Tableau Essential Infrastructure Explained (Businesses POV)

Tableau is a BI software that has gained much popularity over the past decade and is desired choice for many companies in all kinds of industries. The software allows for great data management, data preparation and visualizations, while also being able to work with all kinds of data sources. Inside most corporations it is used to streamline and present corporate data (financial, client, sales etc.) within the structure of the company. The different tools that Tableau offers are allowing to achieve all the tasks required.

“This post will now make a review of all the necessary Tableau infrastructure that companies need in order to implement Tableau in their financial reporting.”

We are focusing on larger corporations, for SMEs we will dedicate separate discussion. This post is a must if:

• Your company is looking into upgrading their financial reporting with Tableau;

• You are a senior manager in Finance looking to digitalize financial operations;

• You are a BI Manager tasked with implementing Tableau in the company;

• You are a BI analyst trying to understand the existing Tableau infrastructure at you your company;

So lets dive in

First lets see a grand structure of how the Tableau architecture looks like and then we will talk at each element in more detail.

The entire Tableau infrastructure is aimed at connecting the end users to the available data sources (whatever they might be) while providing the opportunity to develop advanced analytical frameworks. One very good thing about Tableau that is easily overlooked is how flexible is in regards to utilizing foreign tools like Python, which greatly enhance the provided analytics. Overall there are 3 “must-have” tools from Tableau that you need:

• Tableau Server

• Tableau Prep Builder

• Tableau Desktop

Lets discuss each of the tools in more detail.

Tableau Server

Tableau Server is the place where everything is connecting in order to achieve the goals. It provides web-based access to all users and admins. This allows end users to interact with the created content (reports/dashboards) without the need to install additional software on their computers. At the same time it is very convenient for admins to work in the graphical interface environment. Tableau Server is used for 3 main tasks

               Centralized Content Repository – it is the core of the tableau infrastructure where all the created content is stored and can be accessed by different users. This is great advantage as it basically allows end users single point of reference to data that might be coming from different sources.

Example: Lets say report on short-term liquidity is required and it needs to combine financial (balance sheet data) with transactional (paid and outstanding invoices) data. The BI analyst easily can connect to the two sources available on the server and prepare the dashboard, where the FD (financial director) can see it on the web page.

               Data Governance – it serves as a single platform where different kinds of data are stored, manipulated and displayed for end users. This allows for great data governance and organization of sources, prep jobs etc. There is also a tool called “Tableau Catalog” that helps categorize the metadata of all objects on the server. It is very flexible and able to fit any requirements from the users.

               Access Management – arguably the most important capability of Tableau server is the management of accesses. Since it is single platform that combines the data sources, prep jobs and analytical dashboards it is also the best place where access is managed. The accesses can be managed via individual user profiles or most commonly within security groups (bundled users together that are given access). Access can be managed on individual object or folder/project level. One of the great advantages is the ability to apply row-level-security (RLS) both on data sources and dashboard projects. In essence RLS allows different users to see the same dashboard/analysis but only with the data they have access to. RLS again can be managed manually or with functions within the workbooks. One potential drawback is the limited number of security groups that can be created on the server.

Another great thing about Tableau server is how scalable it is. The architecture is designed to work with 10 and 10 000 users so in this regard it is very adaptive. The tool provides also good monitoring capabilities to track load, usage etc.

Tableau Prep Builder

If Tableau server is the backbone of the architecture, Tableau Prep is the working horse. This is ETL type tool that is used for data preparation, cleaning and manipulation. The best thing about Tableau Prep is how easy and intuitive is to use even by relatively novice users. It is graphical interface and basically with drag-and-drop operations you can do most things. It is far easier to use than Alteryx and has more capabilities than Power BI back-end.

Tableau Prep can be connected to many different sources including flat files, but in reality what you want to do is connect it to a data lake so everything flows smoothly end-to-end. Tableau Prep jobs should be designed by BI team who have the necessary skills in data manipulation, but also the business knowledge if they are required to combine and integrate multiple data sources in order to achieve the end goal.

Example: BI analyst is tasked by the FD (financial director) to create a dashboard on depreciation projection of all assets. Most likely he will need to use P&L data for depreciation numbers and balance sheet/fixed assets data for the net value. The approach to this is to drop the two sources into Prep job and join them together most likely on the ID of asset, in this way in a single source we will have both the net value and depreciation amount so we can do calculations on them.

Most common way of using Tableau Prep is to combine data sources and prepare it in a easy way to be used later for vizualizations. Then the Prep job is published and run on the server. This allows to use the resources of the server and also to schedule regular runs. It is vital for any BI project with Tableau how the Prep job is done, and this is why it is super important to think good while doing that first stage of the project.

Tableau Desktop

The main tool that Tableau is known for is Tableau Desktop. This is a pure BI visualization tool that offers great range of functions and capabilities. It is by far the best tool on the market when it comes to dashboard design. Using this tool BI analysts create the analytical dashboards for stakeholders. Most commonly Tableau Desktop will be connected to data source prepared by Tableau Prep and existent on the server. However some companies select to bypass the data preparation step with Tableau Prep and connect Tableau Desktop directly to data lake/data warehouse sources. It doesn’t actually saves costs but it reduces the complexity of the architecture however at the expense of some opportunities to manipulate data.

Tableau Desktop is very flexible and can create almost anything in terms of dashboards. Best thing is that it allows to embed great interactivity of the dashboards. Also it is very easy to use since it is mostly drag-and-drop and can be used by even non-BI people, however for more complex tasks you do need data manipulation skills. One single drawback that I can point is the lack of ability for users to input comments on the dashboards, which is something that is frequently required in order to replace legacy Excel reports. However in the end if your dashboard is build the right way there will be no need for comments as the visualization should be self-explanatory.

If you are managing access on dashboard level this is where functions must be applied to create a RLS model. Also in the perfect scenario when using source from scheduled Prep job on the server the dashboard created via Tableau Desktop will automatically refresh with new data which makes it very self-serving.

Both Tableau Prep and Tableau Desktop are separate pieces of software that need to be installed and maintained by the IT support of your organization just like any other tool. Both of them are used by single analyst to create the necessary project. On the other hand Tableau server should be maintained by the designated server admins and used as the platform that unites the other users.

Tableau Online

You can choose whether your Tableau Server to be on-premise or in the cloud. Tableau Online is the cloud version of Tableau server hosted by Tableau. It has the very same capabilities of Tableau Server with the addition that like any other cloud service it has different cost model. If you have the Tableau Server on-premise obviously you will have fixed cost of maintaining it and also fixed space to work, but with the cloud version you basically pay for what resources you use and this makes your cost dynamic. In general Tableau Online should be more cost-effective, but it really depends on the right management and strategy.

Advanced tool – Tableau Einstein, Integrating Python/R scripts

Tableau also jumped on the AI frenzy and is developing their own analytical AI called Tableau Einstein which is part of the broader Salesforce service called Einstein. It is still in the infancy stages and it is very promising piece that can automate even more things. The idea is to connect it to data source and let the end users interact with it directly.

One great advantage of Tableau architecture is the ability to integrate foreign platforms like for example using Python/R scripts in the tools. This integration can happen both in Tableau Prep stage or directly in Tableau desktop. While native Tableau tools can do very good job in data manipulation and preparation, they lack the ability to do advanced analysis. This is where Python/R scripts come in as they allow to implement more sophisticated stuff like machine learning , regressions etc and then visualize it live in Tableau architecture.

Example: BI analyst is tasked to forecast revenue via neural network model. He will use Tableau Prep to connect to the revenue data in data lake/data warehouse and then integrate Python script with the decision tree model. After that the source (incl. the result of the model) is published on the server where it is visualized by dashboard created via Tableau Desktop.

How much the setup costs

Pricing of Tableau architecture (especially for medium to big enterprises) is very scalable. It does have very little fixed cost and predominantly it is dependent on the size of operations being done:

Basically your company needs to buy the license for Tableau server and then handle the different implementation costs if it is on-premise. If it is the cloud version then it also becomes variable as it based on the usage of cloud resources.

The other big cost pool is the licenses for users. There are 3 type of user-licenses that are available and we will dedicated another post in explaining the economics of it. But simply everyone who needs to use the dashboards/reports created in Tableau will need Viewer’s access and those who will create/edit anything in the architecture will need the more expensive Developer license. There is also the Explorer license that is somewhere in between, it does give editing rights but not everything. Licenses are provided on the user level so it is totally variable cost and dependent on how many users you infrastructure will have and what is the mix.