Celine Chang - UX Designer

Polaris - Data ingestion

Optimizing the data loading experience for ease and clarity

team

1 PM, 1 EM, 1 FE, 1 UX

my role

Design lead

timeline

Dec 2023 - May 2024

overview

As the sole designer on the project, tackled high drop-off rates in the ingestion funnel by co-leading research initiatives to identify key issues. Collaborated closely with the PM and EM to brainstorm and develop solutions throughout the entire design process.

about company

Imply specializes in real-time data analytics using Apache Druid. The platform enables businesses to ingest, query, and visualize large data volumes instantly, offering high performance and swift data-driven decisions.

context

When I joined Imply, we had just started building our SaaS offering, Polaris, from the ground up. I led the data aspects, focusing on data ingestion, table management, and job monitoring. Over two years, we established a foundation aligned with our on-premises and hybrid offerings, prioritizing rapid feature development over user research.

Two years later, we came back to address the critical task of data ingestion, the backbone of our system.

problem statement

High percentage of users were not completing and dropping off ingestion flows at varying stages of the funnel.

user research

Conducted 1) Semi structured interviews with existing customers to understand their usage patterns, initial experiences, and overall perception of the product and 2) usability testing with non-users to understand the usability of the product – validating what works while uncovering issues

Goals

Insights

ideation

After pinpointing areas needing improvement, I collaborated with the EM and PM to prioritize our findings. To better assess the scope of the work, I initiated ideation sessions to explore potential solutions.

The key focus was on onboarding as ingestion was the first touchpoint user had with the product. It was critical that the ingestion process was a simple as possible. However, we also had to ensure that that experience was sufficient for the nth time experience so that we wouldn’t have two separate ingestion flows.

Keeping that in mind, we prioritized the following:

Display minimal and only necessary sources
Simplify and condense the steps involved
Enhance the conveyance of foundational concepts like schemas, primary timestamp, and syntax

final design

Simplified ingestion flow

All three steps into one screen -- select your source, we’ll set the table, and you’ll see data right away

Reduced number of displayed sources

No longer see unnecessary/irrelevant source tiles and only see the ones that have defined connections

Focused on one schema

hiding input schema and highlighting table schema

Lack of column declaration

Detach schema design from ingestion tme

Highlight primary timestamp

Make this critical step unskippable/unnoticeable

Clear syntax

A simple addition of word to clarify the language

data ingestion health app venmo case study