≡ Menu

First Look: i4C Analytics


i4C Analytics is focused on delivering advanced analytic applications for specific industries and business decisions. These packaged analytic applications apply predictive analytics in real-time in a highly automated environment. The applications use automated analytic model creation as well as allowing data scientists and predictive analytic professionals to integrate their own models to solve the business problems involved. i4C Analytics began in 2002 as a consulting firm and found that an ever increasing percentage of their consulting projects could use the same infrastructure, the same product, in many different solution areas. i4C Analytics is based in Italy with approximately 75 employees as well as a network of more than 20 established predictive analytics distributors around the world with half local to Italy and half international. They are a significant IBM partner and are increasingly working with other large system integrators also. About 10% of their revenue is international today and both this revenue and the company as a whole are growing strongly.

i4C Analytics’s applications are focused on delivering decisions, actions, into a business process. This is delivered by their stack – called ACE. ACE stands for the Application Configuration Environment and ACE is designed to solve complex problems that require advanced analytics while relying only on configuration to allow rapid implementation of easy to use business applications. ACE delivers forecasting, optimization and predictive analytics to business-process-specific solution (like Next Best Action or fraud detection). These analytic capabilities are based on algorithms from R, IBM SPSS, Matlab or Lindo. Each specific business solution uses a different combination of these capabilities. For instance a specific application for managing spare parts inventory might be delivered on ACE using forecasting from R or IBM SPSS. ACE contains:

  • Configuration tools
    Files, formulae and wizards
  • Analytic Pillars
    Forecasting, Optimization and Predictive. In addition there is support for data quality, what if analysis and reporting
  • ACE Core
    Logging, entity management, user management, task management, formula and wizard editors
  • 3rd Party components
    Databases, Hadoop, R or IBM SPSS algorithms etc.

Predictive analytic models can be included in ACE in a variety of ways:

  • Easiest is to use an existing model type provided in the analytic pillars. This involves simply selecting an appropriate technique, setting parameters and saving.
  • Next a custom model can be included that uses custom modeling syntax but takes advantage of the data preparation and execution of pillar models. This requires more analytical expertise but the models are extremely easy to integrate.
  • Either of these kinds of models can be saved as a template to allow new models to be added based on the template
  • A new model can be written using a formula editor that allows libraries and variables from inside ACE to be mixed with custom R code, for instance, to create a custom model.
  • IBM SPSS Modeler streams or projects can be loaded up and executed as part of ACE. This is expected to be extended to additional commercial tools in the future.
  • Finally it is possible to integrate new analytic platforms, such as SAS, into the stack if desired.

This variety of approaches means that companies with varying degrees of analytic know-how can use ACE, adding their own expertise in different ways depending on how sophisticated they are analytically.

Crucially with ACE the focus is on providing business solutions not just analytics. Regardless of the kind of analytics being used, they are embedded in business user-oriented interfaces throughout. When a user logs in to ACE they see a set of items, instances, based on the kind of business object being manipulated by the application for a selected business area. Various business processes and actions are available to use against these instances depending on the installed applications and the kind of business object being viewed. Instances can be viewed, visualized, drilled into for additional detail, reported on and more. Each ACE application has its own elements that are pre-defined. The various elements of an application such as available actions, alerts, tariffs or business rules can be configured if the user has the necessary access rights. Different user types are supported, allowing each application to be presented differently to different kinds of users.

For instance, taking the fraud detection application as an example, there could be three roles that play a part –an audit manager, an area manager focused on a specific business area, and a more technical user. When the audit manager logs in she can see various objects such as the company’s stores, the cashiers who work in those stores and the business rules currently being enforced. She could, for instance, report on the risk of fraud for specific stores/locations during a defined time window and visualize the results. Having identified a risky store she can focus on the causes of risk here. A store might, for instance, show a lot more instances of a particular fraud detection rule compared with other stores (all of this analysis is being done in the application). She can even see who is breaking the rule and when, enabling her to identify the causes of risk and inform local managers.

A local manager sees a more restricted view but can perform similar analysis. Because risk may also be reported using a predictive score – an anomaly score – not just rules he can analyze the cashiers based on this score. The drill down can reach down to specific transactions, those causing the anomalies, and he can package this information to report to HR.

Because of the problems the audit manager could decide to create a new business rule. As she creates the rule it shows her the distribution of transactions affected. The more technical user would then be informed and might decide to update the risk matrix that combines business rules and predictive indicators to rank transactions. In this case he could re-weight the rules and predictions to focus in different areas, visualizing the number of transactions in each matrix cell as he works. Once happy with the impact of the change he can push it into production.

More information on i4C Analytics and its applications is available www.i4Canalytics.com .


Comments on this entry are closed.