First Look – G Stat

G-Stat is a privately held Israeli company focused on advanced analytical and data mining solutions. The founder of G-Stat began his career as an econometrician, focusing on large time-series data initially. Then he worked at the Bank of Israel in IT helping translate what the economists and others needed into information systems. He rapidly became interested in data mining and founded a professional services company that supports customer analytics for large companies in Israel.

G-Stat is a privately held Israeli company focused on advanced analytical and data mining solutions. The founder of G-Stat began his career as an econometrician, focusing on large time-series data initially. Then he worked at the Bank of Israel in IT helping translate what the economists and others needed into information systems. He rapidly became interested in data mining and founded a professional services company that supports customer analytics for large companies in Israel. This business focused on using SAS analytics tools initially and Unica more recently. The services business is now over 160 people focused on 1:1 or precision marketing (churn prediction, upsell/cross-sell) and credit risk. The credit risk practice is increasingly in the context of Basel II and this has moved them from doing just the analytic modeling to delivering complete credit risk solutions.

In this professional services work G Stat found that the time to implement data mining and predictive analytic models was too high. Because the barrier to getting models implemented was too high, many clients failed to get these models into production and this led to the development of the G-Soft Xeligence platform. Each Xeligence solution contains a data schema, out of the box processes for data management and data mining, packaged models and rules, a user interface for non-technical users and reporting/dashboards.

Their most mature product is Xeligence NBO – Next Best Offer – that sits between the data warehouse and the CRM or Campaign Management system. Xeligence is designed to produce models in an automated process – taking in data, running analyses, building models and deploying them into the campaign environment. One customer for example had 5 models that were re-calibrated every few years before working with G-Stat. Now they have 70-80 models re-developed and deployed each month with just a couple of work days to supervise and direct. These models are used in Unica to drive 1:1 marketing decisions. They have about 10-12 customers using this product and feeding the results into their campaign management environments and have a global agreement with IBM and various other global as well as regional partners  to supply this model creation and tuning environment alongside Unica and other campaign management tools.

Based on their success in Israel they are expanding – mostly into Europe but increasingly also into US. They have a newer product focused on churn predictions (in use with 2 customers) and have just finished a segmentation and LTV calculation and credit scoring engines.

All the Xeligence products implement the whole process of building, tuning and deploying models in very specific domains. They combine business domain experience with an understanding of how to build and deploy models and business process management to automate the analytical processes in an organization. This combination of business domain expertise with the automation of the analytic processes allows a business team to effectively manage the analytic process.

As they expand into credit risk obviously this complete automation becomes more complex as these credit risk decisions are regulated. For this they have developed an “analytical ETL tool” targeted to risk managers in retail banks. This product loads data into an in-memory column store so users can do all the transformations necessary for credit risk modeling. This exposes the rate of change transformations for instance, critical to credit risk modeling, as simple functions that can be easily used to build models. This eliminates the flattening and transformation of data common to manual data mining processes. The process of reviewing and creating models is thus shortened. The credit risk platform has integrated deployment also, eliminating the time and cost of deployment. Finally the models are also documented as they are created.