Intelligence will out

Intelligence will out

Panel Participants

Dean Blomson

Head of Strategy Advisory Services

Ernst & Young Consulting

Frank Gillett

Senior Analyst

Software Development Research

Forrester Research

Iain Anderson

Program Director Application Delivery StrategiesMeta Group Asia PacificRolf JesterPrincipal AnalystEnterprise Software Markets, Asia/PacificGartnerGroup's DataquestCIO: What business factors are driving the take-up of BI solutions?

Anderson: The biggest factors are things like customer relationship management CRM, the Y2K hiatus, a fickle consumer population, global competition and extended reach.

Jester: Business managers sense the urgency of staying on top of a rapidly changing environment. Structured information (as in BI) is only one piece of what is needed, but it is needed.

The truth is out there. Enterprises have the data -- often more data, from more sources, than they can reasonably handle. The widespread adoption of ERP and other enterprise applications has accelerated that.

The tools exist. Much more far-reaching, usable and useful than in the past. Beginning to look like solutions rather than raw technologies.

Gillett: Increasing availability of data, and the increasing speed of business change.

Blomson: Four main factors are driving the take-up. Firstly, limited time -- the reality is that knowledge and intellectual capital are assets that, in a fast-moving virtual world, could make the difference between survival and failure. Management needs "the facts" available in real timeSecondly, limited resources -- strategic options are (in theory) limitless; resources are always constrained. The latitude for error and the need to be as precise as possible in one's selection of alternatives is ever increasing.

Thirdly, fast-emerging variables -- largely the recognition that organisations need more peripheral vision to track evolving trends, emerging competitors, technology and so on.

Finally, information overload (oversupply of information/undersupply of attention) -- the more information available, the more critical it is to apply intelligence to its use; therefore ensuring the most relevant information receives attention.

CIO: Currently we're seeing many "technologies" lumped under the business intelligence (BI) umbrella. What do you think does and does not belong: Executive information systems (EISs), decision support systems (DSSs), knowledge management (KM), competitive intelligence (CI)?

Gillett: I think they all belong. Some terms, like EIS, are old, and others, like KM, just haven't got respect. The terms and technologies matter less than the overall vision of bringing together information that helps people make better decisions.

Jester: I think EIS is an outdated term. I once heard a vendor say it should be "everybody's" information systems; that's closer to what it should be. As a term, DSS has value in that it sets BI tools apart from transactional systems. But KM, that's not the same thing as BI at all. It's a valuable term for a new and useful type of application but unfortunately, it's on the rising curve of the hype cycle and thus widely abused. BI creates one type of knowledge that should form part of what KM manages.

Again, CI is one of the important forms of knowledge that can be embraced by KM. It is of growing importance in an economy that is global and rapidly moving. The dangerous competitors are those coming from other industries, or from nowhere, that you aren't watching -- companies that add new kinds of value to your customers. BI won't help much with this because the data won't be in the sets that you have collected.

Anderson: CI, DSS and EIS are inherently part of KM. In marketing terms, however, I think DSS evolved into, or was renamed, as BI -- of which EIS is a sub component given the natural hierarchy of information consolidations. CI is typically viewed as a subset of KM.

My concern is that KM and BI are often separate mindsets although both philosophies have the end goal of increasing organisational value through the effective use of information -- structured or otherwise.

Blomson: EISs are, technically speaking, more a way of tracking performance or progress against organisational targets and strategic objectives. In theory, EISs should act as decision catalysts, by indicating what is and isn't performing. EISs and DSSs are the core components of what we call "strategic decisioning". DSSs are a key mechanism for using BI "outputs" as inputs for "resource deployment" decisions. This allocation of resources could be for operational or capital purposes: whether in support of customer management; manufacturing or procurement decisions; or IT system enhancements. These decisions could be achieved via holistic or "point" DSS solutions, and may be OLAP supported. If we define BI to be anything that supports decision making within a business, the decision need not be at a strategic level for it to have a significant proportion of BI within it.

KM does not automatically belong under the BI umbrella. If KM is implemented without a clear vision for supporting business decision-making then the knowledge that is actually managed may be of little value from a BI viewpoint. The best KM process may not necessarily deliver the best knowledge content.

I think CI, however, does belong.

CIO: Downsizing and rightsizing have seen flattened corporate hierarchies push decision-making capabilities "down and out" to division and department managers. Will we see a growing audience of users starting to employ BI tools and applications?

Blomson: Yes. Decisions need to be made as close as possible to the point of client contact. The focus needs to be on developing parameters within which front-line service staff ought to operate, and [then] ensure full latitude and accountability for managing against the measures/variables that they "own".

Good examples are room-service staff working in Ritz Carlton Hotels, and sales personnel with Nordstrom [the US department store]. The key being matching of "knowledge delivered" to measures/variables that can be controlled by the job incumbent.

Gillett: Some level of interactive analysis will be available to most employees in a company, each having appropriate levels of access for their job context. The real issue is the corporate culture: the level of delegation to individual employees will dictate the level of information, not the technology. The growth of analysis users from a few corporate employees to all staff is a major change that is under way right now.

Anderson: [Yes] as a goal, but they're rarely implemented as such. One of the key failure points in business performance management (BPM) implementations is the lack of actionable information, that is, not only is the data usable but the employee it is delivered to has the authority to independently action against it. We certainly see data being pushed out throughout organisations, but then it always was. An effective SLA for BPM implementations should also be to minimise the use of information systems as the information gathered is instantiated into business applications and processes -- the closed loop goal.

CIO: What are the biggest corporate barriers -- be they technological, cultural or other -- to a successful BI implementation?

Jester: There are several challenges. The first is it's such a big task that it may not yield results early enough if you try to do a full enterprise-wide solution. The business is sure to change in unpredictable ways before implementation is complete. Piecemeal approaches also have problems. You end up with islands of sub-optimal intelligence.

Scalability is a problem. When BI catches-on as a result of a successful pilot, demand for capacity (system, network, storage) grows beyond expectation.

Data quality is a challenge. Corporate data is often riddled with holes, inconsistencies and errors. That's not new; it has been thus forever, but BI makes the problems plain. Also vendors who have only a hammer and to whom therefore everything looks like a nail.

Another challenge is multiple incompatible data sources. The problem is not technical incompatibilities (different systems, data formats, databases and so on), but incompatible business data models (that is ways of thinking about the business's information), different semantics. At the bottom, this is often about narrow functional management thinking, power politics and stubbornness rather than practical information issues. But data modelling and metadata management are among the keys to success or failure.

Gillett: The biggest challenge is getting organisational agreement on data definitions and data consolidation. Two divisions must agree on the definition of data elements like "sales" and "profit" or they won't be able to integrate that data. Without executive direction, many companies stall on this problem and fail to integrate capabilities across the company.

Anderson: I think the prime factor for failure is still organisational as the technology and methodology underpins are probably mature enough to deliver demonstrable value to most organisations (appreciating that these best practice may be overlooked). The organisational issues would include:

Broad sponsorship -- obtaining cross-functional buy-in so as to create enterprise-class solutions can create conflicts with budgets and prioritiesBusiness performance management skills -- determining and instantiating business metrics IT methodologies -- reuse of existing waterfall lifecycle models or traditional IS planning to delivery urgent business value.

IT toolsets -- going through the buy versus build (particularly with ETL selection); understanding of the BPM market space; linkage to existing IT apps.

Skill transfer from consultants to internal staff, including finding suitable staffSLAs for managing the outcomes/ deliverable during both the development process and the ensuing operational activity.

Blomson: I'd rather consider the critical success factors for achieving a BI-enabled organisation. Firstly, build a culture of sharing, accessing, enriching and using knowledge. That is, move from synthesis and analysis of data, to enrichment and dissemination of knowledge, making sure that you are properly measuring and rewarding those behaviours.

Secondly, select technology that is a cost-effective enabler and will grow with your needs. Many organisations learn that having the best technology will not achieve the necessary mindset. Finally, recognise that the value does not lie in the data, or even the knowledge; rather it is in achieving valuable insights that lead to action.

CIO: So what's the process? How does an organisation take its first step?

Gillett: The ideal is to create an integrated corporate data model that can drive unification of data across the company, but no company can live up to that ideal. Rather, the reality is that companies should focus on solving immediate business problems, unifying data definitions one business problem at a time.

Jester: Organisations need to tackle both the big picture -- such as architecture and enterprise data models -- and a short-term quick success project from the outset. You need both. They also have to ensure that chosen tools will meet all needs, especially once user demand grows exponentially. Consider the different degrees of sophistication and power needed by different user populations.

Anticipate unpredictable change and build the architecture to welcome it. Ensure that new technologies can plug in. Expect to deal with multiple incompatible systems acquired in mergers. More importantly, plan for radical changes to the business that you are in: markets, customers, products and services, competitors, regulation, partners, distribution, and mergers.

Also plan for external data -- purchased or free -- as well as internal [data] and define their users widely -- not just employees, but suppliers, customers, shareholders, resellers, regulators and the tax department.

Paying attention to data quality is also important. It requires analysis and good processes, not just a one-time effort. There are excellent tools and services available. Otherwise your users will lose confidence.

Where possible, buy rather than build. For example, don't waste programmers' time writing extract routines for which good packages exist. Plan for data warehouse administration tools and labour costs. Just like any other IT resource, the data warehouses and data marts need to be managed.

Blomson: BI is a lot about core competence in an organisation so there is no one "ideal" process. It is not something you can determine without first-hand knowledge of an organisation. The challenge that we see for most organisations is not how to start, but how to pull disparate systems into a coherent structure. My advice for new starters is:

Think big, start small.

Identify a customer, employee, supplier that is of high value and analyse the decision process.

Review available knowledge and look to provide appropriate packets of knowledge to enrich the decision-making process.

Establish measures to review the impact of increased BI in decision-making.

CIO: What about these current BI trends? Enterprise information portals ? . . .

Gillett: A [BI-] related, but separate phenomenon that is really about companies unifying their systems environments in portals.

Anderson: I think it is more about enterprise portals in general, access to applications and information through a common portal infrastructure. People need the right amount of information, customised for their role rather than open-slather access to everything.

Jester: A current buzzword, but useful. A way of tying together the information tools and resources that people need, but beware -- needs change quickly.

CIO: Executive dashboards or cockpits? . . .

Jester: These are also useful if done in an environment where good management practices prevail, and where the users are able to take advantage of the tools. Watch cost-effectiveness.

Gillett: Yes, these are useful, but not sufficient for all interactive analysis.

Anderson: Sure, if there was an agreed set of business metrics. This is a trend across BI tools in general, an extension of the EIS implementations of old.

Blomson: Only a few years ago, balanced scorecard was a hot concept that many companies wanted to implement quickly. Most major organisations have developed some form of measurement system, but still lack a true performance management environment. The recent entry of ERP and data warehouse providers into this field should assist organisations to move away from their "paper based" methods of performance tracking to more automated, and thus "real-time" methods. Nevertheless, many ERP modules have a way to go before they can truly offer an end-to-end solution that enables rapid strategic decision-making on critical performance measures. They are moving down the right track, but it will take time before they catch up with the consultancies that have created and communicated the linkages among processes such as strategic management, performance management, value based management and capital commitments.

CIO: The ongoing move towards the Web? . . .

Anderson: It's a mandatory deployment mechanism.

Jester: This should be taken for granted. One of your BI user interfaces must be a browser -- for authorised internal and external users.

Gillett: Web browsers will be the interface for most users. Only intensive, sophisticated users will have Windows-native applications.

Blomson: As a search engine, the Web is quite powerful. The danger lies in the potential for the organisation to confuse intelligence with being on the Web -- the same argument as e-commerce versus just being on the Web.

CIO: Extending analytics to transactional systems? . . .

Blomson: The technology is now available to analyse customer data to identify and customise value propositions to individuals and then deliver the offer through transaction processing. Wouldn't it be good if, when I put my card in an ATM, instead of receiving a generic home loans rate advert, I could receive a personalised message based on my transaction history.

Jester: BI should be dealing with information that is as current as possible. That means links to transactional systems. But mostly this has to be via extract/ load/transform (ETL) tools as otherwise the BI workload may cause performance problems in the transactional systems and lead to poor service to actual customers.

Gillett: ERP vendors won't be good at adding analysis capabilities. They'll have to buy it or partner for it.

Anderson: Real-time and tactical DSS via ODS-type architectures are used today -- for such activities as fraud detection. They're still not intrinsically part of the transactional environment and will be difficult in the short term given the usage requirements of the data itself -- keyed, fast transactions against a twinkling database (OLTP) versus table scans against a static database (typical data warehouse) CIO: Packaged analytical applications versus in-house development?

Gillett: Most companies want to buy applications and solutions, not tools.

Jester: Packaged analytical apps have a useful role. The BI industry is, to some extent, going through what the business apps industry went through -- from custom systems to packages to flexible packages. The same considerations apply: affordability versus meeting requirements; flexibility.

Blomson: In-house development is needed in some situations to obtain adequate targeting. Off-the-shelf packages may be too generic which always poses the question: "Do you have others?".

Anderson: A growth area, again limited by the ability to "package" analysis requirements. A lot of verticalisation is happening here. Also there are other business initiatives such as CRM that are accelerating vendors' efforts to come up with packaged analytical apps for "standard" customer behaviour analysis such as profitability, churn and propensity.

CIO: What are some of the more common "land mines" in packaged analytical applications?

Anderson: The overall fit and understanding of business problem; the selection of applications; the fit to existing data warehouse infrastructure; embedding the application into business process rather than simple uni-directional data analysis; and management of multiple packages as breadth of analytical requirement broaden.

Gillett: They aren't equipped to combine data for a broader set of problems or they create stovepipes of data that are isolated from other data sets.

CIO: When BI goes wrong, what are usually the underlying factors?

Gillett: Usually, it comes down to trying to do too much in one project, or not focusing on a specific business problem.

Blomson: Two widely divergent problems are either over-reliance on the system, such as expecting that it can replace management's intuition and experience; or conversely, not trusting the data. Lack of appropriate information or oversupply of "irrelevant" information means the system does not provide enough information or provides a wealth of information not relevant to the decision.

Anderson: Many existing BI-focused implementations fail to deliver the benefits they were intended for because the information supply chain (ISC) is installed as uni-directional where as BI implies a need to influence business activities and hence requires closed loop processing.

Too many BI implementation are IT-lead initiatives. Organisations are still getting lost in the technology, schema design and architecture and not delivering against key business SLAs.

Business Intelligence: the Definitive AnswerBy 2000, information democracy will emerge in forward-thinking enterprises, with business intelligence information and applications available broadly to employees, consultants, customers, suppliers, and the public. [. . .] The key to thriving in a competitive marketplace is staying ahead of the competition. Making sound business decisions based on accurate and current information takes more than intuition. Data analysis, reporting, and query tools can help business users wade through a sea of data to synthesise valuable information from it -- today these tools collectively fall into a category called "business intelligence".

GartnerGroup report (September, 1996)

CIO: Do you agree with this definition of business intelligence (BI)?

Jester: I agree. Smart companies, those that will survive and succeed, will be those that are agile enough to ensure that all stakeholders can get at the information they need to add value right now, but other organisations are in danger of becoming irrelevant.

Gillett: Forrester calls this interactive analysis, which we believe has three components: 1) a full range of intuitive, integrated analysis, including data mining; 2) widespread access and distribution, and 3) personalised presentation. This capability is not yet fully available from vendors, who are still working to make Web-based access fully functional, and have yet to address the integration of data mining. It won't really be available until 2002, when data mining is reinvented and incorporated into database engines.

Finally, Forrester believes that interactive analysis will cease to be a separate activity and technology, because it must be built into the decision-making systems, rather than a separate application the user must reference while making operational decisions. This will be facilitated by the creation of intranet portals that integrate corporate resources and applications into a common entry point and interface.

Anderson: Information, yes -- by extension of the data warehousing projects that most forward-thinking organisations started some years ago. Democracy, a qualified no -- depending on the infrastructure established by these forward-thinking organisations as to whether they took an enterprise view of information dissemination. Meta Group sees that BI has become a catch-all term that fails to implicate the required information goals and hence fails in its implementation. We tend to use the term business performance management (BPM) as being the evolution of BI.

We believe that moving beyond one-way data presentation/beautification/aggregation to a solution that integrates front-to-back intelligence/action/learning is the next decade's real winner.

While BI attempted to move away from independent data mart implementation towards hub and spoke, there has still been limited focus on the underlying infrastructure and we would expect that forward-thinking enterprises would focus on metadata-driven deployments for leverage and reuse.

Blomson: This statement can be extended by two dimension. Businesses need to have capabilities in accessing or gathering data, both internal and external (which includes environmental scanning), as a key input; and knowledge management (KM) as the enabling process that runs through this.

Another key new driver of BI is occurring in the customer relationship management (CRM) realm. Companies are starting to devote considerable resources and effort to being far more precise in anticipating, understanding and targeting customer needs with specific offerings.

We find also that inevitably no organisation is going to have perfect or complete information on all of the strategic or operational matters that it needs to make decisions on. Building decision support models is one way for organisations to tackle the challenge. Robust hypotheses -- about those variables they do not have empirical information to support -- are integral to the process. BI therefore needs the latitude to formulate hypotheses around "unknowns" and to work in the realm of "conscious ignorance" by using business judgement, until such time as those hypotheses can be refined or the knowledge gaps closed.

The most successful businesses are likely to be those that can use knowledge strategically to improve the accuracy of the assumptions involved in business decision making. This approach -- "heuristic"', currently bandied around in the knowledge economy -- is relevant to the learning organisation. These businesses are also likely to be the ones that can most successfully incorporate sufficiently intelligent filters into analysis and reporting tools, while at the same time, not attempting to "replicate the executive mind".

BI: business intelligence

KM: knowledge management

CRM: customer relationship management

EIS: enterprise information system

DSS: decision support system

BPR: business process re-engineering

SLA: service level agreement

CI: competitive intelligence

BPM: business performance management

ETL: extract/load/transform tools

OLTP: online transaction processing

Join the CIO Australia group on LinkedIn. The group is open to CIOs, IT Directors, COOs, CTOs and senior IT managers.

Join the newsletter!

Error: Please check your email address.

More about Decision Support SystemsEmpiricalErnst & YoungErnst & YoungForrester ResearchInformation ToolsMeta GroupNordstrom

Show Comments

Market Place