Testing SAP Analytics Cloud
The best way to have an opinion is to prove yourself. Below I describe a small project that I have done with the objective of testing SAP Analytics cloud, as well as the corresponding conclusions.
1 Business Case
2 Developing the project
2.1 Dimensions, attributes and key figures
2.2 Making reports. Restricted KF, aggregations …
2.3 Simulations and data exploration
2.4 self-consumption of data
(note: the screenshots at following are in spanish language)
1. Business case
From a structured excel file, we want to easily publish the data of this excel. We want the published information to be dynamic (the user can select the information according to their interest), and also the user can make certain ad-hoc reports.
For this test, I have done a small project based on the open datasets of the Generalitat of Catalunya “Monthly budget execution” (255,477 rows, 18 dimensions, 12 indicators).
Note: I take for granted that the reader of this post has a minimum knowledge of Business Intelligence and knows similar tools (cloud or non-cloud)
! I exclude from the test the different integration modes beyond the loading of excel files, as well as security elements and Content Cloud analysis.
2. Developing the project
I will not go into details of all the necessary steps (manuals already exist for this, and the tool is also reasonably intuitive), but I will focus on those aspects that have attracted my attention, as well as on the results obtained.
2.1 Dimensions, attributes and Key Figures
The first step in terms of modeling is to define which fields are dimensions, which with key figures and which are attributes of dimensions. The tool makes a proposal of attributes and key figures (which in our example, the majority was right). Logically it can not know which are attributes, although it will be very easy to define, especially for the texts of the dimensions)
The first thing that strikes me is that it automatically detects errors in the data
In our case, it seems that the dataset used contains some data errors:
These “errors” I can see are due to two reasons:
- Texts of dimensions that have changed over time: for example, changes in the name of the departments of the Generalitat.
- Integration errors: I observe that several rows for the same code are accompanied by similar descriptions but not the same (abbreviations, spaces, etc). I guess the information that has been collected from several departments has not been integrated correctly.
I can conclude this quickly because I can navigate to the erroneous data detected:
Here comes my second and pleasant surprise: BO Analytics Cloud proposes for each case, how to do the data cleansing:
That is, it proposes a transformation to be applied (which will apply in this loaded excel and also in the excels that we load later). In this dataset, the problem is reproduced in several dimensions, so that what could have been a couple of days of work, I resolve in about 2 hours.
2.2 Making reports. Restricted key figures, aggregations …
The realization of the reports is very intuitive and reasonably powerful.
The first thing to be clear is that SAP Analytics Cloud is NOT SAP Web Intelligence (for example), and therefore we can not expect the same level of functionality. It must also be borne in mind that the modeling possibilities are small, and therefore the report must be based on a correct data model.
My third pleasant surprise is that for example, in the report of the following image, I raised for each indicator, the typical comparison between periods that the user selects. Well, without knowing the tool, I immediately saw how to perform the restricted ratios and how to link them with the corresponding user filter:
Report with restricted key figures
screenshot how to perform restricted key figures
Also comment that the different reports that are built on the same screen, automatically are attached to the filters that are on the screen or to the general filters of the entire document if they exist.
Here are some captures of reports made (with real data):
Comment that the idea of the previous screenshot report was to have a series of filters that the user can select. But I also wanted to see if the user could dynamically choose the indicator to be visualized (not in all the reporting tools it is possible), and indeed it is possible and also in a simple way.
2.3 Simulations and data exploration
SAP Analytics cloud, in addition to the “traditional” reporting features, incorporates what it calls “intelligent discovery”, which allows both to discover relationships between data and to simulate what impact a data change would have on the other data:
Partial screenshot of the report generated in the “intelligent discovery”
Also comment that SAP places this tool in the Data Discovery section. My first (and probably) hasty opinion is that for this objective it still lacks a bit of maturation, although it certainly seems that SAP is on the right way
2.4 self-consumption of data
It is possible to publish one or several data analysis views in which the user can create and export self-consumption reports. It is possible to place the dimensions in the desired position, as well as to filter by any value of any dimension.
There are also various functionalities available, such as the classification, N maximums, visualize table or graph … and logically the excel export.
The following conclusions are the result of the project described above, and therefore we have a limited experience. Due to this, these conclusions are susceptible to be corrected / extended based on major experiences
High productivity: worth as an example that I without prior knowledge of the tool, took only one day to make the reports exposed (including the effort in the data-cleansing).
Intuitive frontend: the interface is pleasant and intuitive.
It is not necessary to install or configure anything (*): obviously it is the most interesting and inherent aspect in a cloud product. It is enough to have a user, design the project and publish it so that it is immediately accessible to all users.
High performance: although I insist on the boundedness of the test (255,477 rows loaded), I have not perceived any type of latency in the sample of results.
Data Discovery: We integrate “simple” reporting with data discovery possibilities in a simple way.
Real-time integration with on-premise local databases such as BW4 / HANA, Oracle, SQLServer … (although with some functionality restrictions today), as well as with other cloud tools.
(*) Except for connections to external source systems (on-premise or cloud) and SSO if applicable
It is not “pixel-perfect”: the final result is not fully controllable. In addition there are some strange behaviors when moving certain objects, although it is expected that these small inconveniences will be solved in future releases. However, this “weak point” is acceptable if we consider that SAP classifies this product as data discovery.
Limited modeling functions: Although it is possible to make calculated dimensions and key figures, integrating several data sources in the same report, as well as linking dimensions, is a tool that will not solve any deficiencies that we have in our data model.
It is difficult to understand easily the predictive analysis: Although I have already commented that it was not an objective of this document, I get the impression that just as the tool offers high productivity and is very intuitive for the development of reports (predefined or ad-hoc), it is not so evident to obtain fast results and easily interpretable in the section of predictive analysis
We would be wrong to think that SAP Analytics Cloud is (for example) an “SAP Web intelligence in the cloud” (in fact SAP does not intend to compare the two tools). It is another different tool with many interesting aspects but also with limited functionalities if we compare it with other on-premise tools.
Therefore let us know in advance if the functionality offered fits with the approach of our project.
Let’s discard SAP Analytics Cloud if we want to make a scorecard or a reporting with high presentation requirements. Instead, let’s keep it in mind if we want to disseminate data to many number of users, with pre-defined reports or with the objective of data discovery. We must also take it into account if we want a lower infrastructure cost compared to traditional architectures
I await your comments