PHASING THE DEPLOYMENT OF A DATA MINING
SOLUTION TO YOUR END-USERS
By Jennifer Parker Laflamme
Who will be the end users of a data mining solution is one of the considerations in the deployment of a data mining solution. Getting a solution into their hands involves data access considerations, deployment options regarding technological infrastructure, different scopes of a project implementation, and different end-user training requirements. But if done correctly, a phased rollout that provides a solution to various user-sets can result in the best chances for better decision making throughout the company and a continual competitive advantage.
In a typical business organization with an abundance of data (e.g. in the insurance or banking domains), possible end users of a data mining solution may be: quantitative analysts, strategic marketing specialists (responsible for marketing products across boundaries), specific marketing personnel (responsible for marketing within a specific business or investment area) and field agent/representatives (responsible for customer sales via interactions with customers).
For the quantitative analysts as end-users, a standard client/server data mining application distributed to a handful of users within the company may suffice. Since there are few users, software deployment, maintenance (upgrading software versions, etc) is manageable. A phased deployment can be achieved in manageable phases by first starting out by supporting the core analysis and then enhancing this with additional analysis afforded by the data mining software. Development time can be minimal (i.e. nothing more than the installation, and a connection to data stores) and training should be negligible since analysts are already familiar with sophisticated analytical routines, as well as the business domain.
Strategic marketing specialists represent another set of end-users for a data mining solution. This would typically be a more customized client version of the application used by Analysts. Since their expertise is marketing rather than analytics, some of the details of the quantitative analyses can be made transparent.
They could have access to the same data as for the back office analysts, since the end user would be interested in cross-selling at an enterprise-wide level. The scope of this project is also on the small scale -- with most effort and time devoted to developing an application that an end user at this level would find usable. Since these users already know the business domain, training would be limited to learning to navigate the application's user interface.
Marketing end-user (person responsible for a specific business/investment area) would require a very similar or even the same application as the Strategic Marketing person, but would only require the data pertaining to his marketing area. As such, database and data access specifics are transparent, and geared to his/her specific job responsibilities. Rather than accessing data from a data warehouse, access can be from a data mart. Since a company may have several end users matching this profile it would be beneficial to deploy this as an Intranet application--making deployments of new releases and interim updates very manageable and transparent to the user. The scope can be characterized as a medium project, piggy-backed on top of the previous two deployments as another phase.
Field agents represent another set of end users. This solution could be deployed across the Internet or as a customized client on laptops or remote desktops. Deployment is of course much more manageable across the Internet, but increased security concerns become an issue. The end user should be able to run the data mining application on data he has collected on his customer base as well as data provided by the home office pertaining to his customers and products. Training is more of an issue with the deployment since the users are in remote sites, but deploying in phases to agents in specific customer/product regions can reduce this.
A phased rollout of a data mining application deployed to the above sets of users should ultimately result is better decision making at all levels of the organization and sustained competitive advantages.
Jennifer Parker Laflamme is a Senior Analyst at Virtual Gold, Inc. She heads the Virtual Gold development team for IBM's Advanced Scout, a data mining program used extensively by coaches of the National Basketball Association to devise new strategies based on the automatic identification of hidden patterns in game data and video. She is also part of a team which offers consulting services to help organizations differentiate themselves with the strategic use of data mining technology.
Prior to joining Virtual Gold, Inc., Jennifer worked as a business systems consultant for American Management Systems, Inc. She also worked for IBM at the T.J. Watson Research Center.
For more information, see http://www.virtualgold.com.