By Alberto Roldan - Copyright 2010
Introduction
Companies are eager to implement business analytics to help them reduce costs and increase revenues. These companies face an unchartered territory in the area of analytics and need assistance in how solve implementation issues. Among those issues are mapping business objectives to analytics, resources, budget, data understanding, and planning. The objective of this article is to assist companies in dealing with those issues.
For the purposes of this article, we will equate analytics with the ability to predict the probability an occurrence or an event in the future. Companies are using analytics in many ways to predict the probability of
1. Transactions that have a potential to be fraudulent in market surveillance institutional trading and in health care claims processing.
2. Clients that should be targeted to buy different software products that are bundled together.
3. A CPG competitor’s product gaining track in the marketplace through the use of data from social media and its internal data.
4. The patients that are more susceptible to multiple chronic diseases.
5. The failure of machine parts within a specific product at the customer site and during the manufacturing process.
Organization Structure
Companies want to know the number of resources needed for a successful analytics project. Although the number of resources varies from project to project, the rule of thumb is that the core team consists of three people: a statistical modeler, an analyst, and developer. A fully functional team is also going to include a leader, a manager or project leader, business analysts, and evaluators/testers.
One of the main issues is how to find and budget for the proper resources. Individuals with statistical backgrounds are hard to find in the marketplace, as well as developers in specialized analytics software. A common mistake is to equate a developer with a statistical modeler. These skill sets are different, and understanding this difference is one of the keys for successfully implementing business analytics.
Outsourcing the analytics project is a good and cost-effective solution if the outsourced company has industry-specific knowledge and statistical modeling experience. There are different business models that allow a company to be successful using this approach:
1. Start with a proof of concept (POC) and move to larger projects.
2. Ramp up with the outsourced company, but then move to bring the analytics area into the full control of the company.
3. Fully outsource for the long term using a revenue-sharing model.
Other companies prefer a staff augmentation model. Personally, I have only seen this model work for companies that already have a well-organized organization structure for their analytics.
Budgeting for these resources and the analytics software is another challenge. It is a traditional market forces issue: high demand for these resources but low supply in the marketplace. The result is that you are paying a premium for these resources. In order to justify these resources, you need to show a return on investment (ROI). It is difficult to show an ROI in an area that you have no historical experience. The best solution is to start with a well-defined (including the budget) small project and use the results to extrapolate an ROI. Outsourcing the POC is a potential solution to this issue.
Mapping Business Objectives
One of the keystones in the successful implementation of business analytics is to specifically map the business objectives to the question that you want to solve. Although on its face this seems like a fairly simple issue, it is recurring flaw that I see. Companies want to know how they can use predictive analytics instead of defining the issues that they are seeking a solution. The first step in implementing analytics within a company is to have a clear business understanding of the issue. Analytics are based on a combination of mathematics and business knowledge. This knowledge is precise to solve a specific question.
The leader of the analytics team must ensure that the best practices are followed, including mapping out business objectives to the specific questions that analytics answers. The manager of the analytics team is responsible for identifying the metric for success used to evaluate the project. An analytics project should always be measured in business terms. How much potential revenue or cost savings does it identify?
Data Understanding
Two of the main issues are data quality and availability. The results of an analytics project are directly related to these two issues. The best practice in dealing with a data-quality issue is to deal with these issues before undertaking an analytics project. The data does not need to be perfect for an analytics project to be successful, but leaders and managers must understand and agree upon the limitations facing the project. Also, there are statistical techniques that can be used to refine the results and minimize data-quality issues like segmenting the results into “expected” results or those with a high probability of data-quality issues (“unexpected” results) for evaluation purposes.
The issue of data availability is more complex since there are some issues that we can extrapolate results even without having the data available, and sometimes this is impossible without additional data. A CPG company could extrapolate the results of an ethno demographic trade promotion analytics project into stores that they do not have data, but that falls within the same classification as similar stores where that data is available. The issue to validate the extrapolation is whether the segmentation or classification is valid or not. For example, a small store (by sales volume) in Southern California may not be comparable with a small store in Miami because although they both are catering to Latino customers, the characteristics of both populations may be different. On the other hand, a segmentation of small stores within a specific geographical area of Southern California could be useful to extrapolate the results within that geographical area in stores that do not have data available.
Implementation Planning
Implementation is essential in analytics because that is the actual transformation of data into actionable information. Planning how you intend to use analytics and the nature of your audience becomes essential to maximize your ROI. An understanding that analytics has different meanings within a company is fundamental. Executive management may want to know whether tactics are properly aligned with strategic goals. Line management would like to know how to best accomplish their specific monthly objectives. The employee on the field would want to know how to accomplish today’s target.
The visualization of analytics results, through dashboards, provides the means for decision makers within a company to quickly grasp the meaning of the information. Companies should think about how they want the information layer to be presented before embarking into an analytics project. The information layer should be directly correlated to the business objectives. Also, it should be flexible to add new requirements.
Companies should consider utilizing the advances in visualization techniques when planning a dashboard. This includes the ability to see the condition of a company through business metrics in a 3-D manner. The utilization of a 3-D graph that incorporates dollar value, statistical control process comparison, and predictive analytics is a powerful tool that allows using analytics in strategic and operational areas. The utilization of 3-D graphs increases perception and the ability to detect patterns by over 40 percent. Companies should ask, “Is there a value in increasing our ability to detect patterns in the data by 40 percent?”
Conclusion
As companies embrace and streamline analytics projects into their operations, they should plan to face issues that can derail their goals. Just a few years ago, it was common to hear that 50 percent of all IT projects were never completed. One of the lessons learned during that time was the importance of limiting scope, budgeting for resources, and project planning. A modified version of those lessons should be used when planning analytics projects. Think big, but start with small measurable projects. Avail yourself of best practices, identify potential issues early, and use experienced resources to ensure the success of your project.
About the Author - Alberto Roldan, the author, is responsible for the Enterprise Analytics Practice- North America, at Cognizant Technologies. He a published author with over 18 years experience in business analytics. He has a BA from the University of Michigan and a JD from the University of Puerto Rico Law School. He can be reach at alberto.roldan@cognizant.com
Copyright 2010
