My Role: Research| Character development | Persona | Interactions | Testing |
2017 - Duration: 1 Month


Product Summary
Product Walkthrough

Product Summary



The most valuable assets in a retail store are its dressing rooms, where customers are three times more likely to make purchase decisions than someone on the sale’s floor.

But, in our research, we learned that dressing rooms are treacherous grounds, as they bring customers face-to-face with their own self-image.

When you go in the dressing room and think it’s going to look great on you and then it either just doesn’t look good or it’s not the right size. It can feel defeating.
- A 32 year old shopper

How dressing rooms address such emotional ups and downs can impact customers’ attitudes to the brand they are shopping and their decision to buy.



Verbal Outfitters

  • No sales floor, only dressing rooms.

  • AI powered virtual assistant.

  • Minimal interaction with sales associates.


Powered by emergent technologies, CUIs (Conversational User Interfaces) promise to transform our interactions with computers into more intuitive, human-like conversations.

Our solution relies on CUIs to assist the customers, creating a personalized experience that has the power to mitigate customer’s emotional state, and engender trust.


Product Walkthrough

A. Make an outfit: The assistant creates a tentative outfit based on an initial conversation. 

B. Mix and match: The assistant presents the customer with similar outfits. The customer can  choose whole outfits, or mix and match different pieces.

C. Try it on: The lights in the room can adjust to mimic the environment, in which the outfit will be worn. 

D. Purchase: The customer can add chosen items to his shopping card and pay at the register.




In this one-month long project at Carnegie Mellon University, in a team of three, we examined the potential for deploying CUIs to a dressing room. 

  • Research Objectives

  • Research Insights Recap

  • Ideation Overview

  • Prototyping and Testing

  • Layout

  • Security and Accessibility

  • Rapport

  • Final Prototype and Testing

  • Conclusion

Research Objectives

To understand how customers experience a dressing room, and what technologies could feed our CUI design, we did:

  1. Literature reviews consisting of 15 articles, 2 forums, and 1 book.

  2. Interviews with four people

  3. Competitive Analysis


1. Literature Review

We learned that dressing rooms are the most valuable touchpoint in a store.

  • They show the customer’s intent to buy.

  • Customers in a dressing room are three times as more likely to buy than someone on the sale’s floor.

  • Dressing rooms are unique environments that can cause discomfort and anxiety.

Where else can we take arms full of other people’s clothing into a room where we’ll take off all of our clothes and put on theirs.
- Margaret Laney

We also learned how to effectively deploy CUIs:

  • It is recommended to use mixed media, when considering conversational user interfaces. This allows combining people's auditory and visual strengths.

  • The strength of voice interfaces is in their ability to mimic a conversation. By relying on conversational conventions, it can create a natural dialog flow.

  • CUIs can help foster trust, because they are non-judgemental.


2. Interviews to build empathy

The goal of our interviews was to better understand the emotions at stake in a dressing room. We interviewed:

  • 2 males, and 2 females.

  • In their 20s and early 30s.

  • From three different countries.

We learned that a dressing room can be a particularly anxiety-inducing space:

  • Customers have to face their own body image.

  • Customers’ own body images may contradict what really fits them.

  • They have to make a purchase decision under stress.

We saw that to mitigate these anxieties our interviewees had some common methods:

  • Relying on friends and family for emotional support.

  • Shopping with someone.

  • Connecting to one’s community by texting or through social media while in the dressing room.


3. Competitive Analysis

Next, we examined two different types of technologies used in dressing rooms: Traditional, without any sophisticated software, and modern tech-driven solutions.

Our goal was to see how these respond to the emotional experiences inside dressing rooms. We saw that:

  • Most stores did not use any interactive technology in the dressing room.

  • Buttons were used as a low-tech solution in some stores.

  • Only a few high-end stores had installed high-tech mirrors in their dressing rooms.

Mistrust was an unfortunate after-effect of some of these high-tech solutions. By trying to improve the image that the customer saw of themselves, they presented a distorted version of reality.

Given how vulnerable the customers may feel inside a dressing room, it is important to strive for honesty and breed trust.

Research Insights Recap

We found dressing rooms to be the most crucial touchpoint in a retail store. Today most dressing room technologies explore logistical aspects of improving the experience. What they leave unattended are issues around:

  • Emotions

  • Support from social network

  • Trust

Ideation Overview

We used insights from our research in a three-step ideation process to come up with a product.

  1. Exploratory Scenarios

  2. User Persona

  3. CUI Character


1. Exploratory Scenarios

We translated our insights into 16 four-part scenarios. These stories helped highlight the different ways to create value with a CUI inside a dressing room, emphasizing the problems we had encountered during our interviews.


2. Persona

We created Roberto to narrow the scope of design and situations that the CUI would have to respond to in our prototype. Roberto’s persona was crafted such that it would exhibit most of the issues we saw people struggling with in a dressing room.  


3. CUI Character

Finally, we created the CUI character, keeping in mind Roberto’s low self-esteem and social anxiety. Our CUI was to be helpful and polite, and attentive to Roberto’s demeanor and his emotions, as expressed in his responses.

Prototyping and Testing

Based on Roberto’s needs, there were four types of principal tasks that the CUI would have to support:

  1. Indicate CUI’s turn in conversation

  2. Create outfits and allow item selection

  3. Complete a purchase

  4. End the conversation

We used paper to create visual assets, and turned to Wizard-of-Ozing to iterate on the experience. 

CUI Iterations

We did two rounds of design and testing. 

  • Bodystorming: We created a mock-up dressing room, and enacted the possible conversations between Roberto and CUI. This way, the script we created kept the flow of a natural conversation.

  • Table reads: We read out loud the script for others to catch any awkward moments. For example, the CUI started every answer with “Ok”, making it sound repetitive and unnatural.

  • Paper prototypes: To support Wizard of Oz for interactions that needed visuals, we used paper mock-ups. For example, we displayed lists visually to avoid overwhelming Roberto’s short-term memory.


After two rounds of tests and modifications, We learned that, even though our design was good at providing service, it couldn’t engage people emotionally.

In a traditional outfit store, voice interactions seemed superfluous, since buttons and screens could have done as much. We needed a new shopping paradigm: 

What if buying outfit was more like getting advice from a friend?

So we untethered our design from the conventions of a brick and mortar outfit store. Verbal Outfitters became a store without a sales floor.


Security and Accessibility

During our design we kept in mind security and accessibility. Because of privacy concerns, we decided against using cameras pointing at customers. For customers with a hearing impairment, we thought of turning to a texting framework.  


Even a simple transactional conversation can convey friendliness and care, and help put Roberto at ease. For example an elegant error recovery is a crucial opportunity to reassure Robert that he can trust the system.

Final Prototype and Testing

For the final prototype, we we created a mock-up consultation room out of boards, and used a projector to show the screen interactions. Voice interactions were achieved through Wizard of Oz.

We tested with three participants the ability of the CUI to engage the people emotionally. We paid attention to participants reactions to error recoveries, and the overall flow of the conversation.

We saw that participants have a positive reaction to error recoveries, and there were no jarring disruptions in the flow of the conversation. 


We learned that to effectively draw on the human-like quality of CUIs, we had to create the right context. That's how we came up with Verbal Outfitters. The next step for Verbal Outfitters would be to automate the Voice Interaction and take the consultation room to an actual store. As machine learning technology improves to faithfully depict outfits on all types of human bodies, the CUI could turn into an assistant app, and customers could access it anywhere.