In the area Data preparation there is a constant challenge: how to ensure that the Data best prepared for Tableau? And do tools such as Knime or Tableau Prep or in-house programming make more sense? With luck, the companies have decided on a tool path and the IT service provider can follow this. But how do you proceed if there are no guidelines? Based on the following five cases, we have taken a closer look at three "tools" and shed light on the Pre and Disadvantages of the individual solutions:
- Very large data sets (files with at least 3 million data sets are tested)
- Complex logic (are certain logics supported, e.g. loops)
- Error analysis (how and when errors are displayed by the tool)
- Connection to different data sources (which connections to different data sources are possible)
- Automation (update)
Our supporter for this purpose is an HP Elitebook (RAM 16GB; Intel® Core™ i5-8250U CPU 1.60 GHz).
Overview: When is which tool useful?
Tableau Prep - Nothing for complex logic
Case 1: Very large data sets
With very large data sets, this can lead to Restrictions the PC performance (CPU and RAM) and from Tableau Prep - right up to a complete PC crash. The Work status lost and you have to start all over again.
Our test result: not recommended.
Case 2: Complex logic
It can be No complex logic at all map.
Our test result: not recommended.
Case 3: Error analysis
Errors are immediately displayed either in the Editing window or in the area of Notifications indicated, but these can be very cryptic and short.
Our test result: recommended with reservations.
Case 4: Connections to different data sources
Files can be imported from MS Access, CSV, PDF, XLS, statistical data or Tableau extracts. There are not as many connections as with Tableau Desktop.
Our test result: recommended with reservations.
Case 5: Automation
If the Tableau Prep Conductor is not licensed, an automated execution on Tableau Server with Tableau Prep Conductor would be possible in the test. Local automation can be carried out locally (Batch, JSON), but the PC must be continuously switched on be.
Our test result: recommended with reservations.
Knime: complex logic and large data sets
Case 1: Very large data sets
In the area of large data sets, Knime has to check at the end whether all calculations and changes work. The result: it leads to Performance restrictionsbut not a crash. As with Tableau Prep, there is the bonus that the data in the sample is limited to a certain number of restricted (up to 1 million freely selectable). Another advantage of Knime is that the data is only processed when it is triggered by the user.
Our test result: recommended with reservations.
Case 2: Complex logic
With over 2,000 nodes and the possibility of Complex logic Knime is recommended for this case. And if there is no node to map a logic, it can be stored in a Node with Java or Python mapped and executed. The large online register for nodes and the helpful online community also outweigh the disadvantage that know-how is urgently required due to the many functions.
Our test result: recommendable.
Case 3: Error analysis
Warnings and errors are displayed in the form of Symbols or Traffic light colors immediately displayed in the node, whereby a Checking from node to node is possible. Messages can also be sent on a Console are displayed. The only drawback: error messages are sometimes not easy to understand, as the Error handling of a node is subject to the creator.
Our test result: recommendable.
Case 4: Connections to different data sources
Connections to a wide range of data sources can be established - Text formats (CSV, PDF, XLS etc.), Unstructured data types (images, documents, networks, etc.) or Databases and Data warehouse solutions (Oracle, ApacheHive, Azure etc.), Twitter and Google etc.
Our test result: recommendable.
Case 5: Automation
The advantages of automation usually cancel each other out with Knime. Although automatic execution on the Knime server is possible, it is in turn Licensed. Likewise, on the one hand, an automated run (e.g. Batch), but on the other hand none can be implemented for workflow chains.
Our test result: recommended with reservations.
Python: With patience to maximum flexibility and preparation of large data sets
Case 1: Very large data sets
In the case of very large datasets, it is possible to save the entire Code without touching the data and there are hardly any performance restrictions. There is one but: the Error analysis takes a lot of time.
Our test result: recommended with reservations.
Case 2: Complex logic
Python is in the context of complex logic very free in development. Functions such as FOR- or While loops are available. Know-how in the language is therefore absolutely essential in order to be able to make full use of the complex logic.
Our test result: recommendable.
Case 3: Error analysis
Through a Debugger is a direct Data access and better Error comprehensibility given - and the use of Python Console is also possible during debugging. However, Python only displays errors when they occur. With a very large data set, this can take up to half an hour.
Our test result: recommended with reservations.
Case 4: Connections to different data sources
A connection to File formats (e.g. CSV, XLS, TSV, file etc.) is possible. Databases and Data warehouse solutions are directly integrated in some cases.
Our test result: recommendable.
Case 5: Automation
Automated execution is possible directly on the servers and local automation (e.g. Batch) is also offered. However, databases and data warehouse solutions must Python scripts and the PC must be switched on.
Our test result: recommended with reservations.
Conclusion
All three "tools" or solutions offer the possibility to prepare data in different ways. After examining the solutions in five cases, a pattern quickly becomes clear. Tableau Prep is for a quick and Simple data preparation from smaller data sets useful. As the workflow can be integrated into Tableau Server, it can be accessed and edited from anywhere. Should complex logics and larger data sets are processed, it is recommended Knime or Python.
Find out more about our Data Driven Services
You might also be interested in these blog posts:
Automated quality assurance in the context of data analytics
