Write a data dictionary with real examples
Makes units, nulls and ownership explicit.
[DATASET COLUMNS][SAMPLE ROWS][SOURCE OWNER]
Topic · 26 prompts
Data & analysis
Explore data & analysis prompts with clear inputs, complete instructions and ways to adapt the result. Explore “Write a data dictionary with real examples” and “Design a manual-entry accuracy sample”.
Design a manual-entry accuracy sample
Checks quality without inspecting every record.
[ENTRY PROCESS][RECORD COUNT][RECORD IDS][RISK LEVEL]
Round a public table without breaking totals
Shows how display choices affect interpretation.
[TABLE VALUES][DISPLAY PRECISION][AUDIENCE]
Validate event order in a funnel
Detects impossible paths before conversion analysis.
[EVENT SCHEMA][FUNNEL STEPS][SAMPLE EVENTS]
Investigate an outlier without deleting it
Uses context and measurement checks before exclusion.
[DATA SAMPLE][EXPECTED RANGE][MEASUREMENT PROCESS]
Explain unmatched records after a join
Keeps missing matches visible instead of dropping them.
[LEFT SCHEMA][RIGHT SCHEMA][MATCH KEY]
Choose a safe map resolution
Balances local insight against re-identification risk.
[LOCATION DATA][USE CASE][AUDIENCE]
Sanity-check survey weights
Finds extreme influence and unsupported population claims.
[WEIGHT VALUES][SAMPLE SUMMARY][TARGET TOTALS]
Document a category recode
Shows what changed when labels were consolidated.
[OLD CATEGORIES][NEW CATEGORIES][SAMPLE ROWS]
Find overlapping reporting periods
Prevents double counting across files.
[FILE SCHEMAS][DATE FIELDS][REPORTING RULE]
Review possible duplicate customers without auto-merging
Separates exact matches, plausible matches and risky false positives.
[CUSTOMER RECORDS][MATCH RULES]
Compare geographic data across changed boundaries
Exposes non-comparable areas before producing a trend map.
[OLD BOUNDARIES][NEW BOUNDARIES][MEASURE TABLE]
Code open-text feedback without erasing dissent
Creates a transparent themebook with counts and counterexamples.
[FEEDBACK RESPONSES][RESEARCH QUESTION]
Find where a forecast fails, not just its average error
Breaks forecast performance into meaningful segments and horizons.
[PREDICTIONS][ACTUALS][BUSINESS COSTS]
Check an experiment's assignment before reading outcomes
Tests randomisation balance, exposure and exclusions independently of results.
[ASSIGNMENT LOG][EXPOSURE LOG][EXPERIMENT PLAN]
Audit a dataset for mixed units before analysis
Finds unit mismatches and proposes reversible conversions.
[DATA DICTIONARY][ROWS]
Report a small table without exposing individuals
Designs suppression and aggregation rules for sparse data.
[TABLE DESIGN][POPULATION][DISCLOSURE RULES]
Define a retention cohort with the right clock
Separates signup, first use and return events before calculating retention.
[EVENT SCHEMA][TIMEZONE][RETENTION QUESTION]
Investigate missing data as a pattern
Treats blanks as a research question, not automatic zeros.
[DATASET DESCRIPTION]
Join tables without inflating counts
Designs a safe join and cardinality audit.
[TABLE A][TABLE B][DATABASE]
Write results from a tiny sample honestly
Makes limited tests useful without overstating confidence.
[TEST DATA][SAMPLE SIZE][DECISION]
Investigate an unexpected metric jump
Ranks possible causes before telling a story.
[METRIC SERIES][KNOWN EVENTS]
Map what a survey cannot tell you
Shows who was reached, who replied and what remains unknown.
[SURVEY QUESTIONS][RECRUITMENT METHOD]
Find the denominator hiding in a chart
Stops misleading percentages before they spread.
[CHART OR TABLE]