IIBA-CBDA flashcards

33 cards. Read the term and the definition here. The deck at the top of this page is optional.

Research question vs hypothesis

A research question asks what relationship or pattern to investigate; a hypothesis states a testable, directional prediction that analysis can confirm or reject.

SMART objective criteria

Specific, Measurable, Achievable, Relevant, Time-bound. A SMART objective defines the outcome, metric, target, and deadline so success can be evaluated objectively.

Root cause analysis

A structured method that traces an observed problem to underlying causes rather than symptoms, using techniques such as the five whys or cause-and-effect diagrams.

Scope statement

A scope statement defines what the research will and will not cover, including populations, time periods, systems, and constraints, preventing analysis creep.

Decision context

The business decision the analysis must inform, including who decides, available options, timing, and consequences; it shapes the research question.

Assumption vs constraint

An assumption is a believed-but-unverified condition; a constraint is a fixed limitation. Both must be documented because they affect feasible questions and data.

Measurable success criterion

A quantified threshold or target, such as a 5% churn reduction, that lets stakeholders judge whether the research answer supports the intended business outcome.

Primary vs secondary data

Primary data is collected firsthand for the current question; secondary data was collected earlier or for another purpose and may require validity checks.

Structured vs unstructured data

Structured data fits predefined fields and rows; unstructured data such as text, images, or audio lacks a fixed schema and often needs transformation before analysis.

Data quality dimensions

Common dimensions include completeness, accuracy, consistency, timeliness, validity, and uniqueness; each should be assessed against the analysis purpose.

Sampling bias

Occurs when selected observations are not representative of the target population, making estimates systematically different from the population parameter.

Data lineage

The documented path of data from source through transformations to reporting; it supports traceability, error diagnosis, and trust in results.

Metadata

Data about data, such as definitions, owners, formats, and refresh dates; metadata enables correct interpretation and appropriate use of datasets.

Convenience sampling risk

Selecting easily available subjects can distort estimates; because inclusion is not random, results may not generalize to the intended population.

Mean median mode

Mean is the arithmetic average, median is the middle ordered value, and mode is the most frequent value; choice depends on distribution shape and data type.

Standard deviation

A measure of average dispersion around the mean; larger values indicate greater spread and affect confidence intervals and outlier judgment.

Correlation versus causation

Correlation measures association strength and direction but does not prove one variable causes another; confounding or reverse causation may explain the relationship.

p-value interpretation

The probability of observing data at least as extreme if the null hypothesis is true; it is not the probability that the null is true.

Confidence interval

A range of plausible population values estimated from sample data at a stated confidence level; wider intervals indicate greater uncertainty.

Outlier decision rule

Investigate values far from the distribution or model expectation before removing them; they may be errors, rare valid events, or important signals.

Missing data mechanisms

Missing completely at random, at random, or not at random; the mechanism determines whether deletion or imputation may bias results.

Statistical vs practical significance

A statistically significant result may be too small to matter; practical significance compares effect size with business thresholds and costs.

Audience-tailored reporting

Adjust detail, vocabulary, and visuals to the decision-maker's role; executives need conclusions and implications, analysts may need methods and caveats.

Chart selection rule

Use line charts for trends over time, bar charts for category comparison, scatter plots for relationships, and avoid pie charts for many categories.

Uncertainty disclosure

Report confidence intervals, sample limitations, and assumptions so decision-makers understand the reliability and boundaries of findings.

Data visualization integrity

Charts should use proportional scales, clear labels, and non-misleading axes; truncated axes or area distortions can exaggerate differences.

Limitations section

States constraints such as sample bias, missing variables, or timing that affect interpretation, helping readers judge how far conclusions can be generalized.

Expected monetary value

EMV equals probability times impact summed across outcomes; it compares options under uncertainty but ignores risk tolerance.

Cost-benefit analysis

Compares total expected costs with total expected benefits of alternatives; a positive net benefit or favorable ratio supports proceeding.

Sensitivity analysis

Tests how conclusions change when key assumptions or inputs vary; robust recommendations hold across plausible ranges.

Decision criteria weighting

Assign weights to criteria reflecting business priorities, score alternatives, and compute weighted totals to make trade-offs explicit and comparable.

KPI monitoring plan

Defines metrics, targets, owners, frequency, and review triggers after implementation so results can be tracked and corrective action taken.

Recommendation actionability

A recommendation states the specific action, owner, timing, expected impact, and required resources so stakeholders can implement it.