About Tableau 8

Jen Underwood highly recommended I attend the Tableau 8 tour on Tuesday to witness firsthand its new features. Naturally, I couldn’t resist of comparing everything I saw with Microsoft BI. I took some time after the pretention to take 8.0 for a spin and reconfirm my understanding. Here are the top five things I liked about Tableau.

  1. Simplicity – A few months ago, I blogged about my top 5 Microsoft BI wish list. My number 1 wish was a continued focus on integration and simplification. Tableau 8.0 nailed it down as far as simplicity, at least on the visualization side of things. One desktop tool and its server-based counterpart. A tool whose sole purpose is BI. Not something that was bolted on another tool as a BI add-on. No layers to integrate with and configure on the server side, and error logs to sieve through.
  2. Visualization – Visualizations are yet simple and powerful. The presenter was quick to point out the areas where the tool excels in comparison with Power View. Tree maps and bubble maps were all over the place, as well as the ability to customize them, e.g. change the color of a pie slice. I liked the ability to dynamically group items (similar to custom MDX sets in Excel). For example, the presenter lassoed a bunch of cities in North America and created a North America dynamic group that was subsequently used to analyze sales. Another interesting in-the-box feature is ability to forecast data. In the Microsoft world, this would require some flavor of data mining (not a native feature in both Excel and Power View).
  3. Mobility – This is one area where Tableau has at least a year lead over Microsoft BI (see my number 2 wish item in the above blog). Once the dashboard is published to the server, it can be viewed and edited on mobile devices (iPad was demonstrated). By contrast, due to its Silverlight nature, Power View is currently supported only in Internet Explorer. However, as we’ve heard at conferences, Microsoft is hard at work to change this.
  4. JavaScript API – Developers creating web applications can integrate and embed interactive Tableau content into their applications via the new JavaScript API.
  5. Data reach – Apparently, Tableau customers are asking for specialized connectors to cloud data and emerging data sources. Version 8.0 introduces connectors for Salesforce.com, Google Analytics, Hadoop, SAP Hana, etc.

Being a self-service BI tool, in my opinion the backend is where Tableau trails behind Microsoft BI, including:

  1. No continuum for self-service to organizational BI – While you can publish a workbook to the server, you can’t upgrade the workbook to an organizational BI model. Nor can you connect to a published workbook as a data source, e.g. by using Excel as a front end.
  2. In-memory engine – Based on my experiments, xVelocity (the in-memory engine of PowerPivot and Tabular) excels the Tableau in-memory technology both in compression and speed.
  3. Programming – Tableau supports custom calculations but the potential is not even close to what you can do with DAX. For example, I couldn’t find a way to use many-to-many relationship (not natively supported in Power Pivot but can be handled with DAX formulas).
  4. Scalability – I haven’t tested the server edition yet but I would expect Analysis Services to scale much better than Tableau due to the desktop origin of the latter.
  5. Security – Tableau supports action-level security, e.g. allowing the user to edit workbooks, and basic user filters that allow user access to specific members, such as Bob can see only USA. There is no dynamic data security.
  6. Usability – Besides the presentation layer where Tableau excels, I’ve noticed areas that would be surprisingly difficult for a self-service BI tool. For example, there isn’t option to visualize tables and relationships in a diagram view. For some obscure reason (I guess so it knows what to aggregate and not), Tableau insists on separating entities as dimensions and measures. Tableau stores data and connection definition as external files, and maintains connections per entity. For example, I couldn’t find a way to import more tables using the same connection definition.

In summary, choosing between Tableau and PowerPivot for self-service BI would require careful examination of requirements and comparing pros and cons of the two. On the corporate BI side of things, you should definitely consider Tableau as a front end to Analysis Services if you need an alternative to Microsoft Excel and/or Power View.

  • Thank you Teo for attending the Tableau Roadshow. There are a few inaccuracies in your assessment that I’d like to correct.

    Tableau and Power View may look similar but are completely different with regards to available functionality, both breadth and depth of functionality, and the types of use cases each tool can fulfill. Power View, intentionally by design tenets, is not a deep analytical tool. Tableau, by design, has much deeper analytical functionality than Power View. Tableau can leverage PowerPivot and SSAS as data sources – and – many Tableau customers do use both together. It is not a one or the other choice.

    Tableau’s sematic model is a “shared data source”. The label is misleading. The shared data source is the semantic layer. Keep in mind that Tableau is also often used on top of PowerPivot or SSAS models. There is a whitepaper on this topic available.

    It is not Tableau’s desire to replace xVelocity, ParAccel or other in-memory players. Tableau’s focus is on the front-end, deep analytics, statistics, and visual analysis.

    Tableau is highly scalable. It has a distributed architecture and I suspect there is some education that needs to take place to learn how Tableau scales. here is a link to the docs http://onlinehelp.tableausoftware.com/v8.0/server/en-us/help.htm#distrib.htm

    Tableau does have much deeper security than you describe. Again I think there is some education needed. Tableau can leverage security defined at the data source such as SSAS – or Tableau data level security called “user filtering”. See http://onlinehelp.tableausoftware.com/v8.0/server/en-us/help.htm#security.htm

    Regarding usability, Gartner cites Tableau customers as having the fastest solution time to value over all the 20+ BI vendors for several years now. I would counter back that I have seen my fair share of business users abandon PowerPivot because they can’t figure out basic time intelligence that is out-of-the-box with Tableau and dimensional modeling. Most business users don’t have a clue what dimensional modeling is. They rarely add a time dimension table unless a BI developer showed them. In almost every PowerPivot POC, I saw business users fail until a partner was paid to come in and train them. It is not Tableau’s goal to replace PowerPivot but I would say pointing Tableau at a data srouce and getting functional is far easier and faster than any of the alternatives out there.

    Other key reasons why I liked the Microsoft BI + Tableau combination are as follows:

    – Browser mobile for all platforms and Native Mobile Apps for Andriod and iPad
    – Web Authoring that also works for mobile
    – No AD or SharePoint required, and a variety of APIS for ISVs that need to offer BI in their applications
    -Multi-tenancy and external reporting, something that is really difficult if nt impossible in Microsoft BI world
    -Rich array of data source connectviity with enhanced drivers, Tableau has teams that build additional functions on top of base drivers
    – Deep geospatial mapping and layering capabilities – things like drive time radius, weather overlays
    – All views can be payed over time – not just scatter
    – Deep statistical, linearn and nonlinear modeling
    -Visual grouping, viewing details, exporting details, enhanced tooltips with charts on them or actions with parameters to launch related web apps/views like a Salesforce.com record
    -Ability to render more than 1,000 data points (Power View limit) : Tableau can render over 1,000,000 data points
    -Subscriptions and personalized view saving
    – Wonderfully easy server installation that can be done in a few minutes that contrasts with minimal couple days of SharePoint, SQL Server, Kerberos, AD and all the related tricky, time consuming and complex installation joys of Microsoft BI world.

    I LOVE Microsoft BI but honestly Tableau offers the front-end that Microsoft has been sorely lacking since Proclarity was killed. Excel is just not enough. Tableau and the rest of the data discovery tools have explosive, over 100% YoY growth. Microsoft back-end with Tableau front-end is being widely used in the largest enterprise accounts. If you practice Microsoft BI, than you should also have Tableau in your skillset. There is very high demand for that combined expertise – thus I invited you to learn and get a piece of that business too.

  • Jen,
    Thanks for chiming in. My intention wasn’t to provide a full-featured comparison and I don’t want to turn this into some kind of a vendor diatribe. Obviously, there is a place for both vendors. About the observed inaccuracies:
    1.Scalability – I wouldn’t expect a file-based tool to scale, just like a PowerPivot can’t scale more than 2 GB. The point was that IT can upgrade a PowerPivot workbook to an enterprise model. We don’t have this option with Tableau.
    2.Security – I was referring to dynamic data security which is a common enterprise need.
    3.Dimensional modeling – I don’t see how Tableau (being a modeling tool) shields the user from having to know modeling and relationships. True, users struggle with these concepts but Tableau is no less “dimensional” that PowerPivot is. In fact, I would argue that it’s actually more dimensional as the user needs to know about dimensions and measures.
    4.Easy installation – You have a point here as I mentioned in my blog. Just to point out that there is a good reason to have SQL Server thrown in the mix. Also, from what I can tell, the price pay for configuring the SharePoint and SQL Server combo is less than Tableau server (pricing not disclosed) but you’re getting much more. The question is where you really need the SharePoint features which is a topic on its own. On the Kerberos side of things, you don’t need Kerberos for PowerPivot for SharePoint (equivalent of Tableau server). And, I don’t see how Tableau won’t require Kerberos for the double-hop (Tableau to SSAS) scenario if data security is needed. Finally, you don’t have to configure Kerberos for the Power View -> Tabular scenario. I can see that some education is needed here 
    5.Basic time intelligence – Not sure why end users struggle. Excel pivot reports have been supporting time intelligence for a while. The only thing needed is to mark a Date table. True, Power View doesn’t have it so it has to be added to the model.

    “I LOVE Microsoft BI but honestly Tableau offers the front-end that Microsoft has been sorely lacking…” Things are changing….but I really wonder how much from the Tableau visualization prowess regular business users actually use. I’m inclined to contribute the success of the “explosive” growth of the data discovery tools to very effective marketing (which in many cases borders with outright lies, such as that data warehousing is not needed, OLAP is obsolete, self-service BI is a panacea, etc.) and glitzy front ends. As you know, I have my own stories to share….

    Thank you for being an avid reader!

  • Hi Teo!

    Haven’t been to your blog in years, but it was one of my faves while I was at Microsoft.

    I saw a link to this post on the Twittersphere and just wanted to chime in on the first couple points that you made.

    Tableau doesn’t actually need to load all the data that you want to analyze into RAM. It loads only the columns (or portions of the columns) necessary to answer the question being asked. In a worst case scenario, it’ll actually “page” much like an OS.

    Therefore, it’s not constrained by the amount of RAM on the client. I regularly work with customers using 250M – 1B rows on the client — data that would take up way more space than the 4-8 GB of RAM on a commodity laptop has to offer

    Tableau also connects “live” to data sources and doesn’t need to “own” that data – so to us, upgrading to an Enterprise Data model is largely a no-op. Our users will point directly to the big SQL/Oracle/Teradata/Whatever box out of the gate and have at it…

    Tableau security is also pretty slick — and flexible. You can provision security rules in a number of ways:

    – Simple “User / Group XYZ can see dimension members [P, D, Q]” selections – configured via a dialog.

    – Simple calculations (sort of like SSAS “Advanced” dimension security) to determine which dimension members / measures / cells are available

    – Moderately complex calculations which utilize dynamic “security bridge” tables which define who can do what — the same basic idea as “Security Labeling” ala the SQL Server Security Labeling Toolkit.

    I don’t generally position or use Tableau as a data modeling tool. To me, joining a few tables together and adding some new columns is more akin to throwing up a quick and dirty ODS vs doing true modeling. A Tableau extract is not a replacement for a true DW.

    I personally feel that when it comes time to “go enterprise”, one needs an IT sponsored / created model vs. something a user put together themselves. Call me a traditionalist if you must :)

  • Hi Teo. Love your SSAS 2012 book and your online videos of Tabular modeling. Really appreciate all you do for us BI consultants in the Microsoft world. Thank you for writing this post.

    As a Microsoft consultant myself, I’ve worked closely with a couple clients now that use Tableau and am intrigued by how fast they are able to create analysis, and drill-through to the detail data without having to write any code, or subreport, or drill-through action, or without the help of some 3rd party add-in.

    I’ve also been mindful of big enterprise clients that not only use Tableau as their enterprise analytical tool, but they are also some of Microsoft’s biggest enterprise customers. The ones that come to mind from the last PASS Summit are Yahoo and Klout. Both use SSAS for its performance and semantic layer, but then are able to deliver powerful dashboards and visualizations that can be shared with the rest of the organization.

    The point here is that one cannot compare Tableau to Microsoft BI as a whole. That wouldn’t be fair to either vendor. Tableau can only be compared with what its intended focus is: Analytics, Visualizations, and distributing that discovery to the enterprise. That means we’ve got to compare Tableau with SSRS, PPS, and Power View. Clearly, we are at a disadvantage with our dashboarding tools being completely seperate from the tools used to create OLAP data sources. So to Jen’s point, this allows customer’s to gain value faster just having to develop their semantics in one integrated tool. No seperate add-in or different SKU product. No multi-developer languages to learn: MDX, DAX, DMX, SSRS Expressions, PPS Time Intelligence, etc.

    Now, that’s not to say its all fun and games with Tableau. Tableau has certain things its not good at, like tables or lists. Like, you would never write an invoice report in Tableau, or create a label report for printing bar codes. However, that’s not intended focus.

    What is also apparent is how easy it is for Tableau Server users to share their work and have great drill-through, sort, and filter capability out-of-the-box without a line of code, subreport, or add-in. Tableau Server is easy for a non-IT person to install, and they can start sharing workbooks to the enterprise right away.

    I read your other post of the things you wish Microsoft BI had, and of those, Tableau has got the simplicity thing down. More importantly, it knows who Tableau is. Its listening to customers and their needs instead of developers, as you indicate. So there’s a lot Microsoft can learn from them, IMHO.


  • Great comment, aabundez. Yes, as I said, Tableau+SSAS is a good combo. But you brought a great point about focusing on the visualization side only. This of course brings another question about what BI is? When Gartner plots Tableau and QlikView SxS with the mega vendors, do they focus on visualization only? I bet they throw the self-service BI story in the mix as well, plus organizational BI (dashboards). So, I felt like I’m doing MS injustice to focus on the visualization part alone.

    About MS presentation layer slacking behind, we should all push to change this. MVPs and partners apparently are not enough.