"Data warehousing is not an end, but a means to an end," says Bruce Jones, sales support manager at SAS Institute; the end being the delivery of business value to companies treading the data warehousing road.
Countless businesses should have heeded these words a few years ago before falling into the trap of rushing into data warehousing, often because their opposition did, and with little consideration for the value they wanted to derive from the project. The result was a lot of burnt fingers and a tarnished reputation for data warehousing. However, this image does appear to be changing.
"Data warehousing is something of a swearword in many sectors," says KPMG Consulting system integration director Hubert Wentzel, who talks today of a third wave of data warehousing in a post-dot-com-boom world where businesses are far more circumspect about their investments.
The problem, says Wentzel, is that too many businesses rushed in with little idea of what they hoped to achieve, or even less idea of the kind and volume of data they had to hand. Data warehousing was promised as the silver bullet to solve all their analytical needs. It soon became apparent that even the best analytical tools were useless in the face of inadequate data and a lack of focused business goals. Also, data warehousing implementations were sold with the promise of realising immediate increases in returns from existing client bases.
Wentzel says that in reality, much of this value is not always immediately achieved, if ever. Unlike the data warehouses of previous years, most companies today see returns come from improved internal processes and better efficiencies in operation.
Data warehousing is a means to an end, not an end in itself.
Bruce Jones, sales support manager, SAS Institute
"Research over the past few years might well have concluded that it does not spell failure because up to 80% of projects had to change their design criteria in terms of technology or structure during construction," says Andrew Connold, MD of Synergy Computing.
"What it does indicate is that the construction of a warehouse is such a vast, complex and time-consuming project that designs have to change during construction to accommodate changing business needs," notes Connold. "This is why it is better to build a data warehouse incrementally using conformed data marts that address specific pain points first, rather than try to execute a big bang project that is not only outdated by the time it is complete, but cannot deliver any return on investment until fully commissioned."
"Specific processes should be put in place to consistently check the integrity of the data," says Reg Butcher, project manager at arivia.kom`s Focused Business Solutions division. "If the quality of the information is poor, then the data from the data warehouse will not be useful. Companies must realise that the return on investment is not gained from expensive multi-functional tools, but rather by how the information is accessed and used. A data warehouse does not have to be large with terabytes of data for it to qualify as a true data warehouse. By starting small, with focused data marts, users can achieve quick deliverables within a short period and after a small capital outlay."
Stupid data
Data warehousing is something of an amorphous topic. For a start, sticking a whole lot of a data onto a disk and ferreting it away in a dark room does not solve any problems or offer any value. Without the appropriate analytical tools, data is all but useless, apart from its own intrinsic value. Rather, the business value of data is derived from holding acquired data up to logic models and analysis that can make sense of it. Today this is called business intelligence, or sometimes even analytical customer relationship management. Without the required analysis, data warehousing is useless. And data warehouses are not just copies of existing operational data.
"The major pitfall of data warehouses is that they are often categorised with operations systems," says Butcher. "Customers need to understand that they are actually snapshots of the operational data systems at a specified granularity and point in time. It is of no benefit to the company if a multimillion-rand data warehouse is merely used to draw the same reports and information they received from the previous operational systems."
Butcher adds that the perceived failure of data warehouses is often attributable to the fact that many firms do not continually update the data warehouse with new business changes, or synchronise it with the company`s system so that it continues to serve its purpose.
"There is a common misunderstanding about what a data warehouse actually is," echoes Jones. "In reality, it is a storage place for structured data in a format that makes it ready for analysis."
The key to making the most of information systems, says Wentzel, is a flexible but structured data store, because the emphasis should be on the results delivered and not on the actual contents of the store.
System integration
Who should establish a data warehouse? The answer is not a simple one and largely depends on a combination of volumes, the complexities of systems in use and the values that can be derived from the implementation.
Jones says a smaller business with a largely homogeneous system set-up is the least likely to gain from the implementation of a data warehouse and "can often access the information it requires through the daily operational data". As systems become less homogeneous and as different businesses merge, the need for a data warehouse grows.
One of the typical drivers for a data warehouse is the present dispersed nature of information. In today`s PC-based operating environment, data is often hard to access and is not centralised. Also, many organisations have a range of operating platforms and tools, many of which are incompatible, and while tools may come and go, very often the data that relates to clients that require servicing after the life of the tool does not go away.
This leads to another key factor in the success of data warehouses: system integration. As with business intelligence, data warehousing is all but useless without effective system integration. Jones explains that in previous years, data warehouses were most often set up by IT departments and while the data was in effect "warehoused", the data was most often stored in its original form which made accessing the information difficult or even impossible.
It is of no benefit to the company if a multimillion-rand data warehouse is merely used to draw the same reports and information they received from the previous operational systems.
Reg Butcher, project manager, arivia.kom`s Focused Business Solutions division
What is needed instead is data that is collected at regular intervals from operational databases, which is then cleaned, classified and stored in standard formats for later retrieval.
"Too often data warehouses are extremely complex in their design and as a result are cumbersome to update. A mind-shift is needed," says Butcher. "The key to success is the realisation that the simpler the data warehouse architecture and the more user-friendly it is, the more it allows the user community to use its business intelligence and to access valuable information. Furthermore, by using the correct business intelligence retrieval tools, data can easily be retrieved from the data warehouse and analysed for various business purposes such as developing balanced scorecards and value chains, identifying trends, and forecasting analysis."
Ownership and maintenance
One issue that everyone agrees on is that for data warehouses to be successful, they require buy-in from IT departments as well as the business side of the organisation. A lack of business buy-in often resulted in the failure of date warehouse projects in previous years because the implementations lacked the defined goals that offered business value. Wentzel says the third wave of data warehousing is characterised by the focus on "specific strategic drivers".
In the past, IT departments would set up the data warehousing projects and because business users could not get hold of the data easily, they often set up their own projects with the result that each department had their own intelligence-gathering methodologies.
Pat Holgate, managing consultant of Teradata Solutions, says "years of creating a new data mart every time a department needed to do some analysis are coming back to haunt them". According to Holgate and a survey conducted by the company, "the average annual support cost per data mart is $1.5 million to $2 million. The aggregate cost of all the data marts being funded independently at the department level would shock most CIOs and IT managers; they are only now coming to realise how costly data mart proliferation has become.
"With 59% of the companies in the survey maintaining up to 30 data marts, and some as many as 100, annual support costs easily dwarf the business benefits the data marts were intended to bring."
Return on investment
Most vendors claim that the return on a data warehouse solution is enormous; many talk about 1 000% returns and more. While in some cases these returns are achievable, this is not always the case, with the result that many organisations have been burnt by skewed expectations.
The data warehouse might not be the IT flavour of the month, but it still has a major role to play in the future of companies.
Andrew Connold, MD, Synergy Computing
Most vendors today shy away from the return on investment (ROI) hard-sell and Jones says that when talking to clients, the returns are often played down. Measuring ROI is also a contentious issue. "ROI as a measure has been over-traded," says Wentzel. "You have to be clear on what benefits can be achieved from clean data; what it changes at the coal face." This encompasses the operational efficiencies and improvements that can be derived from clean data.
Holgate agrees: "The real ROI killer is lost business value. When data is spread everywhere, it is difficult to find answers to questions critical to understanding current business performance and making decisions that affect future enterprise competitiveness."
Holgate argues that data mart proliferation leads to time-wasting inter-departmental arguments over who has the correct view of the data. Added to these difficulties is the unproductive and frustrating activity of developing and maintaining new data marts over and over again.
"The data warehouse might not be the IT flavour of the month," says Connold, "but it still has a major role to play in the future as companies, driven by the need to manage corporate performance, have to be in a position to measure critical success factors at an enterprise level. It is just that today, new technology obviates the need for having to wait two to three years before being able to leverage a company`s information resources."
Jones says that data, when structured according to its intended business purpose, plays a critical role in allowing organisations to not only gain from business intelligence but also from advanced analytical intelligence that "looks into the future".
"Analytical intelligence goes one step further than business intelligence by looking into the future as well as the past. It does not merely look in the rear-view mirror, but looks ahead to predict future happenings based on past performance."
If a data warehouse is not changing the way you operate, then it is probably not being used.
Hubert Wentzel, system integration director, KPMG Consulting
By looking towards the future, analytical intelligence enables managers to make sound decisions about where they should steer their organisations to maximise profit and customer value.
There is, however, one cautionary note from Wentzel: data warehouses need maintenance and change because they are "living organisms". He explains that if a data warehouse is not changing business processes, then it is probably not working. If it remains static and unchanged, it probably means that business is not using it in everyday decision-making processes.
Share