Datasets

As a Census user, you have to define your data in Census to sync or segment that data. There are various ways that Census supports these definitions. However, all require a multi-step process involving creating multiple things in the Census product.

Creating these data definitions simply required too many steps.

This multi-step process was ridden with inconsistencies and friction that prevented users from fully using the Census product. It was clear that improving the Census product needed to start with a fundamental rethink of this structure and process.

I worked with the product manager and engineering team to define, design, and ship a feature which would simplify the process and combine the two concepts into one.

The initial hypothesis was that the more the creation flow helped the user completely setup the dataset, the more valuable the Census product would be. This is because to do anything in Census, a user had to have at least a materialized model.

The early explorations focused on different ways that users would setup a dataset. This included not just creating the data definition but also naming the schema and related columns.

I designed a modal-based setup workflow that would minimize friction in setup. Each step in this system could also be isolated for the editing experience of that item. Further simplifying development and reducing cognitive load for users.

The largest challenge with this project was the need for progressively enhancing datasets. Datasets could be setup with only a data definition, but the true value would be unlocked when users performed additional actions on the dataset.

I designed a system that enables users to get access to key components of the workflow through full-page modals.