IIBA-CBDA · IIBA · Analyze Data · Bank updated 2026-07-02
IIBA-CBDA practice questions: Analyze Data
5 free questions from 49 on this area · answer and explanation for each · no sign-up
These 5 questions come from the Analyze Data section of our IIBA-CBDA bank (49 questions on this area, which carries 16% of the real exam). Every question is original, with the correct answer explained and linked to the source it is drawn from.
1/5 · Analyze Data · easy
Before building any model, a data analyst spends time handling missing values, removing duplicate records, correcting inconsistent formats, and resolving outliers in the source dataset. What is the primary purpose of this set of activities?
ATo visualize the data for stakeholders
BTo improve data quality so that subsequent analysis produces reliable results
CTo select which predictive model architecture to use
DTo compress the dataset so it occupies less storage space
Show answer & explanation
B is correct. Data preparation, including cleansing, ensures the dataset is accurate, complete, and consistent. Poor-quality data undermines every downstream step, so the central purpose of handling missing values, duplicates, formatting, and outliers is to improve data quality and produce reliable analysis.
↗ IIBA Business Data Analytics Guide — Analyze Data
2/5 · Analyze Data
A dataset contains a customer's date of birth and the date of each transaction, but the analyst believes 'customer age at time of purchase' would be a more useful input to the model. The analyst computes this new column from the existing fields. What is this activity best described as?
AData deduplication
BFeature engineering
CData anonymization
DSampling
Show answer & explanation
B is correct. Feature engineering is the practice of creating new variables (features) from existing raw data to improve a model's predictive power, such as deriving 'age at purchase' from a birth date and a transaction date. It is distinct from cleansing operations like deduplication, from privacy operations like anonymization, and from record selection like sampling.
↗ IIBA Business Data Analytics Guide — Analyze Data
3/5 · Analyze Data
A data analyst has a single, modest-sized dataset and wants a reliable estimate of how well a predictive model will generalize, without sacrificing too many records to a permanently held-out test set. The analyst decides to repeatedly partition the data into several equal folds, training on all-but-one fold and validating on the remaining fold, rotating which fold is held out each time, then averaging the results. Which validation approach is being described?
AHyperparameter grid search
BA single train/test holdout split
CBootstrapping the test set by sampling with replacement
Dk-fold cross-validation
Show answer & explanation
D is correct. The described procedure — partition into k equal folds, rotate which fold is validated, average the metrics — is k-fold cross-validation. It is favored when data is limited because every record serves in both training and validation across folds, giving a more stable generalization estimate than a single holdout. Bootstrapping and grid search are different techniques serving different purposes.
↗ IIBA Business Data Analytics Guide — Analyze Data
4/5 · Analyze Data · hard
A relationship between advertising spend and sales is clearly curved, but an analyst fits a simple straight-line model to it. The model performs poorly on the training data AND poorly on the validation data, with both errors high and similar. Which diagnosis best describes this situation, and what is a sensible next step?
AThe model is underfitting (high bias) because it is too simple to capture the curved pattern; the analyst should increase model complexity, for example by adding non-linear terms
BThe model is overfitting (high variance) and should be simplified or regularized
CThe data is too small and the only remedy is to gather far more records
DThe validation split was unlucky and the analyst should simply re-shuffle and re-split until errors drop
Show answer & explanation
A is correct. When both training and validation errors are high and close together, the model is underfitting — too much bias, too little flexibility for the curved pattern. The fix is to add complexity (e.g., non-linear/polynomial terms or a more expressive model). This is distinct from overfitting (low training error, high validation error) and is not solved by more data or re-splitting.
↗ IIBA Business Data Analytics Guide — Analyze Data
5/5 · Analyze Data
A dataset contains a 'region' field with the unordered values North, South, East, and West. An analyst plans to feed this field into a regression model and replaces the regions with the numbers 1, 2, 3, and 4. A reviewer warns this will mislead the model. What is the core problem, and what is the appropriate fix?
AThe problem is missing data, so the analyst should impute the most frequent region before encoding
BThe numbers are fine because any numeric encoding is mathematically equivalent for a regression model
CNumbering unordered categories implies a false order and magnitude; the field should be one-hot encoded into separate indicator columns instead
DThe field should be normalized to a 0-to-1 range so the four region codes are on the same scale as other features
Show answer & explanation
C is correct. Nominal categorical variables have no inherent order or distance. Mapping them to consecutive integers injects a fictitious ranking and spacing that algorithms interpret as meaningful, biasing the analysis. One-hot (dummy) encoding represents each category as its own binary indicator column, preserving the categorical nature without implying order or magnitude. Imputation and normalization solve different problems and do not remove the spurious ordering created by integer labeling.
↗ IIBA Business Data Analytics Guide — Analyze Data
Other IIBA-CBDA areas
The same kind of free sample for every other section of the IIBA-CBDA bank:
Study Analyze Data with instant feedback
6 free questions · filter study mode by area and difficulty · error log with spaced repetition · no card, no dumps, no ads.
Create your free account →
Read next
ECBA vs CCBA vs CBAP: Which IIBA Certification Fits Your Experience Level? (2026)
ECBA, CCBA or CBAP? The eligibility gate that picks for you, what the 2026 ECBA rewrite changed, and each credential's three-year cost, verified against IIBA.
PMI-PBA vs CBAP: The Business Analysis Certification Decision, Settled with Data
PMI-PBA ($405-555, 36 months of experience) vs CBAP ($495-650 all-in, 7,500 hours plus references): eligibility, exam mechanics, three-year cost, salary data.
ExamDeck is an independent study tool, not affiliated with IIBA®. IIBA-CBDA and BABOK® are trademarks of the International Institute of Business Analysis. Exam facts checked against official IIBA materials (as of August 2026); confirm current details on iiba.org.