With the continuous increase of technology companies around the world, the quantity of data is increasing just as quickly. In business analytics, it is not the quantity of the data that matters, it is quality.
What is data profiling?
The quality of data is determined by data profiling. As Ralph Kimball puts it “Data Profiling is a systematic analysis of the content of a data source.” In simple terms, data profiling is examining the data available in the source and collecting statistics and information about that data. These profiling and quality statistics have a large effect on your business analytics.
Why is data profiling important?
- With more data comes a greater importance of higher quality, for optimal results of any analysis.
- If the quality of your data is poor, it could affect your company’s success more than you think. It was reported by the Data Warehouse Institute’ that it costs $600 billion a year to American businesses to recover from data quality problems. Moreover, it also leads to delay and failure of large and important IT projects and goals.
- High quality data allows for companies in the retail industry to increase sales and customer retention rates.
- Error free decision making is the goal of any company in any industry. Proper data profiling leads to just that.
Types of data profiling in business analytics
There are three main types of data profiling:
- Structure discovery: Verifying the data is reliable, consistent, and has been arranged correctly based on a specific format – for example, if US phone numbers have all 10-digits.
- Content discovery: The discovery of errors by looking at individual data records – i.e. which phone numbers are missing a digit.
- Relationship discovery: How the parts of data are interconnected. For example, key relationships between tables or references between cells or tables. Understanding relationships is imperative to reusing data. Related data sources should be combined into one or collected in a way that protects crucial relationships.
Best Practices for Data Profiling
Before you begin you data profiling journey, it is important to know and understand some proven best practices.
First, identifies natural keys. These are specific and distinct values in each column that can help process updates and inserts. This is useful for tables without headers.
Second, identify missing or unknown data. This helps ETL architects setup the correct default values.
Third, select appropriate data types and sizes in your target database. This enables setting column widths just wide enough for the data, to improve visibility and performance of the profiling.
Following these best practices will ensure your data to be improved to the highest quality, preparing it for further in depth analysis. The higher the quality of your data, the more precise the results produced by any analysis will be. It is extremely worth any analysts time and money to conduct data profiling steps before proceeding to calculate any information. A single error of an immense amount of data could decrease the credibility of the analysis results.