Data Analysis And Spreadsheets Prompts

INPUTS
[dataset]; [data dictionary]; [analysis goal]; [known collection issues]

PROMPT
Act as a data analyst and audit the dataset before interpreting it. Check missing values, duplicates, inconsistent units, impossible values, date formats, categories, and whether rows represent comparable entities. Explain how each issue could affect the intended analysis. Provide a prioritized quality report and cleaning plan that preserves the original data. Do not delete ambiguous records without documenting the rule. End with validation checks, unresolved questions for the data owner, and a statement of which analyses are currently defensible.

INPUTS
[dataset]; [business question]; [variables]; [time period]; [constraints]

PROMPT
Create an exploratory analysis plan linked to the business question. Define the unit of analysis, key measures, relevant segments, and comparisons. Start with data quality checks, then propose summaries and visualizations that can reveal useful patterns. Explain likely confounders and limits of inference. Provide a sequence of analyses with the decision each supports. If no dataset is available, produce a plan rather than invented results. End with criteria for deciding whether a pattern deserves further investigation or is too weak to act on.

INPUTS
[spreadsheet app]; [cell layout]; [desired calculation]; [sample rows]; [locale]

PROMPT
Write a spreadsheet formula for the requested calculation using the specified app and locale. Explain the references and assumptions, including how blanks, errors, duplicates, and date boundaries should behave. Provide the formula, a plain-language explanation, and expected results for the supplied sample rows. Offer a simpler alternative if it improves maintainability. Do not assume unsupported functions are available. End with three meaningful edge-case checks and identify any detail about the sheet layout that must be clarified before the formula can be used reliably.

INPUTS
[business goals]; [available fields]; [reporting periods]; [stakeholders]

PROMPT
Create a KPI definition sheet that makes calculations consistent across the team. For each proposed metric, specify purpose, formula, numerator, denominator, unit, inclusion rules, time period, source, owner, and interpretation. Distinguish counts, rates, and averages. Flag metrics that can improve while the underlying outcome worsens. Provide examples using clearly labeled illustrative values only if needed. End with reconciliation checks and the open questions that must be resolved before the metrics are used in performance reporting.

INPUTS
[dataset]; [message to communicate]; [audience]; [output format]

PROMPT
Recommend charts that communicate the intended message accurately. Match each proposed chart to a specific comparison, trend, distribution, or relationship. Specify axes, units, ordering, labels, and whether a baseline or uncertainty display is needed. Explain misleading alternatives to avoid and how missing data should appear. Provide a chart specification and a short caption grounded in the supplied data. Do not claim a relationship is causal. End with an accessibility checklist covering contrast, direct labeling, and information that should also be available in text.

INPUTS
[customer or user data]; [cohort event]; [outcome]; [observation window]

PROMPT
Design and, when data permits, perform a cohort analysis. Define cohort membership, the starting event, observation periods, denominators, and treatment of incomplete follow-up. Compare like-for-like windows and flag small cohorts. Produce a cohort table, an interpretation, and alternative explanations for differences. Distinguish retention, activity, and revenue measures rather than mixing them. If inputs are incomplete, provide the calculation specification and missing fields. End with the next question the analysis raises and a check against misleading recent-cohort comparisons.

INPUTS
[experiment design]; [sample sizes]; [outcomes]; [randomization]; [duration]

PROMPT
Interpret the experiment in light of its actual design. Check the assignment method, primary metric, sample sizes, stopping rule, missing data, and whether groups are comparable. Report effect size and uncertainty only using appropriate calculations supported by the inputs. Distinguish statistical evidence from practical importance. Present a findings summary, limitations, and a recommendation tied to the original goal. Do not manufacture significance or call a nonrandom comparison causal. End with whether the result supports rollout, further testing, or no change.

INPUTS
[database dialect]; [schema]; [business question]; [sample records]; [constraints]

PROMPT
Write a read-only SQL query that answers the business question. Clarify the row grain, joins, filters, date boundaries, and aggregation before presenting the query. Handle null values and possible duplicate joins explicitly. Explain the output columns and provide checks to reconcile totals against simpler queries. Use only tables and fields in the supplied schema. Identify assumptions and performance considerations without inventing indexes. End with sample expected output based on supplied records and any unresolved business definition affecting the result.

INPUTS
[analysis results]; [charts]; [business question]; [audience]; [decision]

PROMPT
Translate the analysis into a leadership narrative. Lead with the most decision-relevant finding, explain the supporting data, and state what it does and does not establish. Preserve units, denominators, time periods, and uncertainty. Distinguish observation, interpretation, and recommendation. Provide a concise briefing with suggested chart captions and one actionable next step. Avoid implying that every visible change is meaningful. End with the strongest alternative explanation and the information needed before making a larger commitment based on the results.

INPUTS
[analysis workflow]; [data sources]; [code or formulas]; [deliverables]

PROMPT
Review the workflow for reproducibility and auditability. Identify how source data, transformations, assumptions, calculations, and outputs are recorded. Recommend a minimal structure that lets another analyst reproduce the result without hidden manual steps. Include naming conventions, version notes, validation checks, and a change log. Flag places where copying, filtering, or overwriting data could alter conclusions unnoticed. Deliver a checklist and a handoff template. End with the three highest-priority improvements and a practical test of whether the final output can be recreated.