Capafy logo Capafy
시작하기
Data Diving - expert-depth data analysis report

Data Diving - expert-depth data analysis report

by: Dominic_Huang
A spreadsheet full of numbers isn't an answer — it's raw material. McKinsey taught me to find the number that matters: start from the decision you're making, cut to the one or two insights that change it, then say what to do. This Skill brings that eye to your data — cleaning and structuring it, surfacing the patterns and outliers worth caring about, and writing findings as clear takeaways. Analysis that ends in a decision, not a dashboard nobody opens.
#분석#리서치#컨설팅
평점
4.6
판매
14
사용 방법
Capafy에서 실행
외부 앱에서도 사용 가능
퍼블리셔 제공
Claude Sonnet 4.6

"# Data Analysis
Five publication-grade report styles. Three-layer insight on every finding. The kind of analysis your data deserves — not a chart dump, a structured case with causes and calls.

#What you will get
Input your raw data file and you will get a structured data analysis report like this:
image.webp
image.webp
image.webp

When to Use / Best For

  • You need a report that looks like it came from a strategy firm: five editorial
    styles — Financial Times, McKinsey Consulting, The Economist, Goldman Sachs, Swiss/NZZ
    — each a complete design language with defined color systems, typography, and layout
    rules. Not a template skin; a publication standard.
  • Your data has a pattern you can see but can't explain: the three-layer protocol
    forces every observation through WHAT → WHY → So What. The skill won't let analysis
    stop at ""revenue dropped"" — it has to surface a hypothesis for why, and a decision for
    what to do.
  • You're delivering findings to decision-makers: report structure is conclusion-first
    by default — the answer in the first three lines, data and evidence in the body. Built
    for people who scan, not read.
  • Full-stack coverage without coordinating multiple tools: Excel analysis → insight
    narrative → HTML interactive report → PPTX slides. Brief once, get the whole stack.

How It Works

Five editorial report styles — rendered with ECharts, not screenshots:

Style Signature Best for
Financial Times Salmon-pink ground, serif analysis titles, 4px blue top bar Financial analysis, narrative reports
McKinsey Consulting Navy header, Exhibit numbering, conclusion-as-title Strategic analysis, framework decks
The Economist Red thin bar, editorial density, magazine rhythm Industry insight, opinion pieces
Goldman Sachs Rating badges, gold accent, dense data tables Financial modeling, valuation reports
Swiss / NZZ Black-white-gray-red, 72px display type, extreme contrast Data showcase, design-driven reports

Style randomly rotated per deliverable unless specified — so a weekly report series
doesn't look copy-pasted.

Three-layer analysis — non-negotiable on every finding:
Every major observation must reach WHY (what mechanism explains it). Every top-three
finding must reach So What (what decision it supports, with mechanism + evidence +
expected impact + effort). Reports that stop at Layer 1 (description) are flagged
as incomplete before they ship.

Pre-analysis assumption check: Before running any numbers, the skill surfaces
assumptions that, if wrong, corrupt the entire analysis — fiscal period boundaries,
business definitions (""active user"" = login-based or action-based?), GMV scope, data
lineage. Verified first, not discovered mid-report.

Chart rendering that doesn't break: All chart coordinates are computed from data
arrays via ECharts or D3 scale functions — never hardcoded pixel positions. Bar-chart
Y-axes start at zero. Stacked segments under 3% consolidate to ""Other."" SVG annotations
stay inside the viewBox. What you see in preview is what gets delivered.

Why This Skill / What Makes It Different

vs. BI dashboards (Tableau, Power BI, Looker): BI tools visualize the metrics you
configure. This skill reads the data cold — identifies which segments behave differently,
writes the hypothesis for each pattern, and produces a narrative the dashboard can't.

vs. ChatGPT or generic analysis prompts: No enforced depth — ChatGPT will summarize
Layer 1 and call it analysis. The three-layer protocol here is structural: Layer 2
(hypothesis) is required on every finding; Layer 3 (decision) is required on the top
three. Recommendations without mechanism and evidence don't ship.

vs. commissioning a report from a freelancer: Five editorial styles at publication
quality, data-driven charts that update cleanly when the data changes, and a structured
insight framework — delivered in the same session as the data upload.

Warm professionalism by design: The visual system avoids cyber-neon and dark
backgrounds in favor of warm palettes (cream, coral, dark gold) that read as
professional-yet-human — designed for projectors and client presentations, not
just screens."


View more