February 23, 2011

3 reasons why you should use wiki for BI deployment

As I mentioned in one of my previous posts, BI vendors do not pay much attention to collaboration around BI deployments. However, this doesn't mean that large BI deployments don't have difficulties (or at least inconveniences) with collaboration and knowledge exchange. Some of these problems can be resolved using wikis:

First, and the most often problem -- lack of documentation convenient for both business users and technical team. Traditionally, project documentation (like specifications, glossaries, scopes of works, etc.) is done using Word/Excel documents. This results in network folders filled with tens or even hundreds of documents without any browsing possible. No need to say, that average business user will never look there. As BI-platforms doesn't have any good collaboration capabilities (excluding may be IBM), so user doesn't have many choices if he/she needs to know the logic behind certain indicator in a report -- only to ask somebody from tech team. In case of 5 users and 10 reports -- this is not a problem. But if you have 2'000 users and 10'000 reports -- this would be a problem. Often users, having such obstacles, just don't want to dig into details. And then we have low user adoption. So, reason #1 -- you need wiki to have the documentation searchable, manageable, consistent and convenient for use by business users first of all.

Reason #2 is that users usually don't have good How-To manuals. BI vendors usually do rather good manuals for developers, but in majority of cases they do not produce good illustrated manuals for business users. How to drill data, how to join data from two data sources, how to make ad hoc queries with subqueries, etc. etc. -- for all of these how-tos users need simple easy-to-understand illustrated manuals. And, what is important -- these manuals have to be easily extensible to target specific problems, if they occur.

And, finally, reason #3 is that large BI deployments typically have several BI tools. All of them have more or less decent portals for their own content, but none of them can hold BI content from other BI platform. Cognos knows nothing about BusinessObjects reports or QlikView application, Oracle BIEE knows nothing about Tableau and so on. At the same time a user might need access to several BI suites to perform his/her daily tasks. This is why you may want to have single subject-oriented (not tool-oriented) portal that would have links to various BI-content, or even have BI content embedded just in wiki pages. The latter may require some web-development works, as not all BI tools allow easy content embedding, and setting up integrated security, but it can also lead to much better convenience and user adoption.

There are a lot of wiki engines available -- free and commercial, easy and complex, etc. Wikimatrix can make the task of choice a bit simpler. I found DokuWiki to be perfect tool for majority of cases.

February 21, 2011

CV as dashboard

A few weeks ago I read on HackerNews a story about one designer, who made his/her CV as infographic artwork. This idea inspired me to create my own resume in a form of QlikView dashboard. So, I spent 3 hours to create QlikView application, that visualize my experience, education and some skills. As it appeared rather good to me, I decided to test the idea by publishing it on LinkedIn in QlikView group (link).
The response has surpassed my expectations. I've got a lot of very good and encouraging comments from people all around the globe. I've also got some mails from recruiters and potential employers, but this time without any result (not having US/EU work permit is a huge obstacle). Two weeks in a row LinkedIn had been indicating me as top influencing person in the group. The CV was also published as Creative CV #19 on www.globalrecruitingroundtable.com.

While it was almost a joke, I think the idea has some reason behind it. A random resume looks like typical poorly designed BI report -- too much text, small font, and sometimes, too much formatting. At the same time, every recruiter needs only some key facts and figures from a resume in order to make a decision -- whether to continue with it or throw it into waste basket. Clean and accurate visualization of these key facts and figures could make their life easier.

(click to enlarge)

February 17, 2011

A few slightly pessimistic comments on Gartner's BI Platforms report 2010

Gartner's annual BI reports always create a lot of buzz, when issued. The most discussed (and perhaps, the most contradictory) part of these reports surely is the "quadrants" with vendors positioned inside.

Ironically, these quadrants definitely are not the most interesting and valuable part of the report. Really, I don't know what valuable knowledge can be obtained from them. The latest-greatest BI? Come on. In BI rollouts every organization has it's own set of goals, criteria, preferences and restrictions that need proper analysis and specific solution. How would you apply that quadrants there? In my opinion, the quadrants are just too abstract to be practically useful.

Gartner fulfills tremendous task of identifying market trends, surveying numerous users and analyzing complex product portfolios, and they do it very professionally. I find the sections Market Overview and Vendors Strength and Cautions very compelling and worth in-depth reading. Can't say I agree with everything there, but no doubt, it's a good job done by smart people.

A few comments on Gartner's Market Overview this year:
  • "Data discovery platform momentum accentuates the need for a portfolio approach". Last year Gartner mentioned that idea of "BI-standardization" has actually failed -- more and more customers intentionally use 2-3 BI-platforms to cover needs of business users better. This year the trend continues. As of today, there is no single BI-platform that can effectively cover needs of a large enterprise, solely.
  • "Data discovery" is a new hot thing this year -- did you notice this? Do you remember the hype about "pervasive BI" a few years ago? Did it make any revolution? Well, no. Will data discovery do it? Who knows. However, it's obvious that BI industry still can't overcome it's biggest problem -- low adoption rates among non-technical users. What concerns me is that this year there is absolutely nobody in the Visionary quadrant -- it means that there will be no any significant innovations in the next 3-5 years. No really fresh ideas on BI market. Well OK, QlikView has shown very interesting and innovative approach to BI. But their recent new versions (10, 9, etc.) look more like boring updates rather than new breakthrough.
  • "Shift from measurement to analysis, forecasting and optimization". This is what BI is for -- to drive businesses in the right direction. However, I don't believe in hype about predictive analytics. This is just another wrong attempt to overcome the above-mentioned biggest problem of BI. Building predictive models is not a simple task and it requires from user to have skills in statistics at least to understand what's under the hood. Tools will not magically compensate lack of these skills. No magic.
  • "Mobile BI" -- another one hot thing in BI. Will BI revolution come from this side? Who knows. What's good about it -- is that this trend is inline with global, fundamental shift in lifestyle which is happening now. What's not good about it is that common hype about tablets (iPads, etc.) inflates hype about mobile BI. Hype about tablets will fade as more and more people make distinction between toys and tools. But today this hype prevents us from clear understanding the value mobile BI can give. Really, is it a toy or a tool?
  • Unfortunately, the second current fundamental shift in lifestyle -- socialization -- didn't get proper reflection in BI industry. And this is disappointing, because this is clearly the way to better productivity caused by better knowledge exchange. And knowledge is what business intelligence used to be proud of. I believe, social BI, if designed properly, can have much higher chances to bring valuable innovation to what is called "management information systems". Who's going to do this?
Links:
[1] - Gartner's Magic Quadrant for Business Intelligence Platforms (27-Jan-2011)
[2] - Paul Graham's essay about tablets

February 16, 2011

Where to get free analytical database management system

It's not a secret that majority of popular relational DBMSes (e.g. Oracle, MS SQL Server or MySQL) were originally designed for transactional processing. Besides other features, they employ row-based data storage and SMP-architecture which is good for OLTP systems, but in case of analytical applications (like enterprise data warehouses) that require heavy scans and data aggregation, this is far not the best case. Usually, analytical workloads are handled much better by purpose-built analytical DBMSes. If you want to try such DBMS without hassle of dealing with sales people there're at least 4 options to try an analytical database yourself and for free:

Greenplum Community Edition
http://community.greenplum.com/

This is analytical relational DBMS designed for heavy workloads on terabytes of data. Its core is derived from PosgreSQL, which assumes that data is still stored in rows, however Greenplum has MPP architecture.
The free community edition is limited to 2 CPU sockets or 8 virtual cores.

Infobright Community Edition

http://www.infobright.org/

Infobright is truly columnar DBMS, however not so popular as Sybase IQ or Vertica. The free Community Edition is simplified version of commercial Enterprise edition and is open-sourced. It lacks some features that can be important for large-scale deployment: DML, Temporary Tables, Parallel Query Execution and some other. Nevertheless, even this limited edition can show 5x-10x performance improvement on small datasets (say, up to 0.5TB).

LucidDB
http://www.luciddb.org/

LucidDB is also columnar DBMS which is open-source from origin. I've not heard about any large-scale deployments of LucidDB, however some benchmarks show impressive improvement over MySQL. So, if you use the latter for analytical workloads give a closer look at LucidDB.

MonetDB
http://monetdb.cwi.nl/

Another one open-source columnar DBMS which was designed in the Netherlands. Features include enhanced support for XML and multimedia objects, and support for modern CPU architecture.

Intro to this blog

In 2010 I launched russian-language blog BI Review (http://www.bi-review.ru) which is dedicated to Business Intelligence and Data Visualization, and where I used to share my thoughts about BI tools and market trends with russian-language audience. The blog has got some attention from readers and number of visits crossed the 1000 visits/month mark in a few month. I also noticed that some english-language readers of the blog tried to read it using Google translate services but I'm not sure whether this helped them a lot.

Therefore, to make things a bit simpler for them, I decided to start my english-language blog (the one you're reading now) dedicated to the same theme -- Data Visualization and Business Intelligence tools, Data Warehouses and Corporate Performance Management systems. This blog is not exact copy of the russian-language blog. First, it would be very time consuming to translate every article, and second, there are some thoughts I would like to share with the english-language audience solely.

Gradually, this blog became my main blog related to my professional activity. It is visited by more 3000 readers every month. Since I moved to Canada in 2012 I keep blogging only in English.

About me: I'm BI enthusiast since 2004. My professional bio is available at LinkedIn at www.linkedin.com/in/dgudkov. Feel free to connect with me there or follow me (@dgudkov) on Twitter.

Enjoy and have fun!

Dmitry Gudkov