You are a data analyst. Perform a rapid EDA on the following data.
Context:
- Format: [CSV/JSON/table/text]
- Domain: [ecommerce/marketing/finance/ops]
- Goal: [understand drivers of X]
Tasks:
1) Schema: list columns, inferred types, missingness .
2) Quality: duplicates, outliers (by [method if any]), anomalies.
3) Univariate: top stats for key numeric columns (mean, p50, p95, min/max).
4) Bivariate: 3 strongest correlations with [target] + cautions.
5) Quick insights: 5 bullet observations and 3 follow-up questions.
Output:
- Use a compact table for stats.
- Keep to <200 words + the table.
Data:
[Paste sample rows or attach file]

Role: You are a product analyst.
Scenario: [KPI] changed by [±X%] over [period]. Dataset fields: [list columns].
Goal: Find plausible drivers and recommend verification steps.
Tasks:
1) Decompose KPI by [segment, channel, geo, device, cohort]. Show top 5 movers.
2) Attribute drivers: volume vs. conversion vs. AOV (or relevant breakdown).
3) Hypothesize causes (internal vs. external) with evidence from the data.
4) Suggest 3 experiments or analyses to validate (e.g., holdout, diff-in-diff).
5) Produce a 5-bullet exec summary.
Output format:
- Table: segment → delta, contribution, confidence (low/med/high).
- Then bullets: hypotheses, validations, risks.
Data:
[Attach/describe data; or paste aggregates]

Task: Clean and normalize the following dataset for analysis.
Rules:
- Handle missing values: [impute with median/mode/drop] per column.
- Normalize categorical labels: map to canonical set [list].
- Parse dates to ISO 8601; extract [week, month, quarter].
- Outliers: Winsorize at [1, 99] percentiles for [columns].
- Output a clean schema + transformation steps.
Deliverables:
1) Mapping table(s).
2) Pseudocode for the pipeline (Python/pandas).
3) A compact diff of before → after.
Data sample:
[Paste 30–50 representative rows]

Role: Senior analytics engineer.
Warehouse: [BigQuery/Snowflake/Postgres].
Tables: [table_name(col1, col2, ...)], [table2].
Request:
“[Describe the question, time window, filters, and grain]”
Constraints:
- Use CTEs with clear names.
- Annotate assumptions as SQL comments.
- Include a validation query to spot row count mismatches.
- Return both the SQL and a 3-line rationale.

You are my spreadsheet formula assistant.
Goal: Create formulas to compute [metric] from columns [A, B, C].
Context: [Excel/Google Sheets]; locale: [US/EU decimal].
Tasks:
- Provide exact formulas with absolute/relative references.
- Include an arrayformula version for Sheets if relevant.
- Add a test row example to verify correctness.
Data header + 3 sample rows:
[Paste]

Role: Data visualization designer.
Audience: [execs/PMs/ops]; decision to support: [state it].
Create a charting plan:
1) Recommend 2–3 chart types with pros/cons for this dataset and goal.
2) Provide a Vega-Lite spec (or matplotlib/Plotly code) for the top choice.
3) Accessibility notes (colorblind-safe palette, annotations).
4) One-sentence narrative caption for each chart.
Data description:
[columns, units, time range, sample]

Context: We observed [pattern] in [metric] since [date].
Goal: Design a minimal, valid experiment.
Deliverables:
1) Hypotheses (H1/H0) with expected direction and effect size guess.
2) Experimental unit, randomization, and guardrail metrics.
3) Sample size and duration assumptions; note power trade-offs.
4) Analysis plan: test(s), segments, pre-registration checklist.
5) Risks and mitigation.

Role: Time-series analyst.
Data: [timestamp, metric, optional regressors].
Tasks:
1) Check stationarity and seasonality; suggest transformations.
2) Produce a short-term forecast (point + PI) using [model preference or "auto"].
3) Flag anomalies in the last [N] periods with severity.
4) Recommend alert thresholds to reduce false positives.
Output:
- Table: date, actual, forecast, PI_low, PI_high, anomaly_flag, severity.
- 5-line summary for non-technical stakeholders.

Task: Analyze customer feedback to extract actionable insights.
Inputs: [N] comments with fields [comment, rating, product, date].
Steps:
1) Cluster themes; label top 5.
2) Quote 1–2 representative comments per theme.
3) Quantify prevalence and sentiment per theme.
4) Recommend 3 actions with expected impact.
Output: A table + bullet summary. Keep under 180 words.
Data:
[Paste sample or attach]

Role: Chief of Staff producing an exec brief.
Content to summarize: [paste analysis, charts, or metrics].
Produce:
-  (3 bullets, action verbs).
- Key findings (5 bullets, with numbers).
- Risks/unknowns (3 bullets), Next steps (3 bullets, owners).
- One-sentence narrative for the board deck.
Style: Clear, non-technical, <160 words.

You are an analytics copilot.
Goal: Solve [analysis goal] using the following artifacts.
Artifacts:
- Data file(s): [link or pasted sample]
- Business context: [short brief]
- Constraints: [time, cost, accuracy]
Plan first (10–12 bullets):
- Identify inputs, assumptions, risks.
- Propose steps (EDA → transform → model/test → summarize), each with a deliverable.
- Ask 3 clarifying questions at the end.
Then wait for my confirmation before executing steps.

Add these guardrails to any analysis:
- Cite assumptions explicitly.
- If a calculation lacks enough data, return “insufficient evidence” with what’s missing.
- Provide a simple check: recompute [metric] two ways and compare.
- When summarizing, include a link/reference to the source data fields used.
- Ask: “What would falsify this conclusion?” and answer briefly.