It really depends of the size of the project really. For one client we have for instance ~5 facts and around 12 dimensions. For that kind of requirement the cost is much lower in term of development/maintenance, and free of licenses was a plus in that case (we're using activewarehouse-etl and an "Excel on top of ODBC" front-end).
I don't necessarily disagree with your statement that getting the best BI tool is more important than the cheapest.
I think far more important is for a good BI developer or someone in-house with deep domain knowledge and "power," for lack of a better term, who can clearly articulate what valuable questions are going to be answered by the implementation and tools and also drive the effort.
Of course, if your BI effort is aimed directly at customers and customer data analysis, maybe that is a good enough vision.
My experience is more on the resource management side of data warehousing, stuff like figuring out how to attach revenue/cost numbers to web applications and allow drill down across multiple dimensions.
At my company we've used Jaspersoft and Pentaho and at various times paid for commercial licenses in JasperSoft's case to get support and some consulting (which we only used once) and in Pentaho's case to get a bug fix patch faster.
Both tools are relatively cheap, you can expect to pay around $10K per year if you want the commercial version of the full suite and are very developer friendly. They both use Java extensively, so you have to be willing to have that in your environment.
I would add one more tool to mix - MSFT SQLServer 2008. The standard edition (about $2K, I think) comes with a the full suite minus the data mining tools. The reporting tools are on par with what Jasper and Pentaho offer and the Analysis tools are very simple to use and allow one to use Excel as a front end interface (which is what you end users want anyway).
OT: With this, and the story about phones being the best universal remotes[1], the straw has finally broken my camel's back. There are few ways to get me to hit the back button faster nowadays than to intersperse "See why [related-ish topic]" links between every other paragraph. I'm pretty sure the people who come up with those UX ideas will be the first ones axed when the bubble pops.
Hey, just to throw ourselves in the ring: We're Chart.io, a Y Combinator company. Our product is "Google Analytics for databases" and we've basically stripped out a lot of the complexity of BI (and a lot of the cost :)
We're in private beta now, so email me at dbeyer at chart dot io if you'd like to give it a shot.
+1 for Chart.io. We use it and love it. It addresses the most costly part of BI tools which someone pointed out above - paying one of your engineers to hook up your data with an external tool or have to code something custom from scratch.
Nice article. Other than open source, one more affordable option available today is SAAS BI. Especially for small and medium companies who may not want to allocate precious IT resources for installing and maintaining the open source solutions in-house. I work for Zoho and we have one such SAAS BI solution, Zoho Reports (http://reports.zoho.com). Do have a look.
One problem I have had with BI suites (specifically JasperSoft but it applies elsewhere) is the bolt-on nature of the solution. I found security integration, with a web app, to be a major pain. But the worst part was the UI and feature set. Honestly, I think modern web dev can be a real alternative to these packaged deals ... of course depending on the circumstances. What do others think?
Another issue I have is that the packaged solutions ~tend~ towards some kind of interface where they are eventually essentially presenting a join syntax to the user. This doesn't seem right to me. I think the value of the solution really needs to be at least a level higher than that. But when you think about it, what else is a generic tool going to do?
I'm really interested in other people's opinions on this. I recognize I am in the minority.
I am on the fringes of BI at my company, and do not claim to be an expert. But to me, the value in the BI suites is not the UI, nor the platform on which BI runs. It is the integration points to everything else. This article told me very little about that aspect of the open source tools.
Opensource BI might be interesting for small and medium-sized business, however for large companies price is rarely a disabling factor. Not to forget that Enterprise BI suites often go further than just reports and dashboards and include metadata management, lineage and impact analysis, business glossaries, predictive workbenches etc. Many of enterprise BI suites offer additional benefits from integration with existing infrastructure from the same vendor. So it's not only a matter of price for BI software.
Even for SME sector there are strong alternatives to opensource BI tools -- e.g. MicroStrategy offers its basic product for up to 100 users for free, there is free and very powerful QlikView Personal Edition, etc.
BI field has been moving slowly but steadily into mainstream as more and more businesses recognize the need to be able to analyze internally generated information.
What are the biggest unsolved challenges in the field as of today?
1) Is it difficulty in getting a project up and running quickly due to lack of in-house expertise?
2) Is it expense of getting all the moving parts (software/hardware/personel) together?
3) Is it lack of strong management buy-in?
4) Is it bad/incomplete/fragmented data?
...
A lot of what you're paying for with the "big bucks" vendors is support, maintenance, advice, and depending on your position in the business food chain, some visibility (presenting at their conferences, etc.)
Intangible: what about when your "intelligence" isn't working so intelligently, would you rather be angry with your internal team or some vendor?
In a lot of cases open source BI makes sense, but the cost-benefit analysis should include things other than money.
Support is overrated. My experience with big bucks companies is that even if you do have support, it's very hard to actually get your money's worth out of it. If your problem isn't in an FAQ, you can get stuck in an endless cycle of "we'll get back to you on that".
Unless you have ironclad service-level agreements, good luck getting a big vendor to solve emergency problems for you.
I coded in SAS for a long time and I can say unequivocally that their support was absolutely worth it. I have had techs work practically around the clock to find fixes and offer suggestions. MathWorks is about the only other company that I've dealt with that had that level of responsive support.
I agree, having the source and being able to fix a bug myself is far more valuable than having big blue tell me its not a problem, then me proving it is, then spend months waiting for a promised fix, then the fix breaking two other things...
Anything that has the quality that you don't "get your money's worth" is by definition bad. Simply stating that something as general as "support", which is run differently at every company, has that quality doesn't aim to prove any point.
All depends on the complexity of your project. We tried using open source BI but it just didn't accomplish what we needed. QlikView, on the other hand, has been a great solution. My only concern with QlikTech is that they're a bit arrogant and don't know how to price or write license agreements that support scalability.
It may not be for you, unless you have a some operational complexity behind your web apps. Otherwise your favorite web analytics service will do the job well.
It's ideal for manufacturing companies, companies with some complexity in their supply chain, companies that wish to maintain service standards (e.g., a call center), have sales teams with measurable objectives, etc.
In other words, it's for companies that look more like traditional businesses in some way, versus a super asset-light web app company. By revenue, that is the majority of businesses in the US.
$40k for SW
$500k for development (they brought in a BI shop to get their warehouses setup)
$120k/year for a BI developer
So the delta for making the SW go to $0 just isn't very motivating. Getting the best BI tool is far more important than the cheapest one.