hertenberger.co.za

I’m writing this on the day it’s supposed to be posted. The question of posting on time is up in the air; a note to the aspiring traveller – should you find yourself in the area of the Vaal Dam, or the Emerald Hotel and Casino, to be precise, forget about trying to get online. 3G is non-existent, and the GPRS connection on both the Vodacom HSDPA card and my JAMin are less than reliable. Trying to coax anything more complex than the bare-necessities Google page to appear in the web browser is an exercise in frustration and futility. You’re probably reading this tomorrow. That’s ok. Don’t get too concerned about temporal displacement. Time.Machine.Does.Not.Exist.

Normally, I would be careful to place the words Intelligence and Microsoft in the same sentence. Seeing as this event was hosted by that company, I’ll dispense with that, notwithstanding the fact that more than one presenter, including a self-professed independent analyst, had less than complimentary things to say about SAP’s BI product. Justified, then.

Presentations are divided into a technical and a business track, with the business track including a number of customer case studies. The keynote, delivered by Nigel Pendse, provided some interesting insights gained from a number of studies and surveys. He indicates, quite rightly, that a business intelligence solution should not take longer than six months to complete. Ideally, three months is a good time frame to ensure user involvement and interest and alignment with the original objectives. Something that I’ve noticed a lot in the past year is the fact that data quality issues dog a number of BI implementations. According to results from surveys, SAP’s BI customers show up the highest incidence of data quality issues. Possibly due to the fact that SAP’s BI is seen as nothing more than a bolt-on to an existing SAP ERP system. Hence, data is often simply dumped into cubes instead of being properly cleansed and transformed. An interesting observation comes from the top three problems experienced by BI projects: slow query performance, poor data quality and company politics. Of these, slow query performance is the one that is considered the most serious.

And the reason? Google, who else? With users experiencing sub-second response times when querying billions of rows online and having those results displayed as expected, a two second response time is simply too long. Five minutes is a lifetime. And consider this: Google is free. You pay for your BI system and its infrastructure. Last point made in the keynote: according to surveys, both Oracle and SAP products deliver fewer business benefits and seem to cause the most problems in projects. Thankfully, Pendse refrained espousing the virtues of Microsoft too often! By the way, the survey results are not free.

For a BI conference, the technical track presentation on BI design and architecture was slightly disappointing. Covering the basic concepts of what a data warehouse and data mart is, is not exactly too useful to an audience already well-versed with business intelligence. The two approaches to implementing business intelligence are the introduction of multiple data marts or a single data warehouse. Whereas the cost for a data warehouse is initially quite high, the overall cost is lower and more benefits are derived over time. The introduction of multiple data marts is a cheaper solution, until too many data marts have to be maintained and their data aggregated.

On to the business track where two case studies were presented: the Shoprite and Multichoice BI implementations. Shoprite is expanding at a rate of almost two stores every week! They make use of a wide variety of customized dashboards to display key figures to a varied audience. Multichoice is under threat of losing its monopoly and needed a solution to determine customer distribution and marketing information. Both case studies where informative and provided a peek under the hood.

Last one for the day. The establishment of a Microsoft-led BI user group. A website has already been created and content is being sought. If you have an interest, check it out.

A gift of Freakonomics before bedtime.

Thanks, Microsoft!