you.”). If the participant asked for help, they were encouraged to continue independently. If they were completely stuck, we made a note of it and encouraged the participant to continue trying to finish up their task along with a small suggestion of where they could focus their next action. 

Inside the room, one team member was designated to support the participant if needed, while minimizing intervention to preserve natural behavior. The remaining members were responsible for coordinating the Wizard-of-Oz setup and data collection based on what was stated and observed about the participant. After completing their task, the participant was given more context about our project's goals and asked to fill out a survey about their experience with LuxCopia regarding ease of use, confidence, and convenience. Once the participant left, we had a team discussion about what we observed and learned from each session; we made notes primarily about what went well and what didn't, what the main issues seemed to be, and what changes we could make to the current LuxCopia design. 

Scripts for Evaluation Studies:

Frontline Clinicians

• "Your name is Parker Schule, a frontline nurse who needs to grab items from a supply room. We are going to now move into the task portion of the session. Nurses are usually tasked with 5 patients per day where they can check items. For the task, please work through it as naturally as you can. I may read the task aloud, and then I will mostly stay quiet so I can observe how the system supports you. Please remember that we are testing the system, not you. If anything feels confusing, that is helpful for us to learn."

1. User should get all items for Patient Natasha Ferry

• • • “You are preparing to complete Natasha Ferry’s dressing order. Using the system (tablet) and the supply room setup, gather all supplies (white pieces of paper) needed for Natasha Ferry’s dressing order and leave the room with the correct items. As you work, please say out loud what you are looking at, what you expect to happen, and anything that feels confusing or easy.”

2. User should get single item (gauze)

• • • "You need to find gauze pads for immediate and urgent use. Using the system and the supply room setup, locate and retrieve gauze pads and leave the room with the item."

Supply Managers: 

• "You are working as a supply manager, Dave Chen, responsible for maintaining inventory in the supply room. Throughout the day, items may run low or run out, and it is your responsibility to quickly identify which items need to be restocked and replenish them efficiently. For this task, assume you are starting with a bin of supplies that need to be restocked." 

1. User should find main item needing to be restocked and then restock it

• "You are restocking the supply room for only the highest need item. Do not worry about calibrating scales for now. Using the system and the supplies provided, identify and restock the item of highest priority."

Intervention Policy and General Statements

• If participant pauses for a long time, say things like:

  • “Please continue in the way that feels most natural to you.”

  • “What are you thinking right now?”

  • “What are you looking for next?”

• If participant asks for help:

  • “I’d like to see how the system supports you on its own, so please do what you would normally try next.”

• If participant is completely stuck and does not proceed:

  • “I can note that you would like help here. For now, please continue with whatever you try next.”

• Once participant indicates they are finished:

  • “Thank you. I’m marking that task as complete.”

• If needed, ask a neutral follow-up:

  • “What made you feel that the task was complete?”

Post-Survery

Data collected through evaluation: 

Pictures of us and testers during evaluation:

Findings & Changes Reporting

Summary

A total of 10 participants were recruited for the evaluation. This included 5 participants acting in the frontline clinician role, which are labeled as interviews 3-7, and 5 acting in the supply managers role, labeled as interviews 8-12. All participants in this evaluation were UVA undergraduate students with various backgrounds from CS and pre-med and were either of no relation to the team and recruited through cold calling, or recruited via a personal connection. Other background information may be documented in the individual interview notes.

Procedure

Participants completed evaluations structured around three main task flows. Frontline Clinicians perfomed two of these flows, patient-based item retrieval where the task was to fulfil a patient order by selecting the correct patient and retrieving their required medical items and one-off item search in which participants searched for and retrieved a singular medical item intended for general or emergency use. Supply Managers were only given one task which was to Restock the highest priority item, where they had to identify the highest priority (lowest stock) item, locating its physical position, and completing the restocking process..

Participants were observed throughout the completion of their assigned task(s). Time-on-task, navigation behavior, notable errors, and verbal qualitative feedback either summative or formative were recorded.

Apparatus

For evaluation we used several methods/tools to gather data during testing. Observers were stationed in the room where the evaluation was taking place for the purpose of taking notes during formative feedback. Observed participants during task completion and recorded qualitative notes on behavior, hesitation points, usability issues, and feedback. A timer was also used to measure task completion time for each participant across all task flows, allowed comparison of efficiency between every user and task. Regarding summative feedback, we used a Google Form to administer a post-task questionnaire, used to collect quantitative ratings (e.g. Ease of Use, Trust, Convenience) as well as any qualitative feedback. 

Overall, participants were able to complete tasks successfully, but often required exploration, trial-and-error, or external prompting (in the worst case) .

In addition we noticed several key patterns throughout our evaluations such as confusion about where to begin tasks, the low discoverability of the “Find Items” button specifically, hesitation going from a digital UI interface to physical interaction, and an inconsistent reliance on system guidance (e.g. light cues vs manual item identification through the UI and scanning through labels).

Findings & Changes Table

Interview Notes (Supply Managers)

Supply Manager (8)

Time Taken: 3min 18sec

Got a lot of work cut out for us

I did not need to find the thing without the light so just started filling in stuff.

Did not find their 

Entry of items was slow.

Interview 9: 

Time Taken: 4min 49sec (Restocking Highest Need Item)

Notes:

• FIRST TASK

  • Goes directly to supply bins to attempt to see the highest need item by physically looking at the quantity of bins

  • Clicks “Contact Admin”

  • Goes to frontline clinician due to bug in prototype, quickly corrected

  • Lands on Supply manager screen

  • Clicks on IV bag first (no screen setup as not the highest need item)

    • Misinterpreted “Highest need’ as in the most important item instead of lowest in quantity ranked in priority.

  • Clicks on Gauze next, gets confused on what container to restock (the basket he came in with or the supply bins on the walls)

  • Starts restocking the bins on the wall with items

Conclusion: Was more confused at what the task was than the system, confused that the items in the bin

Interview 10:

Time Taken: 2min 51sec (Restocking Highest Need Item)

Notes:

• FIRST TASK

  • Clicked sign in button first

  • Mentions red and yellow is likely an indicator

  • Clicked on all items

    • Mentions items were listed in alphabetical order

  • Clicks on priority listing then gauze

  • Went back and clicked on text search

  • Searched Gauze pads and clicked on it

  • Clicked around on that page was confused on what to do next

  • Hesitant on going from a digital format to physical

    • Asked “am I supposed to go up there and interact with it”

  • Was given confirmation that anything in the room was allowed to be interacted with

  • Supplied the bin with the items and completed the task

Conclusion: The only point of confusion was the hesitation of going from a digital ux design to physically interacting with our model, realizing priority listing listed items in highest need after realizing all items listed it in alphabetical order. 

Interview 11:

Time Taken: 1min 14sec (Restocking Highest Need Item)

Notes:

• FIRST TASK

  • Clicked on “request item” at first.

  • Goes directly to Gauze Pads. Picks up Bin and stocks item

  • Says “first place is confusing because you can only go there by going to admin view”

Conclusion: Was the quickest, only point of confusion was the ambiguity of requesting an item and viewing priority of items.

Interview 12:

Time Taken: 3min 9sec (Restocking Highest Need Item)

Notes:

• FIRST TASK

  • Clicks on login and goes to gauze item

  • Clicks calibrate scale

  • Goes back to gauze

  • Is hesitant to go from digital to physical.

  • Puts items in the bins

Conclusion: Was confused by the representation of the physical model, specifically how bins were represented and the what the items were

Survey Responses