When data feels like noise: turn spreadsheets into strategic wins
Marketing managers are flooded with dashboards, CSV exports, and vanity metrics. The hard part isn’t collecting data — it’s rapidly turning messy numbers into clear actions you can present to stakeholders and execute this quarter. AI prompts let you automate the grunt work of analysis, surface the most important insights, and draft prioritized recommendations you can act on the same day.
How to use these prompts
Each prompt below is written so you can copy and paste into your preferred AI assistant (large language model or analytics assistant). Start by feeding the model a short context: what dataset you have, the time range, and any definitions (e.g., what counts as a conversion). Then paste the prompt and attach a CSV, table, or a summary of columns when possible. After you get output, ask the model to show sources for any claim or to produce SQL/Python/Excel formulas to replicate the calculation.
Prompts that cut analysis time in half (8 ready-for-use prompts)
1. Data quality audit and anomaly detection
Use this prompt the moment you receive a new dataset. It forces the model to check for problems before you trust any results.
I'm a marketing manager analyzing a dataset. Columns: date, channel, sessions, users, conversions, revenue, cost. Check the dataset for missing values, duplicate rows, out-of-range values, inconsistent date formats, and sudden volume spikes or drops. List the top 5 data quality issues, explain why they matter, and give step-by-step remediation steps and sample SQL queries to fix each problem.
Actionable advice: Run this first. Fix the top 1–2 issues (duplicates, incorrect date parsing) before running performance analyses. Use the SQL snippets the model provides in a staging environment and re-run the audit.
2. Executive KPI summary and one-slide narrative
When you need a concise summary for stakeholders, this prompt produces a crisp narrative and bullets you can paste into a slide.
Given these KPIs for the last 30 days (sessions: [value], conversions: [value], conversion_rate: [value], revenue: [value], cost: [value]), produce a one-paragraph executive summary highlighting performance vs prior period, three supporting bullets explaining why performance moved, and two prioritized recommendations (with expected impact and confidence level) to improve results next month.
Actionable advice: Replace bracketed values with your numbers. Use the recommendations directly in your weekly standup. Ask the model to convert “expected impact” into a forecasted monetary range if you need justification for budget changes.
3. Channel attribution and budget reallocation
Use this to identify which channels are delivering net positive ROI and where to reallocate spend.
Here are channel-level figures for the last quarter: channel, spend, clicks, conversions, revenue. Calculate CPL, CAC (if applicable), ROAS, and ROI per channel. Rank channels by ROI and provide a recommended reallocation of 10% of total budget to maximize conversions while preserving diversity (no single channel >40%). Include assumptions and sensitivity analysis for a ±10% change in conversion rate.
Actionable advice: Validate the model’s assumptions against your tracking setup (attribution window, multi-touch vs last-click). Use the reallocation recommendation as a starting point and test changes with a 4-week experiment instead of a full-scale shift.
4. Cohort analysis and customer LTV segmentation
Turn raw user history into actionable segments for retention campaigns.
I have user-level data with signup_date, first_purchase_date, total_revenue, number_of_purchases. Group users into weekly cohorts by signup_date and produce: retention curves (week 1–12), average LTV at week 12 per cohort, and three user segments (high, medium, low LTV) with recommended retention tactics for each. Provide sample SQL to compute cohorts and retention rates.
Actionable advice: Use the LTV segments to prioritize CRM campaigns: allocate 60% of retention budget to high-LTV, 30% to medium, and 10% to low with re-engagement tests. Ask the model to generate subject lines and offer structures for each segment.
5. A/B test result interpretation and next steps
When an A/B test finishes, get a statistically sound interpretation and concrete next steps.
Test summary: variant A conversion 3.2% (n=12000), variant B conversion 3.7% (n=11800). Average order value equal. Determine statistical significance, compute required sample size for 90% power and 1% difference detection, and produce a clear recommendation: stop/continue/rollout. Provide a justification, risk factors, and three follow-up tests to run if we roll out B.
Actionable advice: If the model recommends rollout, prepare an implementation checklist (QA, tracking, rollback plan). If the sample size is too small, extend the test with a minimum additional sample size specified by the model.
6. Forecasting next quarter's conversions
Use this to create a realistic forecast that accounts for seasonality and recent trends.
Using monthly conversions for the past 24 months: [list values], create a 3-month forecast that accounts for seasonality, trend, and a planned 10% marketing budget increase starting next month. Provide baseline, optimistic, and conservative forecasts with percentiles, and list key assumptions and sensitivity to traffic and conversion rate changes.
Actionable advice: Use the optimistic/conservative scenarios to set stretch and safe targets. Request the model to output a simple table you can paste into a report and to produce the underlying formulae so finance can validate.
7. Cross-channel funnel leakage analysis
Pinpoint where prospects drop out and recommend fixes for each funnel stage.
I have funnel counts by stage for users from paid search vs organic: visit → product_view → add_to_cart → checkout_start → purchase. Compare funnels, calculate drop-off rates and stage-level conversion uplift if each lagging channel matched the best-performing channel. Recommend three actionable experiments to reduce drop-offs at the two worst stages with expected conversion impact.
Actionable advice: Prioritize fixes that require minimal engineering (e.g., messaging or CTA changes) to get rapid wins. Combine funnel fixes with targeted retargeting to recapture leavers.
8. Visualization and dashboard specification for stakeholders
Not all visualizations are equal. Use this to get an optimized dashboard spec.
I'm building a stakeholder dashboard focused on acquisition and ROI. Given KPIs (sessions, conversion_rate, cost, revenue, ROAS) and audiences (VP Marketing, Growth Lead, Analysts), recommend 6 charts with reasoning, a layout priority (top-row for executives), and data refresh cadence. Provide labels, filters, and a short SQL/Looker/Power BI spec for each chart.
Actionable advice: Implement the top-row charts first for executive meetings. Use the dashboard spec to brief your BI team and shorten build time.
Two bonus prompts you can paste directly (copy-paste-ready)
These are additional practical prompts for common needs. Paste into your AI assistant and attach sample data or column names.
Bonus prompt 1 — executive-ready incident report: "I detected a 25% drop in conversions on 2026-02-10 compared to the previous 7-day average. Given columns: date, channel, conversion_rate, sessions, revenue, list the 6 most likely causes, prioritize them by probability, and propose a 48-hour action plan to diagnose and remediate.