Synonyms: data analytics, data analysis, applied statistics, quantitative analysis
Do not index
Definition
Data science is an interdisciplinary field that pulls insight out of data using statistics, programming, and domain knowledge. Drew Conway's 2010 Venn diagram is the classic map: the overlap of coding skills, math and stats, and expertise in the actual problem. For UX designers, data science answers the "what," which flows people take, where they drop off, what the A/B test says, while user research answers the "why."
Use cases
On its own, data science tells you a screen loses 40% of users. It won't tell you why. Pair it with UX research and you get the number and the reason, which is the combination that actually changes a design.
The what without the why: Analytics show checkout abandonment spikes at step 3. Data science flags it, then a usability test reveals the surprise shipping cost behind it.
Personalization: A recommendation model surfaces products, and UX decides how and when to show them so it helps people instead of creeping them out.
Wrong metric: A model chases time-on-page, so the team celebrates "engagement" while users are actually lost and hunting. UX catches that the metric is measuring confusion.
How it's used in practice
Combine quant and qual. Use behavioral data to find where the problem is, then research to learn why it happens.
Co-define metrics with data scientists. Agree on what "success" means before building, so the model doesn't chase the wrong thing.
Plan experiments together. Design A/B tests that are statistically valid and answer a real design question.
Learn enough to be dangerous. Basic SQL, a BI tool, and how to read a funnel let you explore data without waiting on anyone.
Keep a human in the loop. Models flag patterns, people add context and catch when the data misleads.
🪄
Pro-tip: Data tells you where users struggle, never why. If someone hands you a chart and a conclusion in the same breath, ask what qualitative evidence backs the "why."
Challenges & limitations
Correlation isn't cause: Two things moving together is easy to misread, and a false read ships the wrong fix.
Garbage in, garbage out: A model is only as good as the tracking behind it, and event data is often messy or incomplete.
Ethics and privacy: More data means more responsibility. GDPR, CCPA, and basic decency limit what you should collect and how you use it.
