Repeating History

I've been tidying up some of my things in my office and came across an old paper I used as a reference when I was writing my honours thesis back in 1996. The paper[1], written by Elizabeth O'Sullivan back in 1985, was talking about decision support systems (DSS) and contrasted them with the management information systems (MIS) of the 1960s and 70s.

A bit of history first.

MIS, in the 1960s were a disastrous failure. They were developed with the mindset that a techno-centric approach to management was the optimal means of making decisions. They collated data from transaction processing systems, and churned out reports on a regular basis (paper-based, since managers didn't have terminals on their desktops). The systems were designed to support the decision-making needs of all managers in an organisation. Developers at the time argued things like:

With so useful information captured ... we have attempted to make this as freely available as possible to sections of management that can use the information effectively. Past approaches, where special one time programs written by the Systems Department programmers (if they could get time) ...

This appeared in a paper describing the Ford Motor Co.'s MIS approach in 1967. Sound familiar? It should. All of the things people were trying to achieve with MIS are the things they're trying to achieve now with BI and Data Warehousing. And it didn't work. Millions of dollars were spent and lost, and IT departments garnered a reputation that made business people question their hygiene, parentage and mental capacity.

You might be thinking that these systems failed because they were working with machines that boasted a whopping 32kb of memory, and 8mb of disk storage, and reported using paper-based printouts. In fact, the reasons they failed had nothing to do with the technology or its limitations. This is from a paper by Russell Ackoff, again, writing in 1967:

MIS are based on the following false assumptions:

• More information is better
• Managers don’t have the information they need
• Managers need the information they want
• Managers don’t have to understand a system to use it

None of these have anything directly to do with the technology used. In other words, MIS failed because the developers didn't understand managers, the way they work, and the way they make decisions. The same mindset is still surprisingly common with BI.

Back to the paper I found. Here's an extract from the opening paragraphs:

[DSS] represent an attitude that management information systems (MIS) can and should do more than serve routine organizational operations... To support decision making effectively, a DSS must be flexible, user friendly, and interactive. It should go beyond traditional MIS, ... [it] must recognize and accomodate individual problem-solving styles...

[Emphasis is mine]

DSS were phenomenally successful at the time in comparison to MIS. The reason for this is the attention paid to individual decision-support. The same holds true for BI today. There is no way BI will work unless developers understand decision-making as a human process, and pay attention to helping individual decision-makers with tailored support. Unfortunately, this attitude, prevalent in the 1980s DSS movement and key to its success, appears to be missing from today's vendors, and the majority of developers.

[1] O'SULLIVAN, E. (1985) Decision Support Systems: An Introduction for Program Evaluators. Evaluation Review, 9(1), 84-92. (unfortunately, not available online).

SQL Server to run 270TB Multi-node Data Warehouse

Computerworld are reporting that Microsoft are working on a massive data warehousing project for an external client, in an obvious aim to dispel the idea that SQL Server is a lightweight platform. From the article:

At the annual conference of the Professional Association for SQL Server (PASS) user group, Microsoft said it is designing a 270TB multinode data warehouse for a foreign (ie. non-US) government that it declined to identify. The software vendor is also working on a 162TB single-node installation for its own marketing department.

Of course, this doesn't explain how Microsoft are doing this (any tweaks under the hood?), or whether the warehouse itself will be of any use (no metrics on reporting, data mining, etc. throughput - and we all know how Microsoft are on benchmarking *wink*). Still, I quite like SQL Server (and the MS BI tool suite) as a product, and it has singlehandedly been responsible for the biggest shakeup in the BI industry in the past several decades (a much needed one at that). If this kind of data volume can be handled well by SQL Server 2005, as configured by your average corporate DBA, then Oracle and IBM are really going to be looking over their shoulders.

Criticising Edward Tufte

I'm a huge fan of Edward Tufte, and so are many people who are interested in the graphical presentation of information. He's got a lot of good things to say that BI people should listen to, and he says it in an engaging way. Pictured, right, is Minard's diagram of Napoleon's march on Moscow, used by Tufte as an exemplar of good graphical information presentation.

As with any guru, though, his adherents sometimes get bogged down with the minutae of what he says, rather than the underlying principles he tries to communicate. There's an interesting couple of posts over at Emergent Chaos in response to a criticism of the Minard diagram as a communication tool, and Juice Analytic's follow-up. The original criticism was made by Seth Godin in the following video:

Uncertain about uncertainty

One of the key theories that informs our teaching and research into BI at Monash comes from cognitive science. It states that, for a whole variety of reasons, people don't make rational decisions. Instead, we're subject to so-called cognitive biases. One category of bias has to do with how we think about statistics - as a rule, people are terrible at thinking in ways that are consistent with the laws of Bayesian probability. This is a real problem for BI developers, since most BI tools deal with quantitative values, and hence, statistics and probability. Even worse, just about everyone is subject to these biases, even trained mathematicians, and presumably, BI professionals.

I've just finished watching an excellent talk by statistician Peter Donnelly that covers this issue brilliantly. It's one of the recent TED podcast episodes, that I've referred to before, and runs for about 20 minutes.



If you want to find out more about some of the other (really freaky) cognitive biases that exist, check out our bias page at http://olap.sims.monash.edu.au/research/decision.nsf.

Father of BI? Is he having a laugh?!

Computer World are running a story on Howard Dresner, the former Gartner analyst who lays claim to having invented the phrase 'business intelligence'. From the article:

Howard Dresner coined the term "business intelligence" in 1989 while an analyst at research firm Gartner Inc. ... Dresner was seeking a term that would elevate the debate and better define the analysis of quantitative information by a wide variety of users.

I'm sorry, but at best, Dresner re-appropriated the term to rebadge what was then called DSS. H.P. Luhn actually invented the term, not in 1989, but in 1958 in an IBM Journal article called A Business Intelligence System, (vol. 2, no. 4, p.314) that pretty accurately predicts BI systems today. Here's the original definition of Business Intelligence from 1958 (Luhn, 1958, p. 314):

In this paper, business is a collection of activities carried on for whatever purpose, be it science, technology, commerce, industry, law, government, defense, et cetera. The communication facility serving the conduct of a business (in the broad sense) may be referred to as an intelligence system. The notion of intelligence is also defined here, in a more general sense, as "the ability to apprehend the interrelationships of presented facts in such a way as to guide action towards a desired goal."

Sods to Howard Dresner and Gartner for claiming to have invented something they didn't, and to the Computer World people for not doing their research. It's interesting to see that Dresner is now redefining BI as "business process management" (from the CW article), something that's a bit removed from the aims of DSS and EIS, and from the sounds of it, Luhn's original definition too.

IBM and Business Objects form Strategic Global Alliance

Good news for those companies with both platforms - two of the largest software vendors, Business Objects (BO) and IBM have just announced a global strategic alliance. So what does this mean? The word is that the new agreement will provide enhanced support for those companies with both bits of software. What the agreement really means is that it will put BO and IBM in the position to capture even greater market share.
Read more...

Defining Business Intelligence

As academics, we are often working with concepts and ideas, and when we are researching these concepts, we need to have explicit definitions that specify what is, and what isn't, of interest to the study. Sometimes, as with the case of Business Intelligence, there is a fairly strong, but implicit, definition used in industry, which is fine, since an explicit definition beyond a specific project doesn't really matter too much.
However, for academics, this can create difficulties. Without an accepted explicit definition, particularly for a hot-topic like BI, academics jump on the bandwagon claiming the area as part of their expertise without justification. Recently we've been working with some academics from the harder end of the IT spectrum, and we've struggled to make the point that the 'intelligence' bit of BI is not the same as the 'intelligence' bit of AI, or even human intelligence, but rather in the sense that Herbert Simon used it: ie. monitoring and gathering information to support decision-making.

But even that doesn't really solve much. Some people view BI as just the hardware and software tools that support that process. Others see the process itself as BI: Ralph Kimball, for example, famously uses a publishing metaphor in relation to data warehousing. Some see BI as the delivery layer or front-end, and treat data warehouses as something quite separate. Others would include the whole information supply-chain as BI.

So, what is BI to you? Is it any technology that supports organisational decision making? Is it specifically reporting and analytics? Does this include enterprise reporting, or does BI go beyond a kind of passive, push technology? Does BI extend beyond supporting decision-making to the actions that result from a decision? What about operational versus strategic decisions?

Leave a comment and let us know.