The taxonomy team develops and maintains centralized taxonomies, supporting 60 countries and 28 languages. Their work is mission critical to facilitate matching between jobs, job seekers and employers.
When Indeed entered a new market, it took too long (about 1 year) to build parent-level occupational categories.
The team built a machine learning (ML) prediction model to accelerate the process but lacked a proper tool for analysts to interface with the ML predictions.
Reduce the time it takes to create top-level hierarchies by launching a tool for taxonomy analysts to interface with and refine the accuracy of model predictions.
This was a temporary, high priority assignment. As the sole designer, I was given one quarter to launch an MVP. I met the deadline :)
My most impactful efforts:
- Stakeholder and analyst interviews
- Research readout for leadership
- Instilling a user-centric perspective
- Prototype + testing
- Defining a vision for the future
My work accelerated Indeed's ability to open in new and emerging markets.
- 1 year ⇢ 3 months to build parent level hierarchies
- Received the "Indeed innovation award"
1. Country level management
2. Parent level management
3. Child level management
The table was completely redesigned to help managers maintain a birds-eye view and analysts to better prioritize their work. Irrelevant fields were removed, clear status indicators and permission-based quick actions were added.
The revamped information architecture allowed analysts to quickly check, refine and take action on ML predictions. Even though the core experience didn't change much, these tweaks yielded gains in productivity.
There were times when more attention was required to accurately classify a job into a certain category (this would help train the model). The new focused experience allowed analysts to check references, compare and assign categories.
Leveraging insights from the MVP, I put together a 1 year vision on how the team could further improve the product.
