You are currently browsing the tag archive for the ‘modeling’ tag.
For the past several years Ventana Research has focused more on analytics and their importance to improving business performance. We’ve done extensive benchmark research in business analytics, detailing how they are used generally in business and in major functional areas of companies as well as their application in specific industries. We adopted this focus because technology advances are changing the landscape of analytics. Its use in business management, for example, is getting new scrutiny these days because of three important changes in information technology.
One change is the increasing wealth of data that companies can use. It’s not just the data now available in the cloud. Over the past decades, organizations have implemented a range of systems for managing core business processes and collecting the data that go with these processes. ERP and CRM systems were among the first, but especially in larger companies, just about every function and every department uses some system that collects data. Almost all of these systems store this information in ways that make it feasible to access it. Second, so-called big data is making it possible for organizations to process much larger data sets than ever before to gain intelligence and insight into business operations and markets. Third, in-memory data processing is enabling companies to get immediate answers to queries, even through complex analyses of very large data sets, rather than having to wait minutes, hours or days. This accessibility changes the dynamics of planning and review meetings, for one thing, because it enables a far more fluid and interactive dialog around the questions “Why did we get the results we got?” and “What should we do next?” than has been the case in the past.
Yet all of this progress shouldn’t obscure the enduring value of simple ratio analysis. This technique for understanding business performance predates even the adding machine, going back centuries. Although it is widely used in the finance function, I think most companies today underutilize ratio analysis. Our benchmark research in finance analytics shows that finance groups do a good job with basic, well-established metrics such as profit margins or days sales outstanding (DSO) as well as debt and liquidity measures. But they – and the rest of the organization – do less well in monitoring and reviewing ratios that combine financial and nonfinancial data, especially where these involve key performance measures. Ratio analysis can help here.
It is particularly useful for assessing the efficiency of processes and the effectiveness of results, and at its core, business is about transmuting inputs into outputs, such as pounds or kilos of steel or direct labor hour per completed product unit. Indirect cost efficiency also can be measured as a ratio, such as the number of full-time equivalents (FTEs) employed per 1,000 invoices processed. Effectiveness can be measured by, for instance, the percentage of repeat customers, manufacturing defects per 100 units or, in customer support, the percentage of first-call resolutions.
Finance departments tend to focus on financial ratios and overlook operational ones, which may be viewed only by that specific part of the business. Thus, a periodic assessment of the profitability of a particular retail store may only include revenue and costs. However, without looking at the gross profit per sales employee and/or the average revenue per sales employee, it’s difficult to distinguish between the direct and indirect factors that are determining branch profitability.
Because they measure the relationship between inputs and results, ratios are especially useful as quantitative performance metrics. Potentially, there are thousands of these ratios that a company can use for setting objectives, monitoring results and assessing performance. However, it’s important to focus on the “key” performance ratios – those that have the greatest impact on the results of individuals, business units and the company as a whole. Companies can have a difficult time identifying their key factors. This is where driver-based modeling and planning come in because the process of creating these models sorts out the important from the marginal measures.
The use of advanced analytics is growing in importance as technology provides companies another way to achieve an edge on their competitors. At the same time, it’s critical that executives and managers build on the basics. If an organization cannot formulate the most important ratios that define business performance, and if it cannot readily access the data needed to perform this simple division, it’s unlikely to be able to handle large sets of data effectively and benefit from more advanced analytic techniques. Instead it is likely to wind up experiencing the “big garbage in, big garbage out” syndrome.
Robert Kugel – SVP Research
Our benchmark research on business analytics finds that just 13 percent of companies overall and 11 percent of finance departments use predictive analytics. I think advanced analytics – especially predictive analytics – should play a larger role in managing organizations. Making it easier to create and consume advanced analytics would help organizations broaden their integration in business planning and execution. This was one of the points that SPSS, an IBM subsidiary that provides analytics, addressed at IBM’s recent analyst summit.
Predictive analytics are especially useful for anticipating trend divergences or spotting them earlier than one might otherwise. For example, sales may be up compared to a prior period, but is it simply month-to-month variability or the start of an upward trend? Better analytical techniques can help distinguish between normal variation and the beginning of a new trend. By using analytics, one might even discern that while the revenue numbers have been positive recently, the underlying data contains warning signs that point to diminishing volumes, lower prices or both in the future.
Predictive analytic models are created using a top-down or a bottom-up approach, or some combination of the two. SPSS offers tools to handle both. The top-down approach involves creating a statistical hypothesis based on business observations or theories and then testing that hypothesis using statistical methods. IBM SPSS Statistics enables users to build a relevant picture from a sample, as well as test assumptions and hypotheses about that picture. A bottom-up approach unleashes automated data mining techniques on data sets (typically large ones) to distill statistically significant relationships from them. SPSS Modeler is designed for use by experienced data miners but also business analysts to speed the creation and refinement of predictive models. Often, companies employ both approaches iteratively to refine and improve models.
I used to joke that the main value proposition of SPSS was that while its chief rival, SAS, required its users to have a Ph.D. in statistics, SPSS could be used by anyone with a master’s degree. Applying predictive analytics techniques is simple in concept but far from simple to integrate into day-to-day business beyond its traditional roles such as market research. This partly explains why so few companies have woven predictive analytics into their planning and review cycles. It’s possible to create relatively simple predictive models, but for many business issues, such models may be too simplistic to be useful. And they may not be reliable enough because they generate too many false positives (people spend too much time chasing non-issues) or false negatives (missing important developments or breaks in trends).
Beyond the data and technology challenges posed by advanced analytics, there are significant people issues that companies must address to make their use practical. These can be more difficult to tackle than most business/IT issues because of the experience and skills that are needed that our benchmark in predictive analytics still finds lack of adequate resources. Automating general business processes, for instance, requires bringing together business subject-matter experts with people who understand IT. Advanced analytics, however, requires three sets of skills – business subject-matter expertise, IT and statistics – that are rarely found in any single individual. Communication among sets of individuals who have these skills often is difficult because they have a limited appreciation of the others’ domains and often have difficulty expressing the nuances of their own area of expertise.
Today there’s greater focus than ever on analytics, partly because an explosion of available data has made it possible and even necessary to make sense of it. As part of IBM, SPSS has been benefitting from the parent’s “smarter planet” marketing theme. SPSS also has taken steps to expand demand for its tools by reducing the people barriers to adopting advanced analytics. One step has been to automate data preparation for use in Statistics and Modeler. Another is an automated modeler that takes several different approaches to analyze a set of data in a single run and then compares the results. Yet despite these steps, I expect advanced analytics to require specialized skills for many years.
Therefore, I also expect adoption of advanced analytics to happen slowly. Most executives at the senior and even middle levels of corporations have limited familiarity with advanced analytics. Many may have had their last formal education with statistics as a required business school course. To spur broader adoption of predictive and other advanced analytics, IBM and others must foster a “pull” approach to marketing analytics. Business executives need to know that advanced analytics are available and of practical value, especially outside of traditional statistics-heavy realms such as consumer research and fraud detection. Sales planning, financial planning, enterprise risk management, maintenance and customer service are all areas ripe for use of predictive and other advanced analytics. We found all of these as future use of predictive analytics in our benchmark. It’s easy to convince analysts like me of the value of analytics; it’s much harder to get business executives to incorporate them into day-to-day practices. It would be helpful for its own cause if IBM SPSS were to identify promising uses of advanced analytics by function and industry and provide a canned blueprint that can serve as a starting point. Such a blueprint would incorporate a business case illustrating the problem, the suggested steps for addressing it and the scope of benefits that can be realized.
The continuing explosion of data will give rise to an increasing number of ways that business and finance executives can use information to their advantage. But first they have to know that they can.
Robert Kugel – SVP Research