DEPLOYEDAI-era interview training
RankBootstrapping
XP 0 / 250
0
0%
Ready
Data Analyst & Analytics Interview · Data Visualization & BI · Lesson 2 of 2

Dashboard design & storytelling with data

7 min

A dashboard is a product with users. Design it around the decision it supports and the questions a stakeholder asks, not around every metric you can compute.

  • Know the audience & decision: an exec KPI board differs from an analyst's exploratory view. Lead with the answer.
  • Inverted pyramid: most important number top-left (people read Z-pattern), supporting detail below, exploration last.
  • Context on every metric: a number needs a comparison (vs target, prior period, or benchmark) to mean anything.
  • Reduce clutter: high data-ink ratio — remove gridlines, redundant legends, decoration. Consistent colors with meaning.
  • One message per view: if a chart needs a paragraph to explain, split it.

Storytelling

  • Structure a readout as situation → insight → recommendation, not a data dump.
  • Lead with the so-what: what should change because of this number?
  • Annotate charts with the takeaway ('conversion dipped after the April release') so it survives without you narrating.
Numbers need a so-what'Churn is 6%' is data. 'Churn rose to 6%, driven by month-one mobile users, so we should fix onboarding' is insight. Interviewers for analyst roles reward the leap from number to recommendation.
Dashboard as newspaperDesign like a newspaper: a bold headline (the KPI), a subhead (the trend), then the article (detail) for those who read on. Nobody should have to hunt for the lede.

◆ Lock it in

  • Design around the decision and audience; put the key number top-left.
  • Every metric needs context (target / prior period) and low clutter.
  • Tell a story: situation → insight → recommendation, leading with the so-what.
Feynman drill — say it out loudExplain how you'd turn a raw metric like '6% churn' into an insight a stakeholder can act on.
Step 1 rate your confidence · Step 2 pick your answer
What most separates an effective analyst dashboard from a raw metrics dump?
Step 1 — how sure are you?