Simplified dataset modeling for non-technical researchers

Replaced a complicated data-modeling experience with a flexible interface that stays simple until complexity is needed, enabling full deprecation of a legacy tool and shaping how every user on the platform manages their datasets.

Role: Product designer

Team: Product manager, 2 engineering teams

Tools: Figma, Subframe, Qualtrics moderated testing

Problem(s)

Treating every user like a data engineer is expensive and inefficient

#1 — Data modeling is complicated. All data in Qualtrics must be processed in a data modeling tool before customers can use analysis or visualization tools. However, 90% of datasets made in Qualtrics are simple (consist of only one or multiple surveys combined) and should require no data engineering knowledge.

#2 — Novice users are lost and expensive. Most users are not technical, so they are alienated by the complex data modeling requirements and require hundreds of thousands in annual support costs.

#3 — Split user experience and double engineering costs. There are two data modeling tools with separate interfaces and functions, meaning users who learn one tool first struggle to learn the other.

solution

Power on demand — automate as much decision making as possible

Identify user intent

Use existing recommendation logic to detect the dataset a user intends to build, then automatically take users to dataset setup.

Step-by-step setup

Each dataset type follows its own step-by-step setup flow.

Flexible canvas: simple environment for simple tasks

For more complex data transformations, users can toggle to the full data modeling canvas.

Customer impact

Dataset creation cut to 4 steps for 90% of use cases

I redesigned the logic for how 90% of users build datasets. I took a high-friction, technical process (dataset creation) and turned it into a simple user experience that is mostly automated.

Company impact

Second-to-last step for deprecation of legacy tool by funneling all new dataset creation through this redesigned data modeler

The last step will be migrating al existing datasets into the new modeling tool. That is an engineering priority that is scheduled for December 2026.

Reach me at kellyluo.kluo@gmail.com or in Seattle, WA

Reach me at kellyluo.kluo@gmail.com, on LinkedIn, or in Seattle, WA