11 Essential Business Intelligence Interview Questions *

A data cube describes the BI data structure in memory before it is shipped to a BI UI tool to be displayed to the user. It is a multi-dimensional data representation made for better visualization, data slicing, and drill-down techniques. The UI usually does not display a literal cube, but generally 2D slices of it for better human readability:

A data cube is usually based on a single denormalized fact table and some number of dimension tables representing data cube dimensions. The star and snowflake schemas were specifically designed to aid in building data cube structures in memory.

An example schema might consist of:

• Time buckets—time dimension table
• Customers—customer dimension table
• Products—product dimension table
• Sales amount (units sold)—fact table

The data cube structure for this schema can be thought of like this:

A fact table contains dimension keys and numerical values for some measures. Each dimension key represents a dimension that measures are for. Measures can be aggregated across dimensions to build a drillable data cube.

Dimension tables are dictionary tables used to display dimension labels and information on BI visual interfaces.

1. Build company analytical data storage (data warehouses, data marts).
2. Devise an analytical data storage schema based on both actual company data and BI demands.
3. Initially, populate analytical data storage with existing company data, and then update it regularly.
4. Set up BI tools on top of analytical data storage.
5. Develop BI reports.
6. Maintain and modify BI reports according to changing needs.

The candidate should name at least two benefits from those listed below. It can be in their own words, as long as it’s close in meaning. The more benefits they can name, the better.

Data normalization:

1. Removes data duplication.
2. Allows finer transaction granularity. Each referenced table data could be changed independently in its own transaction without affecting its foreign key relationships.
3. Enables clearer referential integrity. The smaller entities produced by normalization allow modeling business objects and their relations as close to the real world as is possible.
4. Allows incremental schema changes. Adding or deleting columns in one table does not affect the structure of referenced tables.

A data mart stores a subset of company data that focuses on a specific department, activity type, or set of subproblems.

Separating data into data marts allows for better performance and separation of tasks for BI analysts and business users.

This strategy is a matter of design and operational convenience. While there is no definitive answer on when to use it or not, it’s usually considered appropriate to build a data mart when a company runs different lines of businesses that are very much independent in terms of their underlying data and reporting needs.

For example, if the same company is building trucks and running an online game application, it likely makes sense to handle these sub-concerns in separate data marts.

The star schema consists of dimension and fact tables. Each dimension table represents a “metric” that can be used in BI reporting. A fact table references dimension tables for each corresponding metric the fact table covers.

The snowflake schema is an extension of the star schema in such a way that dimension tables could be further normalized and split into main and secondary dictionary tables.

OLTP stands for “online transactional processing.” It is used for company business applications. They are most often customer- (i.e., people- or business-) facing.

OLAP stands for “online analytical processing.” It is used for a company’s internal analysis by department leads and company top management to steer the company.

There are numerous BI tools on the market, but among the best-known are:

1. Oracle Business Intelligence Enterprise Edition (OBIEE)
2. IBM Cognos Analytics
3. MicroStrategy
4. The SAS product line
5. SAP BusinessObjects
6. Tableau
7. Microsoft Power BI
8. Oracle Hyperion
9. QlikView

This type of free-form question isn’t about the candidate providing a correct answer, per se. It’s more about sparking a discussion so interviewers can get a sense of the depth of the candidate’s expertise, and where that overlaps with the company’s current needs.

BI provides quick and simple methods to visualize company metrics, generate reports, and analyze data.

These methods, in turn, help top management to:

1. Analyze existing trends.
2. Lay out company development plans.
3. Ensure such plans are executed as scheduled.
4. Detect anomalies and problems.
5. Apply corrective actions.

The candidate should name at least two benefits from those listed below. It can be in their own words, as long as it’s close in meaning. The more benefits they can name, the better.

Data denormalization provides:

1. Simpler initial data schema design.
2. Better data write/read performance.
3. Direct applicability in data warehouses. Fact and dimension tables in data warehouses are usually designed without regard to data normalization to ensure fast and straightforward data retrieval.
4. Precomputation and query performance improvements for data cube BI slice-and-dice and drill-down analysis.

BI developers are generally expected to:

1. Analyze company business processes and data.
2. Standardize company data terminology.
3. Gather reporting requirements.
4. Match the above requirements against existing data.
5. Build BI reports.
6. Analyze the fleet of existing reports for further standardization purposes.

This question can be useful as an opening one—not just to help filter undesirable candidates and put more qualified candidates at ease but also to provide an opportunity to discuss any nonstandard responsibilities that may be involved in the particular job at hand.