De meeste bedrijven gebruiken Business Intelligence om terug te kijken en op basis daarvan beslissingen te nemen. Hierdoor lopen ze altijd een stap achter. Business Analytics maakt het mogelijk om vooruit te kijken en daardoor proactief te handelen.
Is er een totaal nieuwe hype op komst? Of is Business Analytics gewoon een mooi synoniem voor datamining? Lees onderstaand artikel en trek zelf uw conclusies. Een ding is zeker: goed gebruik van uw informatie, proactief dan wel reactief, zal u een concurentievoordeel opleveren.
Intelligent data warehouses and statistical software are hardly new - so why is Business Analytics being sold as such a breakthrough for FDs who want to streamline their operations and take hold of strategy? Sellers point to hidden messages that emerge when larger amounts of data are probed with more sophisticated models. But the biggest gain is often from just knowing what’s happening here and now, and receiving the data in a form that makes immediate sense.
Aircraft used to be flown by the seat of a pilot's pants, but now the cockpit crew wear suits and control their flight path through a range of electronic measurements. Business processes have evolved the same way: gut feel is being replaced by measurement and analysis, helped by computers that can marshall previously overwhelming amounts of data and instantly subject it to statistical assessment that once took weeks to carry out.
A recent survey by enterprise software group Sybase and the National Computing Centre found that 85% of business intelligence (BI) implementations are driven by the need to improve the quality of management decisionmaking, and 66% target performance improvements. As the technology to dig deeper into data sets becomes more widely affordable and better understood, managers are finding commercial value in assessment techniques once confined to the arcane worlds of academic modelling and military strategy.
"Our original customer base was of researchers, weather forecasters, defence analysts – but now it’s migrating to enterprise, because it allows more powerful budgeting and modelling," says Philip Fraher, CEO of data visualisation and predictive analytics provider Visual Numerics. Another fast-growing business analytics provider, SPSS, started out as the Statistical Package for the Social Sciences, but now counts the corporate world among its biggest customers.
Secrets beneath the surface of the data flow
Statistical techniques for establishing causal links and developing forecasts using data have been under refinement for more than a century, and faster computers now allow once-cumbersome procedures to be run through almost instantly and quickly checked for accuracy. It’s now possible to supplement numerical data with 'unstructured' information, such as notes, emails and written documents, which can be scanned in, made searchable, and analysed for key terms. So an unprecedented breadth of material can be assessed in ever greater depth, within the timeframes that most businesses now run on.
As examples of what the new techniques can do, SPSS v-p of customer analytics Colin Shearer gives the example of two currently fast-growing areas, customer relationship management (CRM) and fraud detection. Consumer-focused companies can now analyse their point-of-sale data on individual buyers, and related sales and marketing information, to predict what named customers are going to buy next and where this will leave the aggregate production requirement. They can also identify the customers most likely to make repeat purchases, or to quit, and the profitability of those customers - devising personalised marketing strategies to get more from the ones whose share is most cost-effectively increased, and retention strategies for those on whom it's worth making extra effort to prevent them going elsewhere.
For fraud detection, now a reputation-saver as well as a money-saver for an increasing number of targeted companies, programs that rapidly assess a time-series data set can quickly spot unusual transaction patterns at a particular time or on a particular account, alerting data managers to the cases most meriting a manual check.
At Kognitio, which sells analytical solutions that fit on top of large data warehouses, marketing vice-president Sean Jackson admits that seasoned managers often know intuitively what the analysis later tells them. But intuition can be misleading, and isn't always there when most needed; "now we can validate those business decisions more accurately." Analytics can also give insight into why particular patterns are forming and repeating, extend that insight to managers who haven't been there long enough to develop the intuition, and find regularities or trends that even the most experienced executive wasn’t aware of.
Linking to the underlying hardware
Analysis requires data to be held in electronic form, and much time (and accuracy) has traditionally been lost keying-in information that starts life as hard- or handwritten copy. Many analysis packages also require a particular electronic format, and run into the problem that data held in warehouses must be transferred and translated before it’s of any use. A longstanding obstacle to turning essentially backward-looking business intelligence into more forward-looking business analytics has been that BI data resides in large hardware-encased ERP systems and warehouses, while analytical tools are contained in separately developed applications software suites.
For the applications to interface with the central store and get hold of the data in a usable form, it’s often been necessary to build on an intervening layer of 'middleware'. IBM’s market lead in this area provides the leverage with which it is now staging a late entry into BI, through the $5bn acquisition of Cognos launched at the end of 2007. Oracle, which moved into BI and analytics through the acquisitions of Hyperion and PeopleSoft, has stepped up its addition of the intervening middleware level through the just-agreed $8.5bn acquisition of BEA Systems.
Analytical applications providers are keen to point out that their products' basic building block is lines of code in widely-used computer languages, making them easy to slot into the programs underlying big ERP systems and data warehouses. At Visual Numerics, Fraher points out that the ultimate product is computer code that "can work with any type of data and sit on or under any big data pool." So ERP heavyweights such as SAP and Teradata have been able to licence the analytical software and build it onto their systems, adding such option as forecasting, pattern recognition and simulation into their list of functions.
But while companies that have invested heavily in ERP will expect any analytics to be built on top of it, the numbers don’t have to be run through such a system before they start to speak. "When run in a database, our analytical software will deliver insight you wouldn’t otherwise get. So customers with, for example, Oracle, will run our system on top of it," says Sean Jackson, Kognitio marketing v-p. "But some just run with us, as we can replace the database. It depends what the end-user wants to do." That ability to work with or around the large installed bases of hardware and middleware underpins Kognitio’s imminent launch onto the US market.
Depth vs speed
Knowing where the company is, in terms of resource deployment and income generation, is a necessary first step to knowing where it could be through modelling and simulation. So gathering and visually presenting data, quickly and accurately, is analytics' first mission, before any complex calculations are made with it. Analytics providers warn against running regressions before you can walk and talk through the basic data.
A recent survey by BI group Information Builders suggests that the speed with which people can obtain accurate, relevant data may be at least as important as the sophistication of the analysis that then gets performed on it. Over half of European employees said that lack of complete and consistent information was the biggest obstacle to accurate decisionmaking, and the average UK worker spent an hour a day looking for it. "Compliance is no longer what's driving the demand for BI. There's now an interest in getting a 360-degree view of the customer, by building systems that consolidate the data," says Information Builders vice-president for products Kevin Quinn.
Providers whose software uses in-memory processing (analysing the whole data set), rather than online analytical processing (which works with structured summaries of the data), argue they can deliver accurate forward-looking data in near-real time. "Companies need not just to achieve compliance and transparency, but to become more predictive. In the past it used to take three weeks to extract the right data, one month to produce a report. Now life's too short for that," says Andy Honess, UK managing director of in-memory pioneer Qliktech.
Operationalise before you optimise
Despite ever faster methods of data preparation and processing, there can still be a trade-off between instant presentation and in-depth processing of data – graphically illustrated in the more academic business journals, whose studies have often taken months to refine the data and report the calculations made with it. Time is needed to collect, collate and clean up data so that it's reliable enough to send to the warehouse and do complex calculations on. More time is needed to do the calculations, especially when they have to be run many times under variant scenarios to gauge the sensitivity of a result or find the standard error around a central forecast.
So analytical competitiveness is still principally a high-level, strategic activity, which managers use to build on the operational insights from business intelligence when time and space for investigation allow. They often don't - the Sybase/National Computing Centre survey found more than 20% of respondents less satisfied than expected with their BI because of slowness in implementation, in flow of data after implementation, or in people's getting accustomed to using the system and getting results from it.
Critics of BI say that even the generation and presentation of numbers is often too 'strategic', failing to achieve efficiency and agility gains which derive principally from managers and employees having an up-to-date picture of the business at their fingertips. They refer to 'operational intelligence' to distinguish it from strategic intelligence that might eventually tell you where you could have been and where you could have gone next, but arrives too late to make the best of where you actually are.
"It’s working in real time that makes the difference for business performance," says EG Solutions founder/CEO Elizabeth Gooch, whose operational intelligence system promises immediate results by making business intelligence data more readily and readably available. Her experience in financial service companies – now a major client sector - conformed that what doesn't get measured doesn't get managed: data that are slow to arrive or arrive in unprocessable hard-copy form just get left out of the calculations, whereas including them is vital to getting an accurate picture of what is going on.
Resolving any trade-offs between depth and speed of analysis, and weighing up the often short-run costs against the longer-run benefits, is essential to deciding which approach to analytics will yield the highest ROI. But these calculations are now being re-written, for many businesses, by new ways of delivering BI and data analysis, which is where this investigation turns next.
(bron: Accountingweb)
|
