Data Analysis Techniques Training Course

Participants can assess data sources, structures and fitness for a stated analytical purpose.

• Structured, semi-structured and unstructured data classified by analytical use
• Entity, attribute and record relationships mapped from source system extracts
• Primary keys, foreign keys and grain checked before joining tables
• Transactional, master and reference data distinguished within operational systems
• Source lineage records collection method, owner, refresh cycle and access restrictions
• Fitness-for-purpose assessment compares available fields against analytical requirements

Participants can profile a dataset and document material data quality risks.

• Data profiling measures completeness, uniqueness, validity, consistency and timeliness
• Frequency tables expose unexpected categories, blanks and coding variations
• Duplicate records identified using keys, matching rules and business logic
• Missing values classified as random, systematic or structurally absent
• Range checks compare numerical values against plausible operational limits
• Quality issue logs record impact, treatment, owner and unresolved risk

Participants can apply governance and ethical checks before using data for analysis.

• Data minimisation limits collection and use to stated analytical purposes
• Personal data fields classified using sensitivity and access requirements
• Consent, purpose limitation and retention requirements assessed before analysis
• Bias sources examined in collection, labelling, sampling and historical decisions
• Access controls separate raw data, prepared datasets and published outputs
• Reproducibility records preserve source versions, transformations and analytical decisions

Participants can write SQL queries that retrieve relevant, traceable analysis datasets.

• SELECT, FROM and WHERE clauses retrieve records using explicit criteria
• Boolean conditions filter transactions by dates, status codes and thresholds
• INNER JOIN and LEFT JOIN combine tables through validated keys
• GROUP BY summaries calculate counts, totals and averages by category
• CASE expressions standardise categories and create analytical flags
• Common table expressions organise multi-step extraction logic for review

Participants can prepare analysis-ready datasets using documented Python transformations.

• Pandas dataframes load, inspect and reshape tabular source files
• Column types converted before calculations, joins and date-based analysis
• Missing values treated through deletion, imputation or explicit category assignment
• String cleaning standardises spaces, case, punctuation and coded values
• Merge operations combine validated datasets while preserving expected record counts
• Transformation scripts document each preparation step for later reproduction

Participants can use exploratory analysis to identify patterns requiring further investigation.

• Distribution plots reveal skewness, clustering, gaps and extreme observations
• Cross-tabulations compare customer, service or product groups systematically
• Time series plots distinguish trend, seasonality and irregular variation
• Segment comparisons identify material differences between defined populations
• Correlation matrices screen relationships requiring business and statistical review
• Exploration notes separate observed facts from preliminary interpretations

Participants can calculate and interpret descriptive statistics for operational datasets.

• Mean, median and mode selected according to distribution characteristics
• Variance, standard deviation and interquartile range describe numerical spread
• Percentiles identify service levels, thresholds and unusual performance values
• Weighted averages account for differing volumes across reporting groups
• Rates and ratios use consistent numerators, denominators and reporting periods
• Control charts distinguish common cause variation from special cause signals

Participants can design basic samples and estimate population measures with stated uncertainty.

• Population, sample and sampling frame defined for analytical studies
• Simple random, stratified and systematic sampling methods compared
• Sample size considers precision, variability, confidence and operational constraints
• Sampling bias identified through coverage, nonresponse and selection mechanisms
• Confidence intervals estimate plausible ranges around sample statistics
• Margin of error communicated without claiming unwarranted precision

Participants can select, run and interpret common tests for business hypotheses.

• Null and alternative hypotheses express measurable business claims clearly
• t-tests compare means between independent or paired operational groups
• Chi-square tests examine associations between categorical variables
• ANOVA compares average outcomes across three or more groups
• p-values, confidence intervals and effect sizes interpreted together
• Multiple testing risks managed through planned comparisons and cautious conclusions

Participants can model relationships between variables and assess regression results.

• Pearson and Spearman correlation selected by variable type and distribution
• Simple linear regression estimates relationships between outcome and predictor
• Multiple regression controls for relevant explanatory variables and confounders
• Regression coefficients interpreted using units, direction and practical significance
• Residual plots assess linearity, constant variance and unusual observations
• Multicollinearity checked using correlation patterns and variance inflation factors

Participants can build and assess classification models for binary business decisions.

• Binary target variables define events such as churn, fraud or default
• Logistic regression estimates event probability from selected predictor variables
• Training and test sets separate model development from evaluation
• Confusion matrices quantify true positives, false positives and missed events
• Precision, recall and F1 score reflect different operational error costs
• Classification thresholds selected according to capacity, risk appetite and value

Participants can produce and evaluate short-term forecasts from historical time series.

• Time series components separate level, trend, seasonality and random variation
• Moving averages smooth short-term fluctuations in operational demand data
• Exponential smoothing weights recent observations for short-range forecasts
• Seasonal indices adjust forecasts for recurring monthly or weekly patterns
• MAPE, MAE and RMSE compare forecast error across candidate methods
• Forecast assumptions document known events, constraints and expected changes

Participants can identify unusual observations and distinguish errors from meaningful exceptions.

• Box plots flag observations beyond interquartile range decision limits
• Z-scores identify values distant from the sample mean
• Control limits detect unusual process performance over consecutive periods
• Isolation Forest identifies anomalous records within multivariable datasets
• Business rules distinguish data entry errors from legitimate exceptions
• Investigation queues prioritise anomalies by financial, service or risk impact

Participants can create interpretable segments from customer or operational data.

• Segmentation objectives define actionable groups before selecting clustering methods
• Feature scaling prevents large numerical ranges dominating distance calculations
• K-means clustering groups records around internally similar centroids
• Elbow plots and silhouette scores inform suitable cluster numbers
• Cluster profiles describe size, behaviour, value and distinguishing variables
• Business validation tests whether segments support differentiated actions

Participants can design visual outputs that support accurate, timely decisions.

• Chart selection matches comparisons, distributions, relationships and time patterns
• Bar charts, line charts and scatter plots use meaningful scales
• Pre-attentive attributes direct attention through position, colour and size
• Dashboard layouts prioritise decisions, exceptions, trends and supporting detail
• Annotations explain material changes, assumptions and relevant operational events
• Accessibility checks address colour contrast, labels and nonvisual interpretation

Participants can communicate findings, uncertainty and recommendations to decision-makers.

• Situation, complication, question and answer structure analytical briefings clearly
• Findings distinguish observed patterns, tested results and informed interpretations
• Recommendations link evidence, expected benefit, owner and implementation condition
• Uncertainty statements explain data limitations, assumptions and confidence bounds
• Executive summaries present decisions required before methodological detail
• Review questions test whether conclusions match evidence and stated scope

Participants can define controls for maintaining analytical models and recurring reports.

• Model documentation records purpose, inputs, method, outputs and limitations
• Data drift monitoring detects material changes in input distributions
• Performance monitoring compares predicted outcomes against realised results
• Version control tracks query, script, dataset and report changes
• Approval workflows define review responsibilities for high-impact analytical outputs
• Retirement criteria identify when models require revision or replacement

Participants can assemble a structured analysis pack from their chosen business problem.

• Analysis brief confirms decision owner, scope, measures and success criteria
• Source inventory identifies datasets, fields, access conditions and refresh dates
• Preparation log records cleaning decisions, exclusions and derived variables
• Method selection aligns analytical technique with question and data structure
• Interim findings reviewed against assumptions, quality issues and stakeholder context
• Presentation outline connects evidence, implications, recommendations and next actions

Participants can build, review and present a complete data analysis addressing a defined business decision.

• Business question, measures and hypotheses finalised for capstone analysis
• Prepared dataset validated against source records and documented rules
• Selected statistical or predictive method applied with stated assumptions
• Visual findings communicate patterns, uncertainty and decision implications clearly
• Peer review challenges method choice, interpretation and recommendation evidence
• Final presentation submits analysis pack, actions and monitoring measures