Data modeling deals with the creation of data models for purposes of data storage. Demand for data scientists and professionals has sharply risen over the years thanks to the incorporation of data into business for insight generation purposes. If you are eyeing a data-related job, you are better off knowing a thing or two about data modeling as you may be interviewed about it.
To help you prepare for your upcoming interview, this article will look at some of the data modeling questions you should expect in your interview. Take a look at the following:
1. Define A Data Model
Data models organize various data elements and define the connections between them. They are made up of entities represented as tables and stored in a given database. These entities generally have attributes, which refer to the details to be tracked. Some of the most common entities in the business world are manufacturers, products, and sellers.
2. Can You Mention The Different Data Models?
There are three main data models: physical, conceptual, and logical models. In physical data models, the framework describes the physical storage of data in the database, while conceptual models only focus on offering users a high-level view of the data. Lastly, logical data models allow data to be represented logically and physically. They act as a bridge between the physical and conceptual models.
3. Can You Mention The Different Types Of Critical Relationships Found In A Data Model?
Parent and child tables are connected via a relationship line in data modeling. In identifying relationships, a thick line connects both tables. This only occurs when the reference column of a child table forms part of its primary key. For non-identifying relationships, the reference column in the child’s table does not form part of its primary key. A dotted line, therefore, connects both. The last relationship is self-reclusive, where the primary key on a given table is connected to a standalone column of the same table.
4. What Errors Have You Faced In Your Work?
There are four main errors that I have faced in my data modeling project works. The most common is the creation of excessively broad data models, which occurs when tables are run higher than 200, making the data model too complex. Another common error is a missing purpose due to a lack of knowledge of a business’s goals or missions. Therefore, a data modeler must fully understand the entity’s business model. Other errors include unnecessary surrogate keys and inappropriate denormalization, which leads to redundant data that is challenging to maintain.
5. Define OLTP
Fully known as online transactions processing, OLTP is an operational system that supports transaction-oriented architecture. It is mainly focused on data integrity maintenance, query processing, and the effectiveness of transactions per second in lathe environments with multiple access. These systems are commonly used to enter orders, add items to shopping carts, book online air tickets, and for online banking. You also rely on OLTP to send text messages.
6. Tell Us About The Benefits Of Data Modelling
Data Modelling has its fair share of advantages. It is used in software development to limit the number of errors and improve application or database performance. It also allows easier data mapping between multiple processes and the reusability of data models in repetitive tasks. Other benefits are improving data quality and enhancing the strength of an application, reducing maintenance costs for a large-scale system, and improving communication between the developers and business intelligence.
7. What Do You Understand By Data Modelling?
It is creating a data model representing a given dataset and its relationship to other data for storage purposes. Data modeling is also referred to as database modeling. It comes in handy in different data-related jobs such as data engineering, software development, data science, and any other job that requires data to be prepared, analyzed, and processed. Data modeling helps reorganize, optimize and reorganize data to fit different needs of a company or an organization. A data modeling project results are data models categorized into conceptual, physical, and logical models.
8. Do You Know The Importance Of Metadata?
Metadata refers to the information about a given set of data. It reveals the type of data, its purposes and the intended parties. Various types of metadata exist as they are categorized based on purpose. Business metadata applies to business and defines business regulations, policies, and data ownership, among many others. Operational metadata carries information regarding the business regulations and is normally used for task performance by management. Technical metadata sheds more light on a database system. These include names, size of tables, data types, and different attributes. Lastly, descriptive metadata carries information about a book, file, video, or image. It will capture the titles, date, author, size, and date of publishing, among many others.
9. Define Nosql Database
NoSQL databases do not require any fixed schemas, which makes them non-relational. Instead of schema variations, they have joins that improve their performance. They are also easier to scale. However, they cannot store huge amounts of data distributed. There are four types of such databases: column-based, key-value, document-based, and graph-based. The graph-based databases organize data into edges and nodes and are therefore multi-relational. Column-based NoSQL databases use columns instead of rows to organize data; in document-based NoSQL databases, semi-structured data are stored in document formats. Lastly, key-value NoSQL databases have linked keys and values. They are more efficient, highly scalable, and mostly applied in session management and caching applications.
10. Differentiate SQL And Nosql
SQL and NoSQL differ in flexibility, models, data types, and transactions. SQL has a relational data model, deals with structured data, and follows a strict schema and ACID properties in their transactions which fully means atomicity, consistency, isolation, and durability. On the other hand, NoSQL has non-relational data models that deal with semi-structured data and dynamic schema, making them very flexible. They follow BASE properties. BASE fully means basic availability, soft state, and eventual consistency.
11. Can You Tell Us The Phases In The Data Modelling Development Cycle?
Data Modelling has five phases. The first phase focuses on gathering business requirements, which are the functional requirements and the reporting needs from end-users and business analysts. The second phase comes up with a conceptual data model based on the requirements in phase one. This model includes primary entities and the associations between them. The third phase is logical data modeling, where a logical data model describing the logical representation of the organization’s business requirements is created. The fourth stage deals with physical data modeling, creating a complete data model with columns, tables, and the relationship between different entities and properties. The last phase is the database phase, where SQL codes are created thanks to data modeling tools and executed on servers to create database entities.
12. You Mentioned That Nosql Databases Do Not Have Schemas. Can You Define What A Star Schema Is?
Schemas are basic representations of data structures and their existing relationships. A star schema has a central fact table connected to different dimension tables. This central table has a foreign key that acts as the dimension tables’ primary key. It is mainly referred to as a star schema because the relationship between entities is visualized as a star, with points diverging from a central fact table.
13. Mention The Different Types Of Visual Data Modelling Techniques
Data Modelling has two types of visual techniques. The entity-relationship model technique designs traditional databases and plays an important role in data normalization performance through data redundancy reduction. It also makes top-level data view and database schema visualization a reality. Such models identify entities, their attributes, and the relationship between them.
The other technique is UML, or Unified Modelling Language, a general-purpose language that comes in handy in database development, modeling, and visualization in software engineering projects. In these models, different diagrams represent software systems. A good example is the class diagrams that define class attributes, class methods, and relationships.
14. Define Recursive Relationships
Whenever an entity is related to itself, the type of relationship that exists is said to be recursive. Such relationships are complex and require several delicate methods to map data to given schemas. In most cases, a foreign key is added to specific attributes. A recursion should have an exit path in the existing entity relations.
15. Mention The Existing Dimensions In Data Modelling
There are five types of dimensions in data modeling. Conformed dimensions apply in multiple database pages, where measures and facts are categorized and described across data marts and facts. Denigrated dimensions are part of the primary key and do not have dimension tables. For junk dimensions, the attributes do not belong to existing dimension tables or the fact table but are texts or flag-like symbols indicating yes or no. Slowly changing dimensions have consistently changing attributes. The last types are the role-playing dimensions, where several relationships exist between a fact table and other tables.
16. Tell Us About The Significance Of The Third Normal Form
Commonly known as 3NF, the third normal form prevents data duplication and the different anomalies. Relations meet the third normal form when transitive dependencies for non-prime attributes exist, as in the second normal form. Certain conditions must be met for every non-trivial functional dependency to achieve a 3NF relation. There should be a super key and a prime attribute where each attribute is part of the candidate key. A table can only achieve 3NF form if it’s first in 1NF and then 2NF. It is also important that the table rows strictly depend on the keys.
17. You Had Defined Star Schema. Can You Please Tell Us What Snowflake Schema Is?
Unlike the star schema, a snowflake schema has dimension tables that are not directly linked to the fact table. However, they are connected to the fact table via other dimension tables. Therefore, these tables represent a snowflake-like structure, with the fact table at the center. It is also worth mentioning that the tables are normalized to 3NF form, with each dimension level representing a level in the hierarchy. A snowflake schema has several relationships. Given the minimal disk requirements, it comes in handy for query performance enhancement.
18. Differentiate OLTP And OLAP
OLTP, fully known as an online transactional system, focuses on several short online transactions. In contrast, OLAP, fully known as the online analysis and data retrieving process, handles a larger volume. The former also uses traditional DMBS, while the latter uses large data warehouses. Other differences include: OLTP has normalized database tables while OLAP has denormalized tables, OTLP takes milliseconds to respond while OLAP responds in seconds, and OLTP is used in real-time business applications while OLAP analyzes data measured in categories and attributes.
19. Differentiate Normalization And Denormalization
Normalization structures relational databases using normal forms that reduce data redundancy and encourage data integrity. Larger tables are divided into smaller tables and linked through references or relationships. Normalization, therefore, reduces repetition and enhances logical storage. On the other hand, denormalization improves the read performance. It adds redundant data copies and compensates for the writing performance. It also comes with duplicate data entries that increase the performance of relatively heavier loads. Both normalization and denormalization are therefore important in data modeling.
20. Explain What Database And Cardinality Are
Relational databases are digital database systems found in relational data models. They are used to store data in different tables. Common examples are Teradata, Oracle, PostgreSQL, SQLite, and MySQL. In these databases, data transforms into rows and columns that can be queried with SQL, given that it is standardized. Relational database management systems, popularly known as DMBS, add and alter the data found in these tables systematically into the database, thus maintaining data integrity.
Cardinality refers to the exceptional levels of data values found in a given column. A column with a high cardinality value has a large percentage of unique values, while the opposite means that a given column has repeated data.
21. You Have Mentioned Data Mart Several Times In This Interview. What Is It?
Data is grouped into datasets and then stored in data marts. Therefore, a data mart is a condensed warehouse version, making it a subset of the data warehouse. They are generally made for different units and departments. The marketing, sales, finance, and HR departments will have their data marts in an organization. However, some data marts can also be empowered to interact with a given data warehouse for data processing purposes. This can only happen in given departments.
22. Which Among The Two Schemas Is Better?
There are two schemas in data modeling: star and snowflake schemas. Star schemas exist in denormalized forms, meaning that one only needs a few joins for querying purposes. It is also simple and fast to query in this schema instead of Snowflake schemas. This is because snowflake schemas are normalized and therefore require several joins, complicating the querying process and making execution quite slow. On the other hand, snowflake schemas are easier to maintain, given that they don’t have redundant data. Star schemas have redundant data, which makes them harder to maintain. The best type of schema will depend on the type of project one intends to work on. For dimension analysis, the snowflake model works, while the star schema comes in handy for metrics analysis.
23. Mention The Different Types Of Measures In Data Modelling
There are three types of measures in data modeling. Non-addictive measures do not allow any application of aggregation function on top of it. A good example is a ratio, percentage of indicator column in a given fact table. semi-additive measures allow the application of some aggregate functions on top. However, some do not make the cut. Lastly, additive measures allow the application of all aggregation functions on top.
- Top 25 Big Data Interview Questions and Answers
- Top 25 Data Quality Analyst Interview Questions and Answers
- Top 25 Data Scientist Interview Questions and Answers
- Top 25 Data Structure Interview Questions and Answers
- Top 25 Amazon Data Engineer Interview Questions and Answers
- Top 26 Data Warehouse Interview Questions & Answers
- Top 20 Data Migration Interview Questions & Answers
- Top 25 Data Analyst Interview Questions and Answers
- Top 25 Clinical Data Management Interview Questions & Answers
24. Have You Ever Used The Erwin Tool? Mention Its Use
I have used the ERwin tool in several projects. It is one of the most reliable software for data modeling. It can create databases from physical modes and streamline different processes. This tool also allows you to customize your database as much as you need as you get to pick fonts, colors, and layouts, among many other attributes. Lastly, ERwin can be used in reverse engineering, a technical process.
25. Explain What An Artificial Key Is
An artificial key is also known as a derived or surrogate key. It violates the principle of stability, unlike the natural key that changes equally when other keys change, making it an organic part of the database. It is of no value to the system and only impacts the output. The surrogate or artificial key stores information that lacks semantic meaning. Such information can be transferred easily when an artificial key is used. It can also track variations in employee performance and other related activities.
Our 25 recommendations should help you ace your upcoming data modeling interview. Remember to have all the answers at your fingertips and answer the questions confidently. Groom yourself well and create a good first impression for a one-on-one interview; you will be good to go. We wish you all the best.