Data Profiling with No Limits

All this power comes with true simplicity. The user interface was designed by practitioners to provide the most streamlined and intuitive navigation possible. Users require no technical skills at all and can be up to speed and productive in one day. Pandora is zero administration providing a low impact solution. Even installation is simple - just tell Pandora how much memory and disk space to use, and it figures everything else out for itself. There is no configuration, schema definition or administration and it is totally self maintaining.

It supports up to 2 billion tables, each with up to 2 billion rows per table, works with all international character sets and runs on commodity hardware.

Pandora allows colleagues to work collaboratively in a secure, role-based environment, and keeps a history of all activity, allowing users to log on and immediately pick up their work where they left off.

Pandora profiles data automatically at point of load, deriving the true content, quality and meaning of every data item, producing an extensive set of over 170 statistics automatically per column such as:-

• Frequency distribution of column values, formats and phonetic patterns
• Unique value, format and phonetic counts
• Actual and most prevalent data types
• Smallest, largest, least common, most common values and formats
• Shortest, average and longest value length
• Numeric scale & precision
• Sum, average, variance & standard deviation
• Null and blank counts

Pandora automatically standardises values to a common business interpretation whilst also retaining the original variants. If a column contains, for example, mixed alpha-numeric and integer values, Pandora will calculate statistics per data type too, so you could view the column as a whole, from an alpha-numeric view point or from an integer view point.

You can interactively click on any statistic to view related information. Pandora provides interactive drill down to the data associated with every statistic instantly, regardless of volumes, thanks to the revolutionary way that it stores information. A few mouse clicks are sufficient to interactively filter, sort, count and sum any data, metadata or profile. Unexpectedly frequent values or formats across multiple columns are highlighted using statistical analysis of values, formats and lengths in the Outliers Report.

Pandora automatically calculates structural and phonetic patterns, and provides unique distributions of these as Frequency Analysis views at column, table and even enterprise level. This includes information such as the value, its frequency, its format pattern and any phonetic patterns as well as unique information such as the count of columns that this value appears in throughout the whole enterprise, and how many times it appears in the entire database (overall database occurrences). Each column can also have documented expectations, which may be manually configured or automatically derived from, for example, existing database catalogues or Data Definition Language. These are automatically compared against profiling results and any inconsistencies are highlighted accordingly.

Example documented expectations are :-

• Expected data type
• Expected length
• Expected format pattern
• Expected value range
• Expected scale & precision
• Expected nullability
• Expected uniqueness (i.e 100% for a single column key)

At a click of a button, Pandora will also produce exportable quality reports.

Pandora provides a number of reports for automatically assessing imported data, such as the Table Quality Report, Relationship Report and Outliers Report. Users can collaboratively extend business knowledge about the data by creating notes attached to any item in the repository, including attaching related views of data. All of this information can be exported as a composite report.

Pandora supports a number of export formats including web pages, allowing knowledge and data to be viewed by users outside of Pandora.

Apart from the data manipulation capabilities, Pandora also provides the ability to change the presentation of data views by setting colours, type-faces, widths, custom formatting and much more - to provide a bespoke appearance to output data.

Discovering multi-column keys and dependencies between columns in a table has always been an Achilles heel for products of this nature as traditional techniques can take hours or even days for small samples, resulting in highly misleading results. Pandora uses a revolutionary and unique approach to discovering dependencies and keys of any quality using the entire dataset in less time than existing products can analyse a small sample.

Dependencies and keys can be discovered incrementally and additively at differing quality levels, and the user can drill down to both valid and invalid detailed results, and then further drill down to the actual rows that are in error. Critical dependencies can be named, and saved as part of the repository.

Users can also simply define keys or dependencies to be checked individually, which is an extremely fast operation on 100% volume of the data. No other product can come close to the accuracy and performance of Pandora for this type of analysis.

Pandora automatically and incrementally relates the enterprise together simply by loading data. It memorises every usage of every value, and allows immediate navigation to where those values exist or to where they are embedded regardless of volumes, number of tables or columns. Pandora provides detailed information about common value domains and cross-table joins automatically. There is no time lost waiting for what-if analysis processes to run, and relationships can never be missed or misjudged. The statistics produced are 100% accurate, covering every aspect of a common value domain or cross-table join. Exceptions to a relationship are instantly available for interrogation as full interactive drill-down is provided to any aspect of any relationship, for example:-

• View the joined rows between two related tables, including unmatched records (i.e. outer joins)
• View the common value domain between two columns of different tables