Launch the People Analytics feature on the Command system in 2 months to be tested for advanced clients, and fully launch in 3 months to all users.
I'm the lead designer on this feature.I defined user goals, pain points, designed interface, and interactions, and prototyped rapidly and worked with one product manager and several engineers to launch this feature.
As an enterprise security system, users often face complex tasks, and the interface still designed in an unintuitive way. When they need to search for one person of interest or a vehicle, they find the searching task is not only time consuming, but also hard to make the result accurate.
We were tasked with evaluating and revamping the current Verkada search function to support users to find the objects more efficiently and accurately.
Originally, the product manager only planed to develop the People Analytics on a single camera basis. During the initial user interviews with Verkada's loyal clients, I discovered the potential need to expand this feature scale.
I initiate the feature exploration with computer vision engineers, back-end engineers, and the product manager. We successfully map out the expanded product development roadmap based on technical ability and resources.
We designed, validated, and delivered a new, intuitive People Analytics feature on the web application and iOS application. This feature focuses on allowing users to easily navigate and search for one object in the complex enterprise surveillance system.
While this time-bound project didn't start with a defined set of requirements, we iterated rapidly to drive the project with design. Not having time for rigorous discovery research, we actually began designing at the very beginning. And along the way, every stakeholder demo and user testing at conferences became our source of data. We iterated and delivered hi-fi prototypes to prompt use cases, pain points, and feedback for interactions and UI.
This approach helped us to validate assumptions at different scales, from evaluating project goals at the beginning to examining usability later on.
At the beginning of designing this feature, I received many ideas or suggestions from my stakeholders. Every team has its requirements and preferences. The product manager would like to set the entry point at the left navigation; the engineering team wanted to show the full frame results; the computer vision team wanted to highlight the person's face, etc.
After considering the user's needs and experience, I explored several options for entry points and layouts. It took multiple rounds of user testing and many iterations for the People Analytics project to evolve. The final result was a simple, clean, and visually driven single-page of people history that showed only the cropped results contains a person under the camera. This also provided inspiration for the later smart analytics features development.
Exploration: Results display
Exploration: Entry point
Exploration: Display method for faces
I decides to display all the faces that been detected without asking the user any information. This is a quick way for them to start the people search task. In the meantime, there are multiple paths for them to narrow down the search range and find the target person.
I defined 3 major user journeys based on user interviews:
"There is way too much footage. It's hard to search."
Previously, users had to triage through the sea of footages to find if people of interest show up. One major pain point we found out is that locating the date and time range and distinguish people of interest from normal employees.
Users can quickly scan all the people who entered their physical space. Without any click, users can quickly view all the activities on the hover states. The blue bars indicate when there is an activity.
Offering the People Analytics feature on iOS platform is also very crucial, security managers are frequently use this feature on the go.
"I don't know what time the person came into my store yesterday afternoon, but I remember it was a man ware a red T-shirt."
To help our users to find the relevant results with a vague description, we need to provide an experience that can help users to search with attributes. Such as gender, appeal characteristic, and time range.
The design will never get perfect, by iterate with feedback from clients and stakeholders, the design can get closer and closer to perfect.
I pay attention to every detail in my design process. The following is an example of how the clothing color filter got iterated.
Users can only remember some characterizes of the searching target, such as gender, apparel's color, and date range. By prioritize these searching keywords, we organize the attributes into 4 groups. Users can easily select the relevant criteria in the filter dropdown, and the results will change accordingly. The selected criteria will become chips to enhance the current search keywords. Users can remove any criteria by click the chip.
"I want to know when this person shows up in all my history footage."
Feedback for the early iterations turned out that after knowing one person could be suspicious, users often want to know when did this person shows up for the first time, and where else.
Due to the development constraints, searching in multi-cameras is a phase 2 project.
Users recognize the target's face, the users can view all the results related to the person by clicking the avatar.We choose the circle shape to represent people's faces as it can reduce the user's recognition time. Each result only shows the targeted people, the algorithm helped to crop the frame to the specific area.
One design challenge I was facing is once the system only shows results related to one face, often the results look very similar in a short period, such as a person stand in the kitchen area and doing small movements. I explore different ways to display results, and the final solution is offering a small album looking stacked results. This design saves spaces on the interface and reduces the working load for users to manually go through all the similar results.
We also discovered that there is another use case through the early usability tests. School union clients brought up a use case that they have the target's photo(such as sex offenders on the policy record), clients want to upload and search through the whole footage storage.I quickly iterate my design based on the feedback. The upload and search function in the release version.
To make People Analytics more intuitive to users, I proposed a new feature: People of Interest Notification. Taking advantage of users already finished the search task and find the target, I encouraged users to create profiles for faces.
After an easy setup, users can easily browse the target person's results without start the searching process. And they can set up an alert on any faces they want to and receive instant notifications once the target person shows up.
Based on the different characterizesand user behaviors of web applications and mobile applications, I designedthe same function with different experiencesfor website and mobile applications (iOS and Android).
For mobile applications, I emphasized the uploading search function. Since it is easier to use a mobile phone to take a phone then start the search process. And to be more considerate of the constraints of mobile screen sizes, I also decide to only offer fewer functions on the mobile application.
The People Analytics feature was announced in the high-value partner conference and rolled out to 100% of customers with a lot of positive feedback. This project also defined a natural way to combine machine learning technology and IoT more efficiently. In addition, it indicates good potential for how much more can we do in the future.
"I like the structure and simplicity of the interface. It makes finding any abnormal activities easy."
- Security Team Member
"This makes less work on my part to do new employee training. Besides, I have more power in controlling monitoring people of interest"
- IT Security Manager
This feature was launched and became a strong selling point.See the blog post.
By the end of the quarter, this feature released to more than 10k cameras across the United States. It increases 121% of the sales revenue and 140% of the trial requests.
One of the biggest challenges of this project is the changing requirements. Initiating as a "feature could be useful", the project goals changed several times by considering clients' strong desires, technical constraints, and resources limitation. As the lead designer on this feature, I articulate the essential problems and empathy the client's needs. By balancing different needs from the cross-functional teams, optimize the user experience, and focusing on the ultimate goal, I decided to make the project into several phases.
I strived to maintain thorough documentation, systemize design components, and work closely with engineers to update the component library. I also realized that presenting the design to stakeholders is not just taking in feedback and iterate, but it is also a way to communicate back to stakeholders about my design vision and design philosophy. With a thorough understanding, it can benefit the collaboration in a long run.