DiY BI - Toys, Tools and Power Tools
When I was a kid, I received a set of toy tools as a Christmas present. They were great, I could twist in plastic bolts and screws, "saw", rachet, hammer in thumbtacks; I had a blast as a kid playing with these toys and acting like I was really building something.
Of course, I couldn't actually build anything with these tools, I could only pretend to build things. So, when I graduated from college and got my first apartment, my dad took me out to the garage and put together a set of real, honest-to-God tools, a hand saw, several different screw drivers, a level, pliers, channel locks and a bunch of other stuff. Channeling Jeff Spicoli, "My dad had an awesome set of tools!". In fact, I never really understood why he had so many tools, an entire garage full of them, most of which I had never seen him use.
These tools served me pretty well starting out, I could hang pictures mostly level, make simple repairs around my apartment and even put together quality, pre-manufactured Swedish furniture from time-to-time. But, when I purchased my first house and started becoming more of a "Do-it-Yourself" (DIY) kind of guy, those simple tools that my dad started me out with just didn't "cut it", in manner of speaking. I mean, I wasn't going to build a deck by cutting 4x4's, 2x4's and 2x6's and screw in every blasted deck screw by hand. No, the manual, "starter" set of tools wasn't going to do the job and that's when I started using power tools like a corded drill and circular saw. Here now, I had a set of tools that I could do real work with and build real things.
OK, so I would have to imagine that by this time there is at least one reader out there wondering, "What in the name of all that is holy does this idiot's infernal ramblings of toys, tools and power tools have to do with business intelligence?". Glad you asked, I was just getting to that...
Here's what I found out about power tools. I should have started there. The funny thing about power tools is that they can handle the real work like building a deck but they can also handle the simple chores, like hanging a picture on drywall. These days, sure I could hang a picture with a screw driver and a level, but I tend to reach for my laser level and drill. And hence, I now understand why my dad had a garage full of tools that I never saw him use, because now I'm the one with a garage full hand saws, screw drivers, ratchets and other assorted manual tools that I rarely, if ever, use. Its because now I have power tools that can do the job better, faster and have a wider variety of applications.
The parallels between my progression with tools and today's self-service, or what I like to refer to as "Do it Yourself" (DiY) business intelligence (BI), are striking; particularly when it comes to the ongoing debate about "which BI tool is better". Some of the more recent dialog in this space has been about Tableau vs. Power BI. This all started with Tableau posting their "10 Ways Power BI Falls Short" post, although now it seems to be 8 except the slide deck still states 10. And, predictably, there was some response to that and then even a rebuttal in the ways that Tableau falls short. I'm not here to rehash all that, I'm here to explain why, when it comes to DiY BI, Tableau is a toy while Power BI is a power tool.
Quite simply, when it comes to doing data cleansing, data mashups and data modeling, Tableau simply can't cut it. You can't build real things with Tableau whereas you can with Power BI. I've recently heard this from several new customers, one of whom actually stated, "We're regretting our decision to go with Tableau because it can't handle complex data models. Once you get beyond one fact table and one dimension, it's next to useless.". But, don't take my word for it, even Tableau advocates say the exact same thing. Granted, that article does show that Tableau can handle a single fact table and; count them, three whole dimensions, but requires "an expert's knowledge of Tableau" to accomplish.
So, if Tableau requires an expert's knowledge to pull off a single fact table and three dimensions, what level of knowledge is required for a recent model that I worked with Dallas Salazar to put together around commodity pricing correlations to organizational health in the oil and gas industry:
This model has three fact tables, multiple dimensions and a couple of many-to-many relationships thrown in for good measure. Furthermore, the visualization of the model is crucially done by Power BI's SandDance visual:
The reality is, Mr. Salazar's thesis could not have been realized and presented without the SandDance visualization.
Finally, the data with which the model was put together was certainly not "clean" data. The data source files had to be appended together, unpivoted, mashed up and manipulated, operations that Power BI's Power Query handled easily but are impossible in Tableau.
The bottom line here is that Power BI's data modeling engine is SQL Server Analysis Services (SSAS), an enterprise-class data modeling power tool. SandDance is the power tool of all power tools when it comes to advanced data visualization. And Power Query is a data ingest, manipulation and mashup tool whose data transformation capabilities rival enterprise class ETL engines such as SQL Server Integration Services (SSIS) and Informatica.
The reality is, when it comes to getting real work done, Tableau is a nice toy that can certainly be used to solve some problems but Power BI is the power tool that makes it easier to take on the largest jobs and save you time in the process.
I do not want to see customers end up with a garage full of tools that they never use, I'd rather see them buy the power tool and forgo the toys and manual tools that are only useful in highly specific circumstances. Of course, that's where the analogy breaks down. Whereas power tools cost much more than toys and manual tools, Power BI is free while Tableau will cost; well, let's just say it will cost you way more than a circular saw...