This project was a part of my graduate degree’s capstone project. I was on a team of four other students who worked together, over a course of 8 months, to apply and grow our design thinking skills.
Zensors is a AI powered visual sensors that provides users with tools to collect and analyze data about their physical space. For my capstone, my team and I were tasked with determining the most compelling product market fit for Zensors and create a user tested MVP to support future business growth and development.
Business Strategy, User Interviews, Usability Testing, Interaction Design, Task Flows, Personas, Experience Mapping, Wireframes, Prototypes
Sarah Papp, Tony Wang, Marvin Kennis, Yutong Chen
Who is Zensors?
Zensors was developed from research conducted at the Future Interfaces Group at Carnegie Mellon University. The technology harnesses camera’s video stream, computer vision, and machine learning to gather about an environment or events in a space.
For example, point a camera at table, ask the following question: “How many people are sitting at the table,” and then over a period of time the Zensors algorithm will be able to provide you with data that might help you understand peak usage times.
In the most basic sense, you can ask Zensors a question and we will harness the camera’s video stream to answer the question.
Working with our client to understand their research journey until now, and their future business goals. We were able to uncover the following goals:
ONE: Find a Market fit
Zensors is still a nascent, complex technology, thereofre it's crucial the first market fit we choose to develop within is set up to be successful for both the platform and its users. We wanted to explore a myraid of businesses that rely on understanding the way their space is used, and then see if their data needs and business goals map to Zensors offerings.
TWO: Build an MVP
To transform this from a research project into a viable business, we wanted to build an MVP that put our product into the hands of customers. We wanted the user to be able to successfully set up questions and gather data from a camera's video stream. Building the MVP tailored to a market's needs would not only help turn set up the product to acutely solve a need and prove it's impact, but would give the Zensor's team a lightweight platform to gain user insights and continue to grow on.
Defining Our Market Success Criteria
Zensor’s nascent, complex technology has the capacity to change the way people harness data from their environment, but to ensure that Zensors is being built to support users and their goals we needed to, first, define the tech’s capabilities and shortcomings.
My team worked closely with Zensors to achieve the following:
understand how the tech worked
distill its core value proposition
identify weakness and limitations
create a set of criteria that was necessary to make Zensors a compelling and successful product.
Investigations to Find a Fit
With our success matrix in hand, our endeavour to find the bet market fit for Zensors began! We identified 11 industries that we found relied on understanding their environment to improve their business.
Next, within each industry we conducted extensive user interviews with people who are managing the space and making business decisions - managers, data analytics, floor supervisors, etc. We were trying to gain insights that aligned directly with our success criteria. We were trying to understand the following:
How the business leveraged data
Existing access to cameras
User/employee privacy concerns
Environments to be monitored
Financial incentive to invest in Zensors
Market Fit Found
Coworking Spaces Wins!
Coworking by far was the best fit for Zensors! Coworking spaces were hungry for data & lacked any tool to support that need, it was the best match for both parties.
Moving On to Goal #2!
With our market fit in hand, we jumped into the next phase of our project: building an MVP that supported our user through successful camera setup, question creation, and data collection. In turn this gives the Zensors team a scalable MVP that will start validating their business model, help train their AI models, and a place to continue to build, refine and grow their business.
In efforts to understand the compatibility between Coworking and Zensors more deeply, my team and I conducted some in-depth user interviews. We wanted to understand how the Coworking industry works as a business, how all stakeholders work within the business, and how is success measured.
We conducted more user interviews, but this time inside Coworking spaces, to understand the manager's day to day tasks, how the interacted with members, what metrics they were responsible of, how they currently tracked data, and how the business made decisions. We even interviewed coworking members to understand how they used the space, what their goals and needs were within the space, and what pain points they were experiencing.
The insights we gained were centered around how managers could leverage Zensors and the type of questions they needed answers for. These insights became a guide to the way were were going to build the MVP but it also become a great tool to align the team and business.
Creating Assets to Align our Team
An additional benefit of conducting these interviews was creating design artifacts to educate our team and stakeholders. When synthesizing our user data into insights and theme, we found the best way to share the things we learned was to create vizual, tangible artifacts that our clients, research partners, and developers can use to build a connection to the user and their needs.
We create experience maps and personas, each that highlighted the following information:
User goals (Coworking Managers and Branch Managers)
Relationship between stakeholders
These artifacts helped us create high level user flows and aided our ability to prioritize and set scope.
Crafting the Question Creation Flow
For users to successfully craft a question that Zensors can answer requires it requires specific parameters and inputs. To guide our users to successfully craft a question that will no only give them the answers they need, but help our AI models learn we broke down the process into seven steps each with a goal and clear instructions. I have highlighted a few of the major steps in asking a camera questions.
Determining Question Type
This question is two pronged, firstly it works as a nudge to get our user to frame the question correctly. Yes/No, “where”, and “how many” were the major buckets our system is built to handle efficiently.
1. Give clues and define each question type
2. Show a sample “Data View” so user can set expectations
3. Leverage this to nudge users to frame question so the system can understand
Framing the Question
The question type is followed by typing the question, we are able to suggest the correct framing for the question and give recommendations for alternative or additional questions.
1. Recommended question to improve framing
2. Suggestion for additional questions
3. Auto-frame question to previous question type
Video Region Selection
A key part of asking the question is setting it to a specific camera and evenmore, a portion of the view. This step has a universal picture editing interface, where the user can drag to select where they want to target the question towards.
1. Leverage existing metal model for picture editing
2. Give location of camera and sample view
This is one of the more complex portions of the process, we need to get the user to tell Zensors how frequently to analyze the video feed.
1. Instructional text to explain the goal of the step
2. Concrete examples of sample rates that support specific events
User Testing and Learnings
To validate the design decisions and the assumptions we had made so far, we user tested our question creation flow. We asked our participants to imagine they managed a coworking space and wanted to create a question that helped them understand how many people were sitting in the lobby.
Our learnings were extremely impactful, because it completely had us redesign this flow. We found that
Users did not understand terminology. For example, “sample rate,” “frame region,” and “binary” were difficult to comprehend and act on - no matter how much help text.
Users failed to accurately set video analysis speeds. They weren’t able to build a mental model around the fact that Zensors only analyzed the video frame periodically. Furthermore, they didn’t understand that more frequently frame rates meant more expensive questions.
“The process was too long.” Most of the participants complain about the length of the process, although most were straightforward steps, it still felt too long and arduous to complete.
Version 2 of Question Creation
With the insights we gained from our testing session, we made some major updates that allowed us to condense the flow down to one page.
Remove redundant inputs
Make the system more intelligent to user inputs
Provide contextual help tips
Ensure successful creation with a sample question
Results and Next Steps
The Zensors team went from just a research project to a tested MVP set to launch with a local co-working space. Through all our testing, we were able to gain a client interested in testing our Zensors at their co-working space. Through this process:
✅ ONE: Find a Market fit: Identified what makes an industry a good fit for Zensors. With this success matrix they can continue to validate and gain more market fits.
✅ TWO: Build an MVP: We build a tested MVP that is great base to validate with real customers to learn and grow on.
Helping your Client Understand the Value of Design
When working with clients or stakeholders who don’t understand the value design brings past making things look consistent or adding delightful interactions, I find it super powerful to create design artifacts to show the value and impact design has on the process and product. In this project for example, the Zensors team only saw our wireframes once a week during review session, but to include them in the entire process and expose the design rationale inside those wireframes, we created and shared persona briefs, experience maps, and task flows. These help the team no only understand the design process but also become value tools in getting the team aligned, set priorities, and build connection to the user’s needs and goals.