Tableau’s pretty cool, right? Self service BI for departments tired of waiting for years for the EDW to come online. Connect to the source systems, mash it all together – WOW – we have some BI!!
Nearly – but not quite. Tableau’s absolutely fantastic at getting started – finally, giving departments some good quality BI and getting people away from hacking reports together using Excel & Powerpoint – and for many organisations, that’s enough.
But – there’s a bunch of gotchas.
- Tableau becomes the data hub – with its huge “cache” engine
- Tableau server needs to be a mega-horsepower monster – or it’s slow
- It’s not creating a golden source repository – Tableau works with Tableau only and can’t be consumed by other reporting & visualisation tools
- Tableau server doesn’t work with Essbase (to add more performance)
- Tableau client DOES work with Essbase – but with severely limited data modelling capabilities.
- That all “financial intelligence” needs to be modeled directly in Tableau
- Its library of visualisations is decent – but not amazing.
Now – for the record, Tableau’s done a really nice job of implementing MDX support for Essbase and done a very elegant implementation – but it’s just not the heart of the product, it’s a bit of an add on.
Why would you want more than Tableau?
Let’s look at some performance statistics. These are produced on a like for like basis, comparing Oracle DVCS on an Essbase datasource to Tableau on a relational datasource.
In both cases, caching is off. (Caching is a cheat – it’s the same thing OBIEE guys used to do to mask poor solution architecture. Caching should only ever be the cherry on top, not the core)
- OAC Essbase/Smartview Performance : <1s
- OAC Essbase/DVCS Performance : <1s
- RDBMS/Tableau : 22s
This was running the exact same data-sets, asking the exact same question of the data. We made sure it was a “hard” question forcing a full aggregation of all underlying data – but the results speak for themselves.
What about Tableau & Essbase?
On paper, this looks really interesting. Leverage the visualisation capabilities of Tableau and underpin it with the performance of Essbase. Unfortunately, only Tableau desktop support this – and while it works very well, we’re thinking bigger here and really want to be able to run on the server and give access to everyone in the org.
Does OBIEE figure?
It’s a gotcha – OBIEE is really powerful, but the power it offers is hidden by high technical complexity and, at best, a clunky UI. It was of its day – and is still incredibly powerful. But too many projects use the power to mask poor data modelling – which then gets worse day by day. So OBIEE IS the powerhouse, but regularly it’s shockingly badly implemented – it gets a bad rap. Drive your car into a wall and it’s the drivers fault, not the car.
What’s our recommendation?
Let’s take a quick look at our requirements here:
- We want lightning fast performance
- We want to leverage the maximum power of the underlying databases and servers
- We want the ability – when needed – to federate distinctly seperate resources
- We want report writers & builders to spend their time focusing on reporting and anlaysis, not data modelling – they don’t care where the data comes from, as long as they can consume it
- We want different departments and teams to be able to have their own subject areas – but still consume from the same golden sources and adhere to standard business model
I’ve really only found 1 solution architecture that meets all these needs. Starting at the top:
Oracle Data Visualisation
(either Cloud or Desktop). This is the new kind on the block. And it’s lovely. It’s still really new, so perhaps not yet quite a richly featured as Tableau, but the usability of it, combined with the out of the box library of visualisations is awesome. And if you need more – there’s an app store where you can down load more visualisations and plugs. Need even more capability? You can write your own visualisations and Oracle Enterprise R plugins to stretch it even further.
The Data Engine
Essbase & RDBMS. Use Essbase for aggregated numeric Data & Relational for low level, transactional and non-numeric data. For extended datasets, also consider plugging in to a hadoop cluster – although I really find those are edge cases, despite the hype of hadoop from a few years back.
The Modelling Tool
This is the real powerhouse – but also the one where it’s important to have the best skilled team. The Oracle BI Server, with its unbeatable RPD data model sits right here. Nothing beats it
This gives you the best solution architecture:
- It’ll beat Tableau all day long
- Allows Ad-hoc & dashboard visualisations – all from the same reporting source
- Can outperform any other reporting solution with its capacity & performance
- Enforces all the “good” best practices (the ones that really help people)