|
|
Computing SA - 29 August 2005
by Samantha Perry
Back to basics for BI
Computing SA gets around the plethora of BI hype in a bid to establish
what BI can do, and how to get the best out of an implementation once you
have decided to take the plunge.
|
Business Intelligence (BI), like too many technologies, was over-hyped in its early days. Again, as with many other technologies, this hype was rapidly followed by some very public implementation failures, before everyone settled down and realised, as usually happens, is that BI is not the be all and end all that it was often claimed to be. Now that a measure of sanity and maturity has crept into this market segment, end-user organisations may want to re-look at BI, or possibly re-look at failed/under-performing implementations, in a bid to derive the value that was originally promised for this investment.
A little bit of History
"If one takes a step back", says Davide Hanan, MD of Cape-based BI Distributor QlikView, "one should try and understand what BI's objectives are, and why we have it all. If you look back 20 or 30 years, the focus of IT has really been in terms of automating and integrating business processes - just about everything in business today has been automated.
"In that process, a lot of information has been collected - business systems are very good at capturing information and putting it into data bases."
Organisations, thus had a lot of information, but no real mechanism to access it - and it is where BI came about because there was a very serious gap there, he adds
|
"BI's aim is to give business information back to the users, and allow them to do analysis on it and get insights into the business and its operations, so that they can make effective decisions. And, while there is no doubt that it can do that, the problem to date has been in the way that information delivery is done" Hanan adds.
Bill Hoggarth, MD of the local office of BI and analytics vendor, SAS Institute, says that just ten years ago, state of the art BI consisted of ad hoc departmental desktop analysis. "Managers had executive information system (EIS) dashboards, which were typically inflexible and difficult to manage," he states. "At the same time, ERP systems were emerging and promising enormous intelligence benefits. However, these function exclusively in the operational space, and despite the promises, are unable to provide intelligence delivery across the enterprise.
"Traditionally," he says "BI concerned itself with a data access and management, as well as reporting. Intelligence solutions today transform the huge volumes of data, voice and text collected by organisations today into actionable intelligence. BI has expanded into forecasting, predictive modeling and optimisation. Thus, instead of answering rear-view mirror questions like "What happened?", intelligence today answers questions such as "What will happen next?" and "What is the best that can happen?"
|
|
"BI is a tremendous growth area in IT, but it is important to understand what it can deliver", cautions Keith Fenner, director Solutions & Business Development at AccTech Systems in Durban. "BI is a way to measure performance, and, when combined with other key performance indicators (KPI's), can deliver accurate, timely information to the right person at the right time.
"But," he says, "a BI project must not be considered in isolation. In order to deliver, it must capably consolidate information from all areas of the business across disparate systems in order to create a single warehouse of data. Furthermore, the dashboards of information must be appropriate to the decisions being made. There is little value in delivering supply chain data to the credit control manager. It is the delivery and interpretation of data that adds value to the organisation."
While BI can provide a useful snapshot of a business' state of affairs, it cannot, as Magix Integration director, Amir Lubashevsky, points out, make business decisions for you. "BI tools are not the kind of tools that will provide analysis and decisions," he adds. "You also need to fit your BI system to another system to get real value out of it - it is not a business solution on its own. BI can do powerful reporting and analysis, but, unless it is integrated into something before that or after that, you will not get the maximum value out of it."
Worst case scenario
"Bi has been guilty of information overload, and it should be remembered that BI is the tool that delivers performance and analytical data to make decisions. It should assist with uncovering trends, good or bad, and deliver this to the right management team at the right time. The presentation of the data is crucial to the project's success, and understanding how the
|
data is to be received, and by whom, is vital," says Fenner.
Many BI tools also hang around on corporate IT systems, gathering virtual dust, until the time comes for the annual report to be generated. Linda Joyce, client liaison of Bateleur Software, says that from a user perspective there may be many valid reasons why the tools are not used. "The reports take too long to produce the results of a query, the tools are difficult to use, or other workload priorities prevent users from installing the software and getting to grips with it," she says
Dave McWilliam, MD at BI vendor, Cognos SA, says that the major reason for BI failures at an enterprise level is that it is often not aligned with the organisation's strategy, and, therefore, is not able to deliver true value as an enterprise BI solution. "In many cases, BI has been implemented to address isolated needs within an organisation or department, and has been successful at that. However, it has not been designed as the delivery mechanism for strategic decision-making.
"In order to provide the right level of visibility into an organisation's performance, BI needs to be elevated to a level where it is viewed as the standard provider of intelligently interpreted information, and used as the basis for decision-making throughout the organisation," he notes.
"BI is of limited use if it is not embedded into business processes, and command over it given to business users, as opposed to IT specialists," states SAP Africa's BI solution manager, Mahomed Asmal. "It is an irony that, while most people will agree that an organisation's information is its single greatest asset, primary access to, and control of, the asset still resides in the hands of IT specialists, who are not directly involved in the day-to-day operations of the organisation. Which is a bit like sending the physiotherapist onto the pitch to face
|
|
Shane Warne. There is no direct relationship between the asset and the business it is supposed to enhance."
Lubashevsky agrees: "BI has been historically implemented at a high level in an organisation, and only in one or two departments - not throughout the organisation. If you do not take an integrated approach in terms of processes throughout the organisation, BI becomes the domain of IT or subject to only ad hoc use my management. BI should become commoditised, and be a normal part of the enterprise's systems and day-to-day activity.
"Measure, analyse, act and improve, needs to be a day-to-day thing" he comments.
A matter of faith?
"In order for users to have confidence in insights derived from a BI solution, they must trust the data on which those insights are based," states Paul Morgan, MD at ASYST Intelligence. "It is essential that the data used to make intelligent business decisions is accurate. Most organisations have a variety of databases and/or data marts storing their information. Before utilizing data for BI, an organisation must ensure data quality through processes that control data entry, validation, and auditing."
Synergy Computing's Andrew Connold agrees, "It is the old story of garbage in, garbage out. BI does not clean up dirty data, although it is often blamed for it," he says. "A BI implementation will also fail to deliver, unless it is driven by management. It is critical to have a champion high enough to ensure that data is input correctly, and that the cubes and reports generated are relevant.
"Additionally," he says, "businesses change, and change often. Unless the reports and cubes are kept up to date, they will become less and less relevant to the business over time."
|
"If you are clear in the objective of BI then it can do what is meant to be doing," notes Hanan, "which is to give insight into the company. BI has failed to date in the way it has delivered information to users, not in the objectives of its implementation. Delivery mechanisms are still very report-based, and very often paper-based - and it makes no sense today to analyse information on paper. A good analogy is that any research done today is done online, initially at least - you have hypertext and indexes and can move from place to place very quickly. Compare the effort needed to do any research 10 years ago, and it is like chalk and cheese. And yet, when it comes to analysing business information, people still work with paper.
"BI answers predefined questions that have been asked before, and you cannot ask new questions - this is its biggest downfall," he maintains.
"Because you cannot ask new questions, you generate 20 or 30 reports to look at the data in different ways - this does not make sense in today's world. Analysis has to be electronic - then it gives you all the flexibility you require."
Hanan echoes Lubashevsky's assertion: "Another reason why BI has not done as well as anticipated in the early days is because it has been driven by IT. BI belongs to the users, it does not belong to IT. The BI manager should not be an IT person - it should be a user, which will give a totally different emphasis to the project," he states.
Success Factors?
"It is critical that all stakeholders are aligned and support a BI initiative," says Morgan. "The owner must champion the BI initiative throughout the organisation, where the organisation may be a department, a division, or the entire enterprise. The champion, in conjunction with a chosen supplier, needs to interview all relevant
|
|
stakeholders to ascertain and document true information requirements. An implementation can only be successful if it delivers the information that the organisation needs to make decisions.
"BI implementations should be kept as simple as possible, with achievable gains that can demonstrate measurable results in short periods of time," says Marc Scheepbouwer, MD of Intellient. "Traditionally, projects undertaken by IT departments are monolithic, driven by the need to satisfy technology requirements and characterised by weeks of implementation, months of ironing out bugs and the absorption of millions of rand, only to be faced with user resistance at the end of it all.
"Business needs are different to those of the IT department, however, and there is no reason why a BI project should follow in the footsteps of its brethren. Rather, BI projects must be driven from a business perspective instead of technology one. To be successful, the information produced through business intelligence must be usable and useful, no matter what format it is presented in," he emphasizes.
Connold re-iterates this point: "On new installations in particular a common problem is that the company tries to make the implementation too complex. It is best to try and focus on few key drivers in any participating area - drivers that are going to have the most impact in overcoming a pain point or driving a new strategy, for example"
Lubashevsky says his company advises that end-user organisations implement BI tools as early in the company's lifecycle as possible, so that the solution can grow with the company, and become part of the culture.
Onwards and upwards
Going forward, technological progress, and the widespread availability of disk space, memory and processing power are changing the face of BI tools. Says Hanan: "BI was born in the late 80s,
|
when memory and access to processing power were limited - the only thing we had in good quantity was disk space. BI had to be developed around these hardware constraints, which is why, when you work with OLAP today, you still find that the first thing people do, when building cubes, is pre-aggregate the information.
"So, for example, if you want to analyse sales by a salesperson, area and product, you typically have to predefine the hierarchy, so that the OLAP cube can pre-aggregate the information, so that when you do the enquiry your PC does not need to add that information for you. The query is thus predefined for you, and you have to analyse the information in that order.
"What are emerging today are new technologies, like associative query logic (AQL), which work on the premise that users do not have hardware constraints, and have access to as much memory and processing power as they want.
AQL thus lets users load the data they are analysing into memory - no predefinition and pre-aggregation whatsoever."
"AQL is truly online," he adds. "It extracts and presents the data to the user in any way they want to fly."
AQL is limited by the amount of information that can be held in memory, a problem which, Hanan says, is resolved by the move to 64-bit environments.
"Technology like this makes it easier to support massive usage of data, and move towards information democratisation (where all users have access, not just a privileged few)," he enthuses. " This moves us well towards guided analysis - instead of having fixed reports, it actually gives you several views of your data and lets you, at a click, filter it in any way you want to. The views effectively guide you to the information you should be looking at," he explains. "There are multiple views
|
|
available to you, and you can filter it in any way you want to. This means you can make information available to many more users.
"The next step is that, instead of users going looking for information, the information will go to the user. This will be the next stage in BI, where people will be working with alert systems and the like," he concludes.
|
|
|
