Tableau Server

If you're in an organization that is all about having absolute command and control over the data and how it's used, Tableau may not be for you. Tableau has a commitment to democratizing data and an incredible focus on the user experience that is very much in line with our organizational values and goals. Tableau has made a business tool that is easy to use, incredibly powerful, and delightful to use. Doing exploratory analysis of fresh data with Tableau is like skiing in fresh powder, and Tableau Desktop is my favorite piece of software, ever.

We are very satisfied with what it has enabled us to do. The only factor in us not having more Tableau licenses has been budgetary constraints as a non-profit.

My department was largely Excel and paper-based when I started, so there's been a tremendous amount of work to get the data into shape, so when I look at the overall implementation of our visualization and reporting infrastructure, I'm less satisfied. Tableau has been wonderful, the one place that would get me to a 10 is if it was easier to tune workbook performance on Tableau Server, there are a huge number of variables to deal with across the workbook/application/hardware stack. In a small environment like ours that takes more effort than I'd like.

First, a caveat: I'm someone who reads the manual, watches the videos, opens 20 tabs in my browser from a set of Google searches, and tries everything I can think of before contacting support. When I need Tableau tech support it's either for some newbie question that I missed the boat on (in which case support is awesome) or some deep in the weeds question about an undocumented piece of functionality that I might not be using for its intended purpose. I've sometimes had a hard time getting the first layer of support staff to understand what I'm doing, it seems like their knowledge isn't as solid as I'd like to be, and I chalk that up to Tableau's rapid growth. However, second level support has continued to be excellent.

While Tableau has the highest ease of use for analysis, the data set has to be ready for that and Tableau hasn't (yet) brought that ease of use to the ETL process.

I use Tableau for one to several hours per day, every day, and it almost never crashes. It's the most reliable piece of production software I've used.

While there are definitely cases where a user can do things that will make a particular worksheet or dashboard run slowly, overall the performance is extremely fast. The user experience of exploratory analysis particularly shines, there's nothing out there with the polish of Tableau.

Tablea is a strong Clinical/Business Intelligence application. Tableau is a good fit for healthcare organizations.

We have been using Tableau for three years now and each new version offers move features and functionality. Tableau grows as our user base grows. Tableau Server is the only product in my experience that takes less than 15 minutes to upgrade.

Some additional details on the Tableau project I worked on at Southern California Edison (SCE):
During implementation, I was heavily involved with the Tableau consultants, stakeholders, and project managers to make sure that all requirements were met. I did testing myself and signed off on testing that others did. I provided my manager with constant feedback on our progress. We were able to get Tableau installed in one month based on a very aggressive deadline set by management. During this phase, I spent countless hours side by side with Tableau consultants to describe the performance issues we were facing at the time. Here are two of the main hurdles we faced during implementation: One of the main issues we faced was the use of Custom SQL to build our visualizations. Because the roles within my team were highly segmented, we preferred using Custom SQL vs. connecting live to the data. The reason for this is because I was tasked with building the visual. I was not tasked with building the query. The DBA I worked with would hand me the query and I would build the visual from there (creating the necessary calculated fields and formatting the data). Tableau advises its clients against the use of Custom SQL for performance reasons. In addition to performance, there are other drawbacks. We would notice that certain functions in the SQL statements themselves would not get translated correctly into Tableau’s Native SQL (there are workarounds for this). This issue was starkly apparent when connecting to Teradata. Towards the end of my role at SCE, I was working daily on improving the connection to Teradata. Before I left, I was able to take certain reports built in Business Objects and build them in Tableau (but this was only accomplished when connecting live to the data). Another issue was refreshing certain dashboards that were meant to run 24/7. Tableau has the capability to refresh the data behind the visual as frequently as every 5 minutes (I believe). But that refresh does not update the visual. In order for the refreshed query to update the visual, you are required to force the browser to update on an interval of your choosing. It is only with this combination that you can come close to seeing what is happening in real time. The workaround for this issue was to create an .html file that had the URL of the dashboard, which was published to Tableau server embedded in it. From here, you simply double click on the .html file and you would have your dashboard updating both the data (handled by Tableau Server) and the visual (handled by the browser).
Here is a list of the data sources that I would connect to when building visualizations:
1. MS SQL Server
2. Oracle
3. MS Excel
4. Teradata
5. MS Access
Data Blending – There were several instances where business requirements made it necessary to join data from two separate data sources (i.e. - MS SQL Server and Oracle or MS SQL Server and MS Excel). Tableau has an intuitive capability to recognize a data blend. This will be indicated by the use of the link icon next to the primary data source (it looks like an infinity sign). The ability to blend data together was very appealing in the conceptual sense. But there are limitations on the number of rows that you can blend together (I noticed a significant deterioration in performance after 100k rows). This seemed to always be our problem at the SmartConnect Operations Center. We would relentlessly push Tableau’s performance only to find out that our local machine didn’t have the RAM necessary to do what we wanted, or that Tableau was not intended for that specific application. Regardless, I learned a tremendous amount about how the application behaves in different scenarios.
Incremental Refresh – Sometimes it can be useful to use a Tableau generated extract. This is essentially a mini-Tableau database. My manager tasked me with building a report that would store historical data. The purpose of this report was to keep a running total of certain events in the network. The benefit of this report would be that it would automate a manual process that the end users were responsible for on a daily basis. Prior to the implementation of this new report, the end users were tracking the historical data using MS Excel. The challenge I faced was to find a way to store data from the daily query results. Because the original data source did not store historical data, it was necessary for me to take advantage of the “Incremental refresh” option available in Tableau. I do not advocate this method if the daily query results are on the scale of several hundred records or greater. This solution was feasible because the number of records were less than 100 per day. In addition, I worked with the end users and their team lead to create a backup method to get to the final row count in the instance that the Incremental Refresh option failed.
Experience with Teradata – SCE’s 5 million meter network produces around 15 million records per day in just one its tables. The data warehouse team built a set of views that could be accessed within the Teradata schema. After several attempts to connect to Teradata (unsuccessfully) using Custom SQL, I worked with both Tableau’s technical staff and SCE’s DW team to diagnose the problem. We spent multiple WebEx sessions documenting application behavior under various scenarios. Initially, my thoughts were that the problem was that the client application was trying to connect to a View versus a hard coded table. But Tableau assured me that the client application has no problem connecting to views. In this instance, the workaround was to connect live to the data. This worked on certain tables (with several joins in place) but not all. Finally, I was able to re-build a few reports taken from Business Objects. In order to do this, I extracted the SQL from the BO reports and then connected live to the same tables. The result was a visual that came very close to what was found in BO. The report reflected data that was aggregated in the form of a percentage. The percentage represented the performance of a particular district over a time interval (i.e. – 98.99% over a 3 day period).
Integration with ArcMap – The SmartConnect Operations Center’s business requirements made it necessary for me to find ways to enhance the mapping capability within Tableau. The default background map that is offered by Tableau falls short in the areas of zooming and satellite imagery rendering. Therefore, Tableau offers its users the ability to connect to a geospatial (WMS) server. To Tableau’s credit, the steps and information needed to connect to a WMS server are very simple. If you have the URL of the WMS, then you can connect to the server. In my experience, the biggest problem with using the layers from various WMS servers is in the rendering of the image. If your application requires a certain level of detail and/or the ability to zoom in to a certain extent, you may find that the static image stops rendering at a certain zoom level. In the SOC’s application, it was necessary to validate the latitude and longitude of meters in the field and their proximity to other meters (aka – the “mesh” network in wireless meter terminology) by visual inspection. When I left the SOC, I was in high-level discussions between Tableau’s technical team and SCE’s GIS team regarding the rendering of the different layers available through SCE’s internal WMS servers.

I can only speak for myself. Even though it has its shortcomings, Tableau is headed in the right direction by making large amounts of data available to the traditional business user. I dont think its simple to learn Tableau; but its not extremely difficult either. With some training and help from Tableau support, you can improve your ability to provide analysis to your team or department.

We got it working which was the main goal, but we had to create a workaround for a key dashboard that would run 24/7. In order for the view to refresh every X minutes with the updated data, we created a .html file that instructed to the browser to refresh itself every X minutes. This was a pretty simple workaround but it was something we were hoping to avoid.

In our case, they hired a private third party consultant to train our dept. It was extremely boring and felt like it dragged on. Everything I learned was self taught so I was not really paying attention. But I do think that you can easily spend a week on the tool and go over every nook and cranny. We only had the consultant in for a day or two.

The sales consultants do an amazing job of introducing the tool and its capabilities. They are also helpful in explaining the layout of the desktop client and its different functionality.

Keep in mind that they use a sample data source (MS Excel) with a very small amount of data to show off what it can do. What you have to remember is that you are buying the tool so that you can connect to large amounts of data (and possibly blend data together from different databases).

I think the folks that work in support are generally pretty good at what they do (when you get them on a WebEx). But the process of reporting issues to them and waiting for a response (via email only) is a hassle. I never understood why you can't just call them up and discuss the issues with them. It would take a handful of email exchanges before they would agree to a WebEx session. That was frustrating.

When it works, it works great. The drag and drop functionality is seamless. Creating calculated fields is also simple.

Tableau Server is very stable. But Tableau desktop has crashed (or become unresponsive) on a few instances.

Again, assuming that you are working with 100k rows of data or less, it works great. When you increase the number of rows of data by a magnitude of 5 or 10+, there is a significant deterioration in performance.

It's a fundamental part of the operation of the business.

It was fine - I did it myself which was a bit nerve wracking at the time but nowadays I don't even think about it.

The Tableau website is full of videos that you can follow at your own pace. As a very small company with a Tableau install, access to these free resources was incredibly useful to allowing me to implement Tableau to its potential in a reasonable and proportionate manner.

In all the years of dealing with Tableau, the support has been exemplary.

It's extremely easy to use. As the depth of its abilities have increased, I think Tableau has done a very good job of maintaining that usability - there are some aspects though that are not as intuitive as the rest of the product - table calculations for example.

In 5 years of use as a customer and 3 years of use of a separate install of Tableau as a Tableau Partner there has been close to zero downtime of the product and it requires no maintenance whatsoever between upgrades.

Tableau Desktop is extremely impressive at rendering data on screen and performing demanding calculations. Users need always to bear in mind that their underlying database and the hardware that sits on are responsible for much of the speed (or not). Tableau extracts (high performance, pre-aggregated, columnar snapshots of the data) help deliver performance improvements that underlying databases may not give. Tableau Server can feel slow to render data in comparison to Desktop but this gap is shrinking all the time.

We host several solutions utilizing Tableau Server being consumed by several global teams.