Pupil Dimensions
During my time at Pupil I spent a considerable amount of time leading the design of internal tools used by the Digital Production (DP) team to verity, clean and annotate spatial data. Dimensions is the result of my analysis of the business needs and the team pain points. It quickly became a priority project to allow Pupil to scale and service customers more efficiently and resulted in a ~25% speed increase on first release.
Team
1x Lead Product Designer - myself
2x Front-end engineers
1x Back-end engineers
1x Product manager
1. The Problem
Pupil offers a property capture service to real estates agents and property owners that allow them to get high quality photos, virtual reality / 360 tours and super accurate floor plans in less than 24 hours from the moment the capture team enters the property.
If the assets are not delivered to the customers in less than 24 hours, Pupil’s policy was to integrally reimburse the cost of the service.
The biggest KPI for the startup was logically the time to deliver assets and at this point in time the production pipeline was barely holding the demand for capture, making the whole business model unreliable.
Hence the tech team biggest problem needing to be solved at the time was:
”How can we make our Capture and Producer teams as efficient as possible?”
Solving this problem meant doing two things for the technology team:
- Short term: Fix the pipeline by building an ecosystem of tools that will boost the teams’ productivity and improve workflows.
- Long term: Through these tools, feed machine learning algorithms to automate as many parts of the workflows as possible in order to increase quality and speed of delivery.
I took the lead to improve the Production part of the pipeline, where the Digital Producers (DPs) needed to:
- Check and fix any issues on point clouds from the Capture team
- Annotate the point clouds (rooms, doors, stairs…) to generate measurements & floor plans
- Create hotspots for virtual tours to switch between rooms / floors
2. Immersion
First I needed to understand how Pupil’s Digital Producers worked in details. I spent 2 days shadowing them in the office to understand their journey, from the moment the received a capture to the moment they uploaded the final assets to the cloud, ready to share with the customer.
At the end of this Immersion phase I mapped out their journey to have a bird’s eye view on their process and start creating concepts around those three objectives:
Build new essential features: Users should be able to fix the point clouds directly in the tool to reduce back and forth between the capture and production team. They also should be able to update the team on the progression of their work quickly and effortlessly.
Reduced and bespoke interactions: manipulating and understanding 3D data requires precise and targeted actions, empowering users to act on the data in a very visual and easy way.
Towards a “read-only” workflow: keeping in mind that some of those interactions will be replaced by algorithms performing the work for the users, moving their roles towards “Quality Assurance”.
An overview of the process that Digital Producers followed. A very bloated process using tools either quickly put together or not fit for purpose, and most of the time quite buggy.
The future process map I created to share with management. It drove conversations on product, operations, and technical levels. It allowed the whole company to grasp better the scope of the new tool Pupil Dimensions.
3. Design
I started my design process by creating the overall structure of the tool, frequently going back to the Digital Producers for validation (one of the perks of designing internal tooling). I attached the main pain points, needs and requests to this structure, making sure I answered them all directly in the essence of the interface.
The Pupil Dimensions tool: a bespoke spatial data editor in the shape of a web app. Digital Producers can take the capture data and transform it into a ready-to-consume floor plan based on millions of measurement points.
Property Tree
Users needed a comprehensive list of all the elements in the property and have direct access to them. I designed this tree to be close to tools they were familiar with (Photoshop and Illustrator)
Object Library
I introduced the concept of a drag and drop library of symbols that producers can use to quickly build out their floor plans directly on the top-down view of the point cloud.
Version Control
Quality Control needed a quick way to see the change history on a property and sometimes revert to a previous version if necessary.
Integrated Feedback Loops
A lot of communication had to flow between the Capture team, the Production team and between colleagues. Pupil Dimensions had to cater for dropping comments onto the point cloud so the whole team could learn from their mistakes but also be aware of the great work they were doing.
Floor Plan Manager
Eventually Pupil Dimensions would auto-generate beautified floor plans from the annotations the Production team would perform on the point cloud. I designed this sub-view to manage auto-generated vs user-added floor plans, so the machine learning team could improve their algorithms looking at the difference between the two.
4. Implementation and release
I followed closely the implementation of the designs for the first MVP release and even helped the engineers on some front-end stories so they could focus on the “meat” of the application. Here is below a video of the first MVP release in action.
While testing the tool with the Digital Production team, we measured the time to completion using the old process vs the new Dimensions tool on similar properties and measured a whooping 25% reduction, which allowed Pupil to take on more orders and build towards a more automatic process.