I had a chance to talk with Rob Walker last week about Chordiant and their decision management platform. Chordiant focuses on large customers – those with lots of decisions in markets such as retail banking, consumer lending, card services, insurance and communications. Their mantra is Customer Experience Management and they aim to deliver an improvement in the quality of customer interaction. Primarily they are focused on what I have called growth decisions – cross-sell, up-sell, retention etc – but they have a growing business in risk-centric decisions and risk management. The company describes itself as having been process-centric in the past and having become decision-centric. It sees its future as becoming “learning centric”.
The decisioning capabilities of their product, which were the focus of our discussions, have a number of interesting characteristics:
- All development and management is usable by business users
Because of the critical role of the business in defining how decisions must be made, it is well established that decision automation must fully engage non-technical users. Chordiant take this seriously and work to ensure that all features can be configured and managed by business users. This helps ensure complete “business virtualization” to use their phrase.
- Channel agnostic but focused on call center, web and outbound
The engine ensures that decisions can be deployed to any channel and any platform and most customers push decisions to non-Chordiant applications. The company’s focus though is on these primary customer interaction channels.
- Next Best Action
The unifying theme is one of Next Best Action (or Best Next Action as it is sometimes called). The platform uses all the available information, rules and analytic models to make the most useful and profitable decision at any moment in any channel and constantly revises this as new data and responses are gathered.
- All pieces built from scratch as part of an integrated whole
This means that there is one repository for decisions (rules and models), one set of reporting and metrics (against decisions) and that everything flows through. This was a very impressive aspect of the platform – total decision centricity.
The core of the approach is the deployment of a Decision Hub (we called this a Decision Services Hub in the book). This centralized component is driven by predictive models and rules combined as decision logic into operational decisions. Real time data is fed in and a wrapper of what they call “conversation management” is provided so that decisions can be integrated into customer conversations to support an interactive dialog in the various channels. A number of pieces come together:
- The Predictive Analytics Director lets you build Behavioral Models
This tool allows non-statisticians to build customer treatment focused models such as churn models, propensity to buy models etc. It is very template and wizard driven and very focused on its particular subset of all the possible kinds of predictive analytics. The tool focuses on analytic techniques with high predictive power and high transparency. It is also possible to import PMML definitions of existing models for use in decisions. They support a variety of adaptive or self-tuning models also.
- The Strategy Director combines rules and models
This tool allows business users to develop their Next Best Action strategies by combined rules and rulesets with executable models. It supports model trade-off matrices, decision tables and decision trees as well as if-then-else logic. The decision is laid out on a canvas where you can add models, define champion/challenger or adaptive control rules, add rules and rulesets etc. The deployment seems to “backward chain” from the decision it needs to make an offer to see which rules and models it must execute to come up with an answer.
- Deployment to Channels
Deployment capabilities allow these decisions to be deployed as executable services so that they can be integrated with a variety of systems and channels. They have a full repository set up that manages rules and models together. They also have many deployment options from stand-alone popups to embedded in their own or other desktops, personalized web etc
The results and consequences of decisions are tracked and stored and ultimately monitored and reported on. This completes the feedback loop to the modeling process.
Overall it looks like a very powerful platform and one that delivers all the components you need – rules, analytics, adaptive control, decision services etc. I liked the “relentlessly top-down” mindset for focusing on customer-centric decisions and the platform has clearly been designed with this mindset. Chordiant also supply an out of the box Next Best Action blueprint to help accelerate adoption.