A beginner’s first session
- Read lesson one and make a prediction.
- Inspect the Course engagement sample.
- Calculate a rate and state what it cannot show.
- Record one independent example in your workbook.
A FREE, HANDS-ON DATA ACADEMY
Ask better questions. Inspect the evidence. Build an analysis you can explain—without handing your judgment to AI.
OBSERVE → QUESTION → TEST
A chart begins a conversation. Your reasoning makes it useful.Begin with what a row represents. Progress toward cleaning, comparing, visualizing and presenting a defensible finding. Every lesson includes two worked examples, answerable practice, a misconception check and an independent task.
Optional local saving. Datasets remain in memory unless you explicitly save an analysis project. No dataset contents, lesson answers or notes are sent to AI or analytics. Shared device? Leave saving off. Export before clearing browser data.
Open a synthetic sample or a local CSV. Preserve the original, inspect missing values, verify a calculation, then export your chart and analysis report.
Go to the workbench →Follow the sequence or revisit a particular skill. Lesson completion records practice and self-review, not certification. Workload is self-paced; repeat a task independently before deciding you can transfer it.
Turn a vague claim into an answerable question and identify the evidence needed.
Classify a field by what its values mean and choose an appropriate operation.
Identify the unit represented by a row and detect when a table mixes incompatible levels.
Document where a dataset came from and assess whether it can answer a specific question.
Explain how turning text into a table creates assumptions that need checking.
Assess fitness for purpose across completeness, validity, consistency, uniqueness and accuracy.
Read a tabular worksheet and separate raw observations from calculations and presentation.
Create a focused view of records while preserving row relationships and stating the selection rule.
Distinguish unknown values from observed zeros and document a defensible missing-data decision.
Distinguish repeated values, repeated keys and duplicate events before removing records.
Separate stored values from their display and prevent silent changes to dates, identifiers and percentages.
Build a derived variable with a stated formula, unit and policy for undefined results.
Calculate three summaries of center and explain why they can describe different aspects of the same data.
Describe spread alongside center and distinguish sample standard deviation from range.
Choose a meaningful denominator and distinguish percentage change from percentage-point change.
Describe a distribution and investigate unusual observations without automatically deleting them.
Explain how selection affects generalization and distinguish sample size from representativeness.
Interpret a paired association without claiming that one variable caused the other.
Match a visual encoding to the analytical question and the type of variables.
Build a category comparison with a meaningful aggregation and a zero baseline.
Represent change across an ordered variable without implying unsupported continuity or precision.
Inspect paired observations, clusters and unusual points before summarizing a numerical relationship.
Construct equal-width bins and explain how bin choices affect a distribution display.
Audit a chart for distorted scales, omitted context and unsupported interpretations.
Convert a broad request into a measurable question with a population, outcome and comparison.
Separate a testable expectation from the assumptions needed to interpret a result.
Compare group outcomes while checking sample sizes, composition and measurement consistency.
Distinguish a repeatable descriptive pattern from a coincidence or an artifact of selection.
Identify distinct sources of uncertainty and avoid reporting more precision than the evidence supports.
Write limitations that constrain a finding and guide a concrete next check.
Separate sales volume, price and revenue when explaining a commercial pattern.
Distinguish transactions from people and construct a behavior metric with a defined observation window.
Interpret session, event and conversion metrics without confusing activity with people or causal impact.
Compare campaign volume, rates and costs using aligned definitions and explicit attribution limits.
Compare planned and actual amounts while defining the sign and scope of a variance.
Balance throughput, time and quality while preserving the scope of operational measures.
Write a bounded analytical prompt that preserves your own first attempt and requests verifiable reasoning.
Create a concise data brief that communicates definitions and limitations without exposing unnecessary records.
Reproduce a generated result independently and trace discrepancies to data, definitions or arithmetic.
Identify claims that require fields, sources or study designs absent from the supplied evidence.
Use AI to critique a chart while verifying the axes, aggregation and underlying values yourself.
Design a review workflow in which people retain accountability for data, calculations and decisions.
Import and export tabular text while preserving quoted delimiters, newlines and field structure.
Translate a reusable formula into explicit inputs and distinguish relative references from fixed parameters.
Build a grouped summary while explaining the aggregation, missing-value policy and retained detail.
Apply documented transformations while preserving original values and checking their consequences.
Join a lookup table without multiplying records or hiding unmatched keys.
Package data, settings and reasoning so another person can inspect how a result was produced.
Design a small dashboard around decisions, scope and evidence rather than a collection of attractive charts.
Choose measures that represent the goal and anticipate ways a convenient metric could mislead.
Create chart interactions that preserve definitions and allow the plotted values to be checked.
Design and document filter rules so readers know exactly which records support a displayed result.
Write a concise finding that names the evidence, calculation, scope and limitation.
Connect a proposed action to verified evidence, trade-offs and a plan for checking the outcome.
Identify a plausible common cause and explain why an observed association may not isolate an effect.
Minimize sensitive information and consider how collection, grouping and recommendations affect people.
Read a treatment-control comparison while checking assignment, outcome definitions and missing results.
Distinguish a projection based on assumptions from an observed fact and evaluate forecasts against a baseline.
Carry one question through source inspection, cleaning, calculation, visualization and a reproducible report.
Demonstrate a complete, inspectable analysis on a learner-selected dataset with bounded conclusions.
All labs work without an AI connection. Their calculations are inspectable and their teaching scenarios are authored, not measured research outcomes.
Choose a brief, preserve your first attempt, and record the revisions and independent checks that support your conclusion.
Questions have authored answers and explanations. Open-ended notes and project evidence are reviewed using transparent criteria, not automatically graded as correct.
Work with rows, columns, data types and provenance. Clean missing values and duplicates with explicit decisions. Calculate percentages, summaries and grouped metrics. Build bar, line, scatter, histogram and appropriate pie charts. Examine selection bias, confounding and uncertainty. Check AI-generated claims against the actual data and communicate a recommendation with limitations.
Worked starting point: Campaign A has 10 conversions from 100 clicks; campaign B has 9 from 30. Their rates are 10% and 30%. A has more conversions, while B has the higher conversion-per-click rate. That alone does not establish which campaign caused better results or had the lower cost.
Every sample is synthetic teaching material. It is not evidence about actual students, customers, companies or public services.
No. Begin with the first module. All calculations, labs and project workspaces run in the browser without an AI API or account.
The workbench reads local files in your browser. It does not upload their contents. Save analysis project stores a copy in this browser when local saving is enabled. Downloads create files on your device.
Completion reflects practice and self-review. Demonstrate independent capability through the capstone and explain your limitations; a progress count is not a qualification.
Yes, within the CSV limits. Use data you are authorized to analyze and avoid confidential or identifying records in practice. The academy cannot determine whether your source is accurate or representative.
Attempt the analysis first. Give an AI assistant a non-sensitive schema or synthetic example, ask it to critique your reasoning, and independently verify every calculation and cited source. You remain responsible for the conclusion.
Vibe Learning framework · Reasoning & Evidence Lab · Learning Evidence Portfolio
Background references: RFC 4180 CSV conventions and the NIST statistical handbook. The academy’s datasets and worked scenarios are original teaching examples.