Business Intelligence—It’s Time to Pass the Baton

Information management is a means; it is not an end. If the information is well managed but does not have an impact on performance accomplishment, then the technology is without value-it’s a toy, not a tool. We have to keep our perspective on the uses of the information, not the information itself. We have to understand the cognitive landscape that permits decision makers to effectively use IT.

Working Minds, Beth Crandall, Gary Klein, and Robert R. Hoffman, The MIT Press, Cambridge, MA, 2006, p, 168.

The business intelligence (BI) industry has done a great deal for information management, but it hasn’t fulfilled the essence of its promise. It has provided a powerful technical infrastructure for collecting, integrating, improving, storing, and accessing large amounts of information, but few tools that directly help people understand and make good use of that information. For years I’ve been hoping that, with the right encouragement, the BI community would figure this out and begin the work that remains. This requires a paradigm shift that I had to make years ago in my own work, so I believed the industry could do the same. BI must shift from a focus on engineering and technology to a focus on the human beings whose work the technology was created to support. After years of constant effort and almost relentless frustration, I now believe that the industry at large will not and perhaps cannot make this shift. It is too entrenched in a techno-centric paradigm. The skills that have enabled BI to build a solid technical infrastructure, which has made so much information available, are not the skills that are needed to build tools for information sense-making and presentation, the activities that most directly support decision making. While most traditional BI vendors have made failed attempts to provide what they don’t understand-data analysis and communication-others have arisen to do the job.

The vendors that are effectively doing what BI has promised fall into two camps: (1) relatively recent start-ups, which are spin-offs of non-BI efforts-mostly academic research, and (2) statistical analysis vendors that have been around for many years but have only recently made efforts to befriend businesspeople. Both of these camps have been working to infiltrate the BI industry with their solutions, but the industry resists them. Not only do they meet resistance, they also find that their association with the BI industry often prevents the very people who need the solutions that they offer from recognizing how these tools are different and better than pseudo-solutions of BI.

I believe it’s now time for the vendors with real decision support solutions to thank the BI industry for the technical infrastructure that it’s provided, but then set themselves apart as a new industry, different from but complementary to BI. Much as groups of people throughout history have arisen and set themselves apart to fix what cannot be fixed within the reigning power structure, the decision-support solutions that people need will only make their mark on the world by leaving the calcified fortress of BI.

Take care,

17 Comments on “Business Intelligence—It’s Time to Pass the Baton”


By Rob Meredith. February 12th, 2009 at 3:58 pm

Spot on Steve. How do you think these proper decision-support organisations should brand themselves? Perhaps a return to the DSS moniker?

By Stephen Few. February 12th, 2009 at 4:45 pm

Rob,

I intentionally refrained from proposing a new moniker, because I wanted to leave the path wide open. Feel free to propose a worthy name for this bold, overdue venture.

By Jim Novo. February 12th, 2009 at 5:00 pm

Calcified fortress, indeed. Care to share who you think are “vendors that are effectively doing what BI has promised”?

By Jonathan Fritz. February 12th, 2009 at 8:13 pm

“I believe it’s now time for the vendors with real decision support solutions to thank the BI industry for the technical infrastructure that it’s provided”

Indeed. While it’s not user-facing, this technology is an important enabler, and should not be taken for granted.

By Jacques Warren. February 13th, 2009 at 7:29 am

Very interesting. I can tell you’ve been brewing this one for a long time.

What type of people (i.e. skills/competence) are going to be better positioned to accomplish that?

By Stephen Few. February 13th, 2009 at 2:24 pm

Jim,

I’ll do my best to answer your question, at least in part. I assume that you’re asking primarily about the software vendors that are in the best position to lead us successfully into the realm of analytics that BI has promised but largely failed to deliver. Given this assumption, I’ll talk on about software vendors and for now ignore consulting organizations, educational organizations, content providers, and thought leaders, who will also play a role in the success of analytics.

I mentioned in my blog post that the only vendors that are currently beginning to support data sense-making (i.e., analysis) in an effective way fall into two camps: (1) relatively recent start-ups, which are spin-offs of non-BI efforts—mostly academic research, and (2) statistical analysis vendors that have been around for many years but have only recently made efforts to befriend businesspeople.

To my knowledge, none of the start-ups of the last few years that have been founded by people with a background predominantly in BI have addressed analytics effectively. There might be exceptions that I haven’t seen, but those that I have seen exhibit the same mistakes that the big BI companies that also exhibit: a techno-centric, engineering orientation that keeps them from understanding the real needs, abilities, and limitations of human beings. As a result, they build solutions that only an engineer could love, which are almost entirely unusable for data sense-making.

A few companies that have emerged in recent years with backgrounds that are not predominantly grounded in BI, especially academics working in the field of information visualization, are succeeding where the BI guys are failing, and are doing so because of their human-centric, design orientation and deep grounding in the sciences (human factors, cognitive psychology, statistics, etc.). I believe that Tableau Software (a spin-off of work that was originally done at Stanford) and Spotfire (a spin-off of work that was originally done at the University of Maryland) exemplify this category of vendor. A few other vendors that are not necessarily direct spin-offs from universities, but share similar backgrounds with them, are also doing good work, such as Panopticon. A vendor that I know less about, which appears to have no direct ties to the information visualization research community, but seems to be doing good work, is Visual I|O. This company was founded by Angela Shen-Hsieh, whose background is predominantly in architecture. I believe that this company is succeeding because of its human-centric, design orientation. Shen-Hsieh’s training as an architect has given her design skills that she has been able to apply to data visualization in meaningful and effective ways.

The other category of vendors that are making inroads into data sense-making consists of statistical analysis companies. Obviously, vendors of this type understand data analysis, but they haven’t always known how to address the needs of a broad audience of business users. Their big challenge, I believe, is to deliver effective analytical tools that don’t require deep training in statistics and do exhibit a friendly, seamlessly interactive user-interface, similar to what Tableau has managed to produce. Companies like SAS Institute are doing good work and are in a position to extend their reach to a much broader audience if they can learn to better understand and communicate effectively with non-statisticians and improve their user interfaces. Of the vendors in this category, I’m mostly familiar with SAS, but it’s possible that others, such as SPSS, have similar opportunities.

I hope this gives you a better sense of the vendors that I believe are poised to take the baton and complete the race.

By Paul Turner. February 13th, 2009 at 4:18 pm

I completely agree – partly the reason I left the BI industry (after working in it for about 10 years) is because it simply came down a ‘who has the prettiest dashboard’ context. With Adobe Flash and Flex it just moved to the next level of prettiness.

I found the SAP Xcelsius ads the perfect example – lots of pretty graphics, but careful not to tell the buyer the full story of the data integration, the calculation – and the fact that it really adds no value over a spreadsheet.

In the world of SaaS BI, the new vendors need to ensure they don’t embrace this mentality – as if it doesn’t deliver value and true insight, then customers will simply switch it off.

By James Taylor. February 13th, 2009 at 6:42 pm

While I don’t disagree on the general thrust, I also think the BI industry is failing to supply the kind of analytic insight that is needed to develop smarter systems – those with analytics embedded so that the code (or better yet the rules) can use the analytics without human intervention at all. Decision Support may be poorly served by many BI tools but Decision Management or Decision Automation gets even less support from them.
JT

James Taylor
CEO, Decision Management Solutions

By Stephen Few. February 14th, 2009 at 12:31 pm

James,

I know that you and Neil Raden have given this a lot of thought, having written a book on the topic, and I have no doubt that decision automation has an important role to play and that the BI industry hasn’t done enough about it. I’m not sure that I would trust the BI industry, which understands so little about human decision making, to build automated decisions systems. Just as I believe that most BI companies lack the skill set that is required to support analytics tools that support humans, I believe the same is true for automated decision systems.

To approach the development of automated decision systems with the techno-centric, engineering mindset of most BI companies would produce a disaster. To build these systems, you must know where to draw the line between decisions that can be automated and those that should involve human interaction. You must also appreciate the fact that decision systems which replace human interaction must do so in a way that doesn’t take humans out of the loop entirely. In most cases, it is necessary to keep humans engaged in monitoring the process to the extent that they maintain situation awareness, even when they aren’t needed, because otherwise, when their intervention is needed, they’ll lack the awareness that is required to respond effectively.

In other words, building effective decision automation requires a human-centric, design-oriented, scientifically-grounded mindset to direct the work of engineers. The engineers should be the enablers, not the drivers. I suspect that you and I agree about this.

By Stephen Few. February 17th, 2009 at 3:03 am

Jacques,

Are you asking about the competencies that will be needed by the people who will produce the best analytical tools or those that will be needed by practitioners who use the tools?

By Hetty4. February 17th, 2009 at 12:52 pm

I think I am having a similar experience but from a completely different vantage point. I was hired by a small brand and advertising agency to start an analytics practice. The focus: provide intelligence to demonstrate success of branding and marketing efforts and to provide strategic direction for improvement. In other words: is it working or not, and if not, what will. So, I’m starting with the “here’s what I want to understand and the decisions that I want to make” and trying to find the tools to allow me to do this efficiently and communicate with a simple, easy-to-understand interface. This has recently put me on the fringe of business intelligence tools.

I have found that end-users of analytics and intelligence are also often more captivated with the tools than with the outcome. They talk as if analytics=dashboard — the shinier, the better. I think good analytics is being able to take information from different places, find the pertinent relationships, and make intelligence decisions that positively effect desired outcomes (seems like a duh, but I am often surprised). BI tools, I hope, can help me bring the information together. Finding pertinent relationships and making decisions requires humans (hopefully me!). BI tools can also help me create visual illustrations of relationships and intelligence so that it can be more easily understood by a wide range of data-friendliness. I do not see “dashboards” as the analytics piece, merely the UI. I am struggling to get my end-users to focus on what the intelligence will do and not how it will look on their screen.

Like I said, I am on the fringe of BI and am in learning mode. I come from 15 years of market research, 8 of those spent on global consumer trends and how business apply them. I’m not a statistician but have spent a lot of time interpreting, tweaking, and applying statistical models. I think you’re correct on a hybrid or symbiotic new industry emerge; can’t wait to see it evolve. Thanks for the info on the companies you see doing this already!

By Jacques Warren. February 17th, 2009 at 5:46 pm

Hi Stephen, I meant the latter.

By Niranjan Pedanekar. February 18th, 2009 at 9:40 am

Very enlightening.

I also have been thinking about (and working a bit on) a framework that tries to look at the decision process from the human viewpoint. It is slightly different than the vision mentioned by Stephen, but nevertheless.

The existing ‘intelligence’ flow seems to be:

Get data (Reports, dashboards, …)
-> Get insight (Humans scratching their head based on what they see)
-> Take action (Humans hope to get it right)

Could this be made even better to prove the real worth of decision support systems? Consider the following model.

Question (Humans pose a question, which is optional)
-> Get insight (Humans are suggested an insight, perhaps in plain English. This is either based on the question, or the ‘system’ itself suggests appropriate insights based on the data seen. Examples would be anomalies, trends, correlations, and a hundred other things that data can reveal. As opposed to the aforementioned model, insights are pushed to the human.)
-> Get evidence (Humans get evidence in an appropriate form such as visualisation, numbers, reports, …)
-> Take action (Humans take action, the ‘system’ may even suggest actions)

The latter route, IMO, may make more people more prone to reacting to insights (or calls to action) which are being thrown at them. And this need for action may compel them to look for better evidence to justify the action being taken. So, more effective visualisation may matter a lot in this case.

Of course, this model would have to deal with the tough problems of ‘what is interesting’, false positives, human feedback, and so on. But then, no one said there will be no hard work ;)

Do you think that this vision is complementary to the one that Stephen describes?

An alternative name to the vision could be a term borrowed from the military: ‘Actionable Intelligence’. Intelligence that makes you act.

By Michael Driscoll. February 19th, 2009 at 12:41 pm

Stephen – As someone new to the BI space, I couldn’t agree more with you. I actually believe that new disruptive technologies — such as Hadoop, R, and grid computing — threaten to make existing BI solutions expensive at best and obsolete at worst.

To make an analogy to the space that I come from, which is the life sciences: in the early days of computational biology, research revolved around techniques for managing and manipulating genomic data — work dominated by computer scientists. But now, biologists are the ones driving the field. I believe the same will occur in the business intelligence space: many of the technical challenges have been solved, the competitive advantage now lies on the business side.

By Matt Cooke. February 19th, 2009 at 1:50 pm

Stephen,

Very good summary of the current state. In fact, we’ve recently founded a company that includes dashboards as part of our repertoire. However, the only reason that dashboards are in our offering, is so we can present data that users can take action on.

Dashboards and measurement are just one piece of how you run a business. Just as important are:
* robust dialogue that uncovers the root-cause of issues found in a dashboard
* action planning that establishes ownership and accountability for improving the results found in a dashboard

Building an application that maps to the workflow of running a business is more exciting to us than building, as you put it:”a powerful technical infrastructure for collecting, integrating, improving, storing, and accessing large amounts of information”.

Side note: As we talk to customers and investors, we’ve learned to stay far away from the words “business intelligence”. They now carry a stigma that we want to avoid. My suggestion for the name of the new market is “Execution Management”. That is what it all boils down to.

Thanks for the insight.

Matt

By Neil Dulohery. February 19th, 2009 at 3:54 pm

Well said Stephen. Most products currently marketed as “business intelligence” offer little more than trivial capabilities for information retrieval and display.

What you are proposing, however, sounds a lot like something that has been around for a long time but perhaps not branded very well. When computers were introduced into business, analysts, statisticians, scientists, and engineers were often placed directly in control of these systems so that they could automate resource-intensive analytical tasks. Products like SAS rose to prominence because with relatively few instructions, one could invoke a vast amount of functionality that had been written by someone else to produce virtually any kind of analysis. You got the data from wherever you had to. The analysis could be completed and automated in days or weeks. Another important capability then and now was to essentially create small data warehouses on the fly by integrating among multiple input sources. You could then perform multi step analyses to get the desired results. What was a necessity then remains relevant today.

The “information management” discipline has done a good job of organizing a lot of information, especially transactions. One conceptual flaw common in information management is that data warehouses can somehow anticipate all or even most future analytical and reporting needs. That is too ambitious and unnecessary. Many analysts have become constrained, not enabled, by data warehouses. Data warehouses in combination with the current crop of “business intelligence” tools work well only for routine information retrieval and reporting. Analytics that respond quickly to management problems and produce graphical displays with high-information densities often require more horsepower in the background.

My requirements list for a suite of decision support tools includes the following: 1) fast access to multiple data sources, 2) ability to create integrated data structures on the fly, 3) multistep processing and modeling of the integrated data, 4) interactive graphical displays rendered exactly as designed, and 5) short development cycle. Several combinations of products satisfy these requirements.

By Stephen Few. February 25th, 2009 at 11:44 am

Jacques,

The perspectives and skills that are needed to develop effective tools for data analysis differ somewhat from those that are needed by the people who will successfully use those tools. Whereas the designers of these tools must be human-centered and design-oriented, with an understanding of human factors, cognitive psychology, visual perception, etc., data analysts who use the tools don’t need the same depth in these areas. I believe that the best data analysts tend to possess the following talents and attributes:

    Interest in the data
    Curiosity
    Self-actualization (that is, they don’t wait around for people to tell them what to do, but pursue what seems interesting and worthwhile without specific direction to do so)
    Imagination
    Open mindedness and flexibility
    Skepticism
    Honesty
    Sense of what’s worthwhile
    Analytical (ability to break things down into their constituent parts)
    Synthetical (ability to see how pieces can fit together to form something bigger and more complex)
    Eye for meaningful patterns
    Knowledge of the data
    Knowledge of effective data analysis practices

This is by no means an exhaustive list, but I believe that each of these contributes significantly to the analytical process.
Even though I compiled this list of talents and attributes to describe good practitioners of data analysis, I also believe that those who design tools for data analysis should also possess these talents and attributes to a fair degree, for without being good analysts themselves, their ability to design good analytical tools will be severely limited.