Skip to content

Latest commit

 

History

History
20 lines (12 loc) · 2.23 KB

Challenges.md

File metadata and controls

20 lines (12 loc) · 2.23 KB

Why data visualization is inherently hard to test

The complexity in our stars

Faster is not necessarily better

Visualization is outcome driven. Very rarely are our users making visceral decisions based on a vis, and while we care about the affective value of a vis, we don’t want users making emotional decisions (as we would on an e-commerce site where we might measure how quickly a user clicks on an add to cart). Visualizations are often designed to help users uncover insights from their data. This process requires exploration that cannot be measured effectively through time. We don’t want to penalize a vis because it makes a user think. Similarly, number of clicks is not a good measure when we want to encourage the user to explore and play with the data.

Too many moving parts

While for some usability studies, there may be an obvious and limited set of interface elements that should be tested. In many cases, however, the system is complex, new to the researcher, or the list of candidate features for testing is long. In many cases, visualizations are exploratory in nature. This would require a fully functional application that the user can dynamically interact with. Creating a prototype that simulates a fully functional application is challenging. The following image demonstrates how quickly typographic elements can scale up in a dashboard, moving well beyond the typical h1, h2, h3, and so on, elements used in websites.

Insights are hard to measure

The ability to draw insights is key for data visualizations. If you want to measure how many insights are derived using one visualization versus another, how do you define what constitutes an insight in the first place? Also, the definition of an insight can vary from user to user.

Datum Ipsum

It is time-consuming to produce hi-fidelity looks-like and works-like prototypes with real-data backing them. While we do attempt to use meaningful placeholder data, static mockups can only show so much. We also face the challenge of users not being able to understand or relate to a dashboard with placeholder data.

Tools- http://datumipsum.com/