Design vs. Data: The Hidden Creative Battle

The Daily Interesting — Art & Design Edition

February 17, 2026

The idea

“Companies say they are ‘design‑driven,’ but designers are actually driven by dashboards filled with metrics like CSAT, NPS, CES, DAU, MAU… as if innovative ideas are hidden in spreadsheets, waiting to be extracted by data scientists.”
— Chuánqí Sun, “The Vanishing Designer” (via barnsworthburning.net) (barnsworthburning.net)

Who said it (and the context)

Chuánqí Sun is a designer‑researcher whose work spans prototyping, computational design, and AI tools. After several years at Microsoft shaping human‑AI interactions, Sun joined MIT’s Tangible Media Group to explore “Tangible AI.” His essay “The Vanishing Designer” argues that over‑reliance on metrics flattens taste, risk, and diversity in design practice. (chuanqisun.com)

A small expansion

Sun’s core move is not “anti‑data”; it’s anti‑substitution. When numbers replace judgment, designers converge on safe local maxima—the “heat death of design.” Metrics are essential as feedback, but dangerous as goalposts when they stand in for felt experience, context, or courage. Put differently: treat metrics like studio lights, not the director. Use them to see better—never to decide alone. (barnsworthburning.net)

Another lens on the idea

  • Goodhart’s Law warns: “When a measure becomes a target, it ceases to be a good measure.” In design, optimizing for CTR or NPS can decouple the proxy from what users actually value—clarity, dignity, meaning. Recent formal work even distinguishes “weak” versus “strong” Goodhart effects, where over‑optimizing a metric becomes useless or harmful for the true goal. (en.wikipedia.org)
  • Tricia Wang’s “thick data” complements big data with deep, qualitative human insight. Her point to product teams: dashboards give altitude; ethnography adds ground truth. Designing with both reduces the risk of mistaking what’s easy to count for what’s worth making. (medium.com)

Why this matters across art and design

  • Visual identity

    • Resist the “metric monoculture” of trend‑tracking. Define a small internal canon of reference work and a taste rubric that articulates why marks, type, and motion feel right—then use testing to refine edges, not choose the center. Keep the “director” (judgment) separate from the “lights” (metrics). (barnsworthburning.net)
  • Product & UX

    • Pair outcome metrics with narrative counterweights: for every quantitative KPI, capture one “thick” artifact (5‑minute screen‑recorded think‑aloud, a diary snippet, a support transcript). Review them together so teams can see when a metric is drifting from meaning. (medium.com)
  • Architecture

    • Performance models (daylight factors, EUI, occupancy) are crucial—but they can’t stand in for lived experience. Borrow Sun’s warning: keep room for judgment, material presence, and use‑phase observation walks after move‑in; treat post‑occupancy metrics as conversation starters, not verdicts. (barnsworthburning.net)
  • Studio art

    • Audience analytics (views, likes, dwell) are weak proxies for resonance. Guard your studio questions—form, rhythm, atmosphere—by scheduling critique that privileges description over scoring. If you measure, measure what helps the next decision, not what flatters the last post. (barnsworthburning.net)

Two quick drills

  1. Metric–Meaning Map
  • List the top 3 metrics your team uses (e.g., activation rate, NPS, dwell time).
  • For each, write the human outcome it’s supposed to represent (e.g., “trust,” “comprehension,” “delight”).
  • Add one qualitative “thick” check per metric (field note, 5‑line user quote, before/after sketch). Review pairs weekly; if the metric and the thick check disagree for 2+ sprints, change the design or change the metric. (medium.com)
  1. Courageous Variant
  • Ship one intentionally opinionated variant each cycle—a tasteful risk that a dashboard would never request.
  • Define success as “did we learn something we could not have learned from A/Bing minor tweaks?” Document the judgment used and the evidence gathered (quant + thick). Keep a reel of “useful risks” to prevent cultural drift toward local maxima. (barnsworthburning.net)

Thread to remember

Across recent editions we’ve explored perception, preparation, and living variability. Today’s lesson: data is a tool, not a taste. Keep judgment and courage in the loop so your work can aim beyond what the dashboard can see.


Sources

  • Central idea and excerpts: Chuánqí Sun, “The Vanishing Designer” (via barnsworthburning.net). (barnsworthburning.net)
  • Background on Chuánqí Sun: personal site and MIT Tangible Media Group profile. (chuanqisun.com)
  • Counterpoints/support:
    • Goodhart’s Law overview and formalizations. (en.wikipedia.org)
    • Tricia Wang on “thick data” (essay and talk descriptions). (medium.com)

Further reading on metric pitfalls: Jerry Z. Muller, “The Tyranny of Metrics.” (kirkusreviews.com)

—
If you want tomorrow’s edition to zoom into concrete workshop methods for blending big and thick data on a current project, I can tailor the drills to your stack and timelines.