First Look: KXEN’s Cloud-Based Predictive Offers

March 20, 2013

in Analytics, Data Mining, Decision Management, Product News

Share

I last got an update from KXEN when they launched InfiniteInsight Genius. Since then they have been rolling out cloud-based analytic applications built around their core modeling engine. In particular they have launched a new product, KXEN’s Predictive Offers, their second cloud-based predictive analytic application.

KXEN has historically been focused on B2C companies, especially large ones, using predictive analytics to improve the customer lifecycle – telecoms, financial services, retail and e-business. They have delivered both their core platform and a series of predictive applications to over 500 customers worldwide. KXEN’s automated modeling technology supports data preparation, model building, deployment and scheduled refresh.

KXEN has begun building applications in the cloud and on-premise using this platform. The on-premise applications are generally embedded in other companies’ complete solutions, with the third party selling a solution that embeds KXEN’s engine. The cloud apps are part of widening KXEN’s footprint to smaller companies and are being sold directly by KXEN through app stores like Salesforce.com’s AppExchange). KXEN’s cloud apps are designed to be configured and deployed by a cloud app admin (e.g. a salesforce.com administrator) or a marketer and consumed by the end user (e.g. call center agent, sales rep, etc.) without any data science know-how or analytical training.

The market opportunity for these kinds of applications exists because cloud or SaaS CRM vendors have generally weaker analytics than their on-premise competitors. Meanwhile more companies, and smaller companies, are focusing on using analytics to deliver better customer relationships. Without an analytic story these SaaS offerings are going to be at a serious disadvantage relative to their on-premise competitors when competing for larger deals like B2C call center deployments where analytics really matter. KXEN plans to enable these SaaS CRM applications with powerful predictive analytics.

The basic infrastructure involves KXEN’s engine running in the EC2 cloud. Data is assembled natively in the SaaS CRM application (initially just Salesforce.com but in practice any SaaS CRM solutions could be integrated) to create the historical dataset needed to build the model. The dataset is then sent to the KXEN cloud which asynchronously builds the predictive model (typically in an hour or two for even salesforce.com’s largest customers) using KXEN’s automated model building platform, and is then pushed back as native code (e.g. Salesforce.com’s Apex) that executes inside the SaaS CRM. This means that the CRM application can score a customer (or an offer, or a lead) in real-time without accessing the KXEN server during a transaction or interaction.

KXEN’s cloud applications are, for the moment, very focused on Salesforce.com and tie to the new Salesforce.com mantra about being a “customer company.” Becoming a customer company means moving from a transactional, inside-out, company-centric approach to CRM (legacy) to one that is more customer-centric, outside-in, and relationship oriented. This requires ending the use of generalized messaging that dominates companies’ interaction with their customers and moving to one focused on personalized, targeted messaging. Customer companies need to adopt a next best activity mindset (something I wrote a white paper about recently) and this drives a need for analytic decision-making in customer treatment.

To adopt a real-time approach to next best activity companies have historically had to build a complex system that required a lot of IT maintenance. KXEN’s cloud apps are designed to let Salesforce.com admins or a marketer install the app, configure the application (set up their offers and any rules associated with them for instance like inventory, eligibility, marketing priority, etc.) and then allow the KXEN predictive engine to do the rest. It will learn what works and build increasingly accurate predictive models. These applications are thus designed to work without the need for an analytic data scientist, leveraging the automation at the core of the KXEN platform.

Today KXEN offers Predictive Lead Scoring (for lead targeting), Predictive Offers (for next best activity). Soon it plans to add Predictive Retention (to make retention offers integrated with Predictive Offers) and Predictive Case Routing (for resolving service issues).

The new Predictive Offers application is integrated into the salesforce.com UI and displays the selected offer alongside the main forms for an end user. It presents the offer recommendation in either banner mode or console mode (a Service Cloud feature for high-volume call center agents)and a script to be displayed for the selected offer. The user can see both why a particular offer is being presented (or not presented) for the current user and the overall model driving each offer’s score. This scoring is done in real-time and natively inside Salesforce.com, so that changes in a customer’s information during a call for instance result in an immediate update of the predictive scores and thus the best offer. These offers can also be delivered into self-service or other environments connected to Salesforce.com.

Offers are defined in an administrative UI built for the salesforce.com admin or a marketer that allows the definition of the offer, association of creative and script, and the definition of rules. No analytical knowledge is required, making the applications easily consumable by the business. With a new marketing offer there is no data so the application will automatically try the offer randomly (in a “learning” mode subject to a defined maximum percentage for learning transactions). This enables it to gather data about what works and what does not. Once enough data is gathered to build a model the offer moves from the control group to the “predicted” offers. In a similar way the administrator can assign a certain number of transactions to a random selection from the predicted offers for “improvement mode”, allowing them to be tried when they would not normally come up as the best offer. This prevents pigeonholing or local maxima. Both of these are clear best practices in predictive analytics and it’s great to see them automated in the application.

KXEN is one of the vendors in our Decision Management Systems Platform Technologies Report, and you can get more information on their products by visiting kxen.com.

Share

Previous post:

Next post: