A project that got me excited about loading states and taught me what a stream URI is. We wanted to enable anyone to connect commercial CCTV systems with our AI software and succeeded!
Background
ThirdEye is a London-based startup creating AI video analytics for loss prevention and operations in the retail industry. I joined ThirdEye in June 2019 as the company’s first designer.
In early 2020 the company was getting ready to expand beyond enterprise retailers. The aim of creating a full package system for small and independent retailers that could be installed and operated by the retailers.
Prior to this, the installation process was highly technical and ‘bare bones’. Servers and software were installed by our own infrastructure engineers or by service engineers trained by us. While this worked for enterprise clients, it was in no way scalable or practical for smaller retailers. This was the starting point for the technical and design challenge: create a tool that allows anyone to install and configure the hardware and software needed to run the video analysis AI software.
Research
On the other side of the problem was the technical challenge of connecting existing CCTV systems with the AI video analytics software. The other side was the intended user, the owner or manager of a small retailer, who would be in charge of the installation process. First, I had to understand both better.
I spent time with my colleagues, the infrastructure, software and machine learning engineers who built the technology. Together we mapped both the current and the ideal installation process to gain an understanding of what is involved. We then broke each step into detailed pieces: what is the purpose of this step, how is it currently done, what needs to be built for this step to be achievable without engineers, and what could the user do. An example would be connecting the AI server to the existing CCTV system.
To understand the user and the physical environment where this process would eventually take place I recruited three colleagues to help gather data. We contacted and visited over 200 small businesses to interview owners and managers about both loss prevention and CCTV systems. This was at times extremely challenging for multiple reasons: small business managers are busy, loss prevention and CCTV are both sensitive topics and tightly regulated.
What we found helped define both the technical process and the first prototype
- Trying to reduce shoplifting is fighting a losing battle in small businesses. The loss is significant, but individual instances are low value making intervention not worth the risk or law enforcement involvement.
- Every single store contacted had a CCTV system in place. However, most had not installed the system themselves.
Problems
- There is a large number of CCTV camera manufacturers and consequently a large number of technical specifications, with little in common. What works with one CCTV system might not work with the next.
- It might be difficult for a user to find information needed to complete the installation process. Some technical details are a struggle for specialised engineers, meaning even technically savvy users would likely not find them.
High-level user journey
The first user journey was created based on the insights from technical mapping and user research. At this stage I knew what would need to happen for the technology to work and had an understanding of the possible challenges for users.
Wireframe MVP
As the challenge was as much technical as it was design, as a team we decided to start with a wireframe MVP. Much of the technology was built from scratch during this process, so we wanted the tool to be focused on the background functionality for testing.
First tests
At this stage COVID-19 had reached the UK and the country was in lockdown. Due to lockdown we were never able to test the MVP installation tool in a real life scenario in a store. Some of the processes were still manual, which meant testing would have required physical presence.
However, the MVP version was tested during an internal demo. Without users the focus was on the technical implementation and usability of the tool from a general perspective.
Problems:
- Execution of some steps took longer, leading to gaps in system visibility as users wouldn’t know if a process was running or if the system had failed.
- The tool was missing an option to cancel and confirm data was not stored. A step in the process requires the user to input login details for the CCTV system, which is sensitive information.
UI and second phase
Following technical testing the tool was updated with a branded UI, in line with other ThirdEye interfaces.
Final words
Due to the challenges created by COVID-19 the tool is still awaiting real life testing and deployment. However, even though the product hasn’t made it to the world yet, the project created an opportunity for me to grow as a designer. Designing a plug-and-play installation tool synchronously with the engineers I work with building the technology was a challenge of complexity and timings. As a designer I am proud of creating a product from scratch, challenging my design thinking and visual design skills, and hopeful that the next iteration is awaiting.