Data exploration
Exploratory analysis: profiling a dataset, finding patterns and anomalies, analyst SQL.
Download all 83
- Reconcile two datasets
Reconciles two datasets that should agree, such as bank versus ledger or CRM versus billing, by matching records, listing mismatches and explaining likely causes. Use for month-end checks.
- Write a dataframe transformation
Writes pandas or polars code for a described transformation with built-in checks on row counts, nulls, key uniqueness and join cardinality. Use when reshaping, joining or aggregating data.
- Analyse competitor price data
Matches your products to competitors' equivalents, builds a price index by category and competitor, and flags where you are priced out of line on the items shoppers actually compare.
- Analyse contact centre performance data
Analyses contact centre data - volume by interval, handle time, abandonment, service level, repeat contacts and contact reasons - and recommends staffing alignment and process fixes.
- Analyse delivery performance
Analyses shipment data for on-time in-full rates and lateness by carrier, lane, warehouse and order type, separating where delays arise and ranking the causes worth fixing.
- Analyse whether promotions paid off
Analyses promotion data for incremental lift, cannibalisation, pull-forward and margin impact against a fair baseline, and says which to repeat. Use after a sale, coupon or discount campaign.
- Analyse nonprofit donor data
Analyses nonprofit donor data for retention, lapsed donors, gift size distribution and upgrade potential, with segment-level actions. Use for annual fundraising planning or a donor file review.
- Analyse an employee engagement survey
Analyses an employee engagement survey with group scores under minimum-group-size privacy rules, eNPS, comment themes and three priorities to act on. Use after an engagement or pulse survey closes.
- Analyse energy usage data
Analyses household or building energy data for baseload, daily and seasonal patterns, anomalies and the savings worth chasing, ranked by money. Use with smart meter exports or a year of bills.
- Analyse farm or orchard yield records
Analyses farm or orchard yield records by field, variety, input and season, separating weather and field effects from management, and designs fair on-farm trials for next season.
- Analyse a file too big for Excel
Analyses a CSV or Excel file too large for a spreadsheet using DuckDB, Polars or chunked pandas, with load checks, sanity counts and the summary tables you asked for.
- Analyse a hiring funnel
Analyses recruiting pipeline data for stage conversion, time to hire, source quality and drop-off, with fair comparisons by role. Use for a recruiting review or when roles take too long to fill.
- Analyse hotel occupancy, ADR and RevPAR
Analyses a hotel's occupancy, ADR and RevPAR by segment, channel, lead time and day of week, nets off commissions, and finds the pricing, mix and channel opportunities in the results.
- Analyse inventory and stock data
Analyses inventory and sales data for stock turns, days of cover, dead and slow stock and stockout risk, with a ranked action list. Use for a stock review, reorder planning or freeing up cash.
- Analyse labour cost against sales by hour
Compares staff hours and wage cost with sales by hour and weekday for a shop, cafe or restaurant to find over- and under-staffed periods, with a revised staffing pattern to test.
- Analyse location data
Analyses location data for stores, customers or deliveries to find catchments, density and distance patterns, with the method, code and mapping guidance. Use for site, coverage or delivery questions.
- Analyse loyalty programme data
Analyses a running loyalty programme's data for member versus non-member behaviour, incremental spend, redemption, points liability and tier health, avoiding the self-selection trap.
- Analyse machine downtime and OEE losses
Analyses machine downtime logs for causes, MTBF, MTTR and OEE availability, performance and quality losses, and ranks the fixes that recover the most production capacity.
- Compare marketing attribution models
Compares last-click, first-click, linear, position-based and data-driven attribution on supplied channel data and explains what each implies for budget. Use before moving marketing spend.
- Analyse price realisation and discount leakage
Analyses list versus net price by customer, sales rep and product with a price waterfall, to find where negotiated discounts and off-invoice deductions leak margin without a reason.
- Analyse process cycle times
Analyses process timestamps for lead time, wait versus work time, bottleneck steps and variability, from tickets, orders or case records. Use when a process feels slow.
- Analyse procurement spend
Builds a spend analysis by supplier and category from purchase or accounts payable data to find tail spend, off-contract buying, duplicate suppliers and consolidation savings.
- Analyse production defect and scrap data
Analyses inspection, scrap and rework records for defect rates, first-pass yield and cost of poor quality by product, line, shift and supplier, and ranks where to investigate first.
- Analyse repeat purchase behaviour
Analyses repeat purchase rate, time to second order and reorder intervals from order data, broken down by first product, channel and cohort, to show what turns buyers into regulars.
- Analyse sales performance
Analyses sales data by product, customer, region and time to find what drives revenue, seasonality, best and worst performers, and the actions worth taking. Use for a sales performance review.
- Analyse sales rep performance fairly
Analyses CRM and bookings data by sales rep - activity, win rate, deal size, cycle length, stage conversion and discounting - adjusted for territory and tenure, to target coaching.
- Analyse school attendance data
Analyses school attendance data for patterns by group, day, term and reason, identifies pupils who may need support without labelling them, and suggests interventions to test.
- Analyse sports performance data
Analyses player or team performance data for trends, consistency, strengths and weaknesses, with fair rate-based comparisons and small-sample warnings. Use for scouting or coaching reviews.
- Analyse stock and fund price history
Analyses price history for shares, funds or indices - total returns, volatility, drawdowns and correlations between holdings - with the maths shown and honest limits on what the past says.
- Analyse store footfall and conversion
Analyses shop footfall, conversion rate and basket size by hour, day and store to tell a traffic problem from a selling problem, and sizes the gain from each fix.
- Analyse survey results
Analyses quantitative survey responses with cleaning, tabulation, cross-tabs and optional weighting, and states the caveats about sample and response bias. Use before reporting survey numbers.
- Analyse user journey paths
Analyses the common paths users take before and after a key event in event data, collapses noise into readable steps, and finds where journeys loop, stall or end.
- Analyse website analytics
Analyses website analytics (GA4 or similar) for traffic sources, landing pages, engagement and conversion, flags tracking problems first and gives prioritised actions. Use as a marketer or site owner.
- Analyse workforce data
Analyses HR data for headcount movement, attrition, tenure and representation under minimum-group-size privacy rules, and flags patterns worth a closer look. Use for a workforce review or board pack.
- Anonymise a dataset before sharing
Plans anonymisation or pseudonymisation of a dataset before sharing, classifying identifiers, choosing techniques and assessing re-identification and residual risk. Use before data leaves your team.
- Answer a question with SQL
Turns a business question and a schema into an analytical SQL query, states the assumptions behind it and explains how to read the result. Use when you know the question but not the query.
- Build an analysis dataset
Combines several sources into one analysis table at a stated grain, such as one row per customer, with point-in-time features and checks for leakage, duplication and coverage.
- Build a cohort retention analysis
Builds a cohort retention analysis from event data (cohort definition, query or code, the retention triangle) and explains how to read it. Use to see whether newer customers stick around better.
- Build a date spine for metrics
Builds a calendar or date spine so daily, weekly or monthly metrics show zero-activity periods, with joins that neither drop periods nor duplicate rows, in your SQL dialect or pandas.
- Calculate active users in SQL
Computes DAU, WAU, MAU, rolling 7- and 28-day active users and stickiness from an event table, with deduplicated users, correct trailing windows and test traffic excluded.
- Calculate customer lifetime value from transactions
Calculates historical and projected customer lifetime value from transaction data by cohort and segment, on gross margin, with young cohorts and the formula's traps handled honestly.
- Calculate MRR movements
Builds a monthly MRR bridge from subscription or billing data (new, expansion, contraction, churn, reactivation) with code, and reconciles opening plus movements to closing MRR.
- Check tables before joining them
Checks two tables before you join them (key uniqueness, orphan rows, type and format mismatches, expected row counts) so totals do not inflate from fan-out or vanish from failed matches.
- Classify text records
Classifies free-text records such as tickets, feedback or expenses into a given set of categories, with a confidence level and an explicit Other bucket, and returns a table.
- Clean a raw survey export with a decision log
Cleans a raw survey export in the project files with reproducible code, checking speeders, straight-lining, duplicates and attention checks, recoding scales and logging every decision.
- Design a clean data collection form
Designs a form or sheet that collects data cleanly at the source, with field types, validation, IDs, required fields and a test entry. Use before launching a form whose answers you will analyse.
- Data analyst
Acts as a data analyst who starts from the decision, sanity-checks data before trusting it and states uncertainty plainly. Use as a standing analyst persona or subagent for data questions.
- Data journalist
Data journalist who checks where a dataset came from before trusting it, distrusts round numbers, finds the human story in the figures and explains methods and limits plainly to readers.
- Data scientist
Acts as a data scientist who frames the decision first, uses the simplest valid method, validates out of sample and communicates uncertainty plainly. Use for modelling, prediction and experiment work.
- Clean a dataset with a scripted, auditable pipeline
Cleans a dataset with a repeatable script in gated steps, profiling, proposing rules, applying and validating them, and exporting with an auditable cleaning log. Use for data that will be reused.
- Decompose a revenue change
Breaks a revenue or sales change into price, volume and mix effects, and into new, lost and retained customers, with the arithmetic shown and reconciled. Use to explain why revenue moved.
- Decompose a time series into trend and seasonality
Decomposes a time series into trend, seasonality and residual, explains each in plain words and shows what a fair year-on-year comparison looks like. Use before reading too much into a monthly change.
- Deduplicate messy records
Plans and writes matching logic to deduplicate people, companies or products across messy records, with normalisation, blocking, fuzzy thresholds, merge rules and a review queue. Use for CRM cleanup.
- Detect anomalies in data
Finds anomalies in a metric or dataset with methods that fit its shape (thresholds, seasonality, robust z-scores), ranks them, and separates data errors from real events. Use when monitoring data.
- Practise pandas in a simulated session
Simulates a pandas session on a dataset you describe, printing DataFrame output, aggregations and errors faithfully so analysts practise cleaning and reshaping without setup.
- Practise R in a simulated console
Simulates an R console with a small synthetic data frame, printing results, summaries, warnings and errors as R would, for learners practising base R or tidyverse verbs.
- Enrich your data with public datasets
Joins public data such as census, income, deprivation or region lookups to your own records at the right geography and date, and reports match rates and the limits of what the joined data can show.
- Explore a dataset
Runs a first-pass exploratory analysis of a dataset (column profiles, missingness, distributions, outliers) and lists the questions worth asking next. Use when you get new data.
- Explore a free-text column
Profiles a free-text column in a dataset (blanks, lengths, languages, common words and phrases, samples by segment) with code, so you can decide how to categorise or analyse it.
- Extract fields from documents into a table
Extracts named fields such as dates, amounts, names and IDs from emails, invoices or letters into a table, leaving blanks where a value is absent rather than guessing. Use to turn paperwork into data.
- Find churn drivers
Finds which behaviours and attributes predict churn in customer data, simple comparisons first and a model only if justified, with an action and a test per driver. Use at subscription businesses.
- Find the story in a public dataset
Finds the story in a public dataset by checking provenance and definitions, computing rates rather than raw counts, and testing the headline before it is published. Use for data journalism or reports.
- Find streaks and gaps in SQL
Solves gaps-and-islands problems in SQL, such as consecutive-day streaks, sessions built from events, outage periods and missing sequence numbers, with the logic explained step by step.
- Fix dates and time zones in a dataset
Standardises mixed date formats, UTC versus local timestamps and daylight-saving shifts in a dataset, and shows how each fix moves records between days so daily counts change.
- Geocode an address list
Cleans an address list, picks a suitable geocoding service for its country and volume, writes the geocoding code with match-quality scoring, and lists the rows that need a manual fix.
- Map an unfamiliar database for analysis
Gives the read-only queries to map a database or warehouse you have just been given access to (tables, row counts, grain, keys, freshness, joins), then writes a short field guide for analysis.
- Pivot rows to columns in SQL
Turns rows into columns, or columns back into rows, in SQL with conditional aggregation or PIVOT and UNPIVOT, and handles a column list that changes over time for your dialect.
- Solve a mystery by querying a database
Runs an original detective case solved by querying a fictional database of witnesses, access logs and transactions, answering every query consistently until the player names the culprit with evidence.
- Query the GA4 BigQuery export
Writes BigQuery SQL on the GA4 export tables for event parameters, sessions, users and conversions, and explains why the numbers differ from what the GA4 reports show.
- Query JSON columns in SQL
Extracts fields and arrays from JSON columns in Postgres, BigQuery, Snowflake and similar warehouses for analysis, flattening arrays without accidental row fan-out or lost rows.
- Read a messy Excel file into pandas
Loads report-style Excel workbooks into tidy pandas DataFrames, handling multi-row headers, merged cells, several sheets, notes and totals rows, with checks on exactly what was read.
- Review analytical SQL
Reviews an analytical SQL query for logic errors that give wrong numbers, such as join fan-out, misplaced filters, NULLs, double counting and date or time-zone boundaries. Use before sharing results.
- Run an analysis loop together, one query at a time
Works through a business question in a loop where the assistant proposes the next query or chart, the user runs it and pastes the result, and the assistant interprets it and picks the next step.
- Run a market basket analysis
Runs market basket analysis on transactions to find products bought together, explains support, confidence and lift, and suggests bundles or placement to test. Use for retail and e-commerce.
- Run a Pareto (80/20) analysis
Runs a Pareto analysis on products, customers, defects or causes, with the cumulative table, chart instructions and which vital few to act on. Use to find where effort will pay off most.
- Screen transactions for irregularities
Runs standard audit analytics tests on payments or journal entries (duplicates, round amounts, split payments, odd timing, Benford's law) and lists items to follow up, not accusations.
- Screen variables for relationships with an outcome
Screens which variables relate to an outcome using correlations, group comparisons and plots, flagging confounding, nonlinearity, leakage and chance findings before anyone acts on them.
- Segment customers
Proposes and builds a customer segmentation (RFM, rules or clustering) with interpretable segment profiles and a suggested action for each. Use to target retention, pricing or marketing work.
- Translate a spreadsheet workflow to pandas
Translates a spreadsheet workflow of filters, lookups, pivots and formulas into a pandas script that produces the same outputs, with checks that totals match. Use to automate a manual routine.
- Translate SQL between dialects
Translates an analytical SQL query between Postgres, MySQL, SQL Server, BigQuery, Snowflake and others, and flags every date, string, NULL, division and rounding difference that could change results.
- Write an analysis plan
Writes an analysis plan before touching data, covering the decision, questions, metrics, data, method, comparisons, pitfalls and the result that would change the decision. Use when scoping a request.
- Write a data request brief
Turns a vague stakeholder ask into a clear data request with the decision, exact metric definitions, filters, time range, format and deadline, plus open questions. Use when a data ask arrives vague.
- Write a SQL window function query
Writes SQL window functions for running totals, rankings, period-over-period change with LAG and latest-record-per-key, and explains the partition, order and frame choices.
Not: building pipelines or tuning queries (data).