Bissantz Deltamaster and some thoughts on guided data analysis
It’s been a while since I’ve written about any third-party SSAS client tools here, isn’t it? This is partly because there aren’t as many of them around as there used to be; if I’m honest, it’s also because I find writing product reviews pretty dull as well. That said, I’m always interested to see demos of these tools, and my customers are always asking me about what’s available so I need to keep my knowledge up-to-date.
I’d like to point out before we go any further that his post is NOT a product review, but some thoughts that occurred to me after seeing one of these client tool demos.
Anyway, few months ago I was given a demo of a tool called Deltamaster, sold by a German company called Bissantz. Now if you’re reading this in Germany (or Austria, or Switzerland), you’re probably wondering why I’m writing about something that’s been around almost as long as SSAS itself – I’ve certainly known about it for years although I’ve never properly played with it. If you’re reading this elsewhere, though, you probably won’t have heard of Deltamaster because it isn’t widely sold outside its home market. It’s a traditional, full-featured desktop SSAS client tool tool that does all the things you’d expect a traditional, full-featured desktop SSAS client tool to do. It does PivotTable-like things. It does charts and sparklines. It allows you to save multiple views in briefing books. It has menus coming out of its ears and hundreds of different options for doing things. The UI is, if anything, a bit too busy and slightly old-fashioned looking, but it does everything you’d want it to.
What really caught my attention, though – and I’m sure this is a feature its had for ages – was the range of guided analyses it has built-in. With just a few clicks and the selection of a few parameters, you can do some very sophisticated stuff. Here’s the Concentration Analysis (aka ABC analysis) report that it produced for months on the Adventure Works cube, complete with colour coding, chart and all the working:
A distribution analysis (again, notice the stats in the box on the right hand side):
Even some impressive-looking data mining stuff that I don’t quite understand (I should RTFM):
I’ve seen this kind of thing before, but Deltamaster does this well and has by far the largest number of different types of analysis available. And all this made me think, why don’t more tools do this? Why doesn’t Excel feature this kind of functionality?
Data visualisation tools like Tableau have done well by making something that’s difficult and easy to get wrong – data visualisation – much easier, by pointing you in the right direction and stopping you doing things you shouldn’t be doing. You think you want a pie chart? You don’t get it, because pie charts are a Bad Idea. You get what’s good for you, not what you want. What Deltamaster is doing (and I think it’s an idea that could be taken a lot further) is the same thing but for data analysis, statistics and data mining. Now I know next to nothing about statistics and while I’m not proud of that fact, I’ve only managed to survive as long as I have in the BI world because my customers know less about statistics and data analysis techniques than I do. So far, the big struggle in BI has been to present the correct figures in a table with reasonable performance. The next problem in BI, once the data has been delivered, is to make sure business people interpret it properly. This is what good data visualisation tools do, and I think this is what guided analysis functionality could do as well. Are sales really going up, or is this seasonal variation? Is there a correlation between running promotions and increased sales? Does a customer’s gender, age, occupation or education level tell us anything about their likelihood of buying from us? At the moment there are plenty of BI tools that give us the ability to answer these questions if we know what we’re doing, but most of us don’t.
So, the key thing though is not to provide lots of types of guided analyses, but to make them easy to use and difficult to make mistakes with. If I was to criticise Deltamaster it would be because it provides a whole bunch of analyses that spit out graphs and stats, but it doesn’t go far enough to help you choose which type of analysis is right for your business problem, to help you choose the right parameters to pass in, and to help you make sense of the results which are returned; it’s still well ahead of most of the competition though. Would some level of user education always be necessary? Would the tool need to know about the data it’s working with, and the business problems associated with that data? To some extent maybe. I still think there’s a lot of room for improvement on what we’ve got today though.