Intelligent systems are aware of the business environment and able to react to it, says Zubin Dowlaty. Smart sensors, check-ins, card swipes and big data of all types contribute to this fantastic yet realistic vision of streaming data that is processed immediately and used to automate business tasks and processes.
Dowlaty, Vice President of Innovation and Development of Mu Sigma and a speaker at the upcoming Analytics 2012 Conference, has implemented innovative analytics technologies and statistical techniques into the Mu Sigma ecosystem, as well as at many Fortune 500 organizations. He will be speaking on "Big Data Analytical Trends, Methods and Opportunities” at the Analytics Conference, October 8-9 in Las Vegas, so if you’d like to learn more, we hope to see you there. But, before the conference, Zubin was gracious enough to provide me with this interview about intelligent systems for analytics ROI. (Content was also contributed by SMS Chauhan who works alongside Zubin at Mu Sigma.)
Michele: Tell us a little bit about intelligent systems.
Zubin: The evolution of intelligent systems has been one of the most remarkable trends in the industry today. Such systems are based on a framework that demonstates various aspects of intelligence. Intelligent systems are aware of the various states of the environment, the changes that occur in these states and other information important for their existence. Not only can they process all relevant information for meaningful interpretation of the environment, these systems can also learn and adapt to such changes by reconfiguring themselves. These systems remember the decisions taken, both favorable and unfavorable, and apply them in case a similar pattern of change appears in the environment.
If you think about such a framework being applied to business from a technology perspective, the chances are that you will immediately start connecting the dots.
What would an intelligent system look like?
Well, for starters, an intelligent system should be aware. It should be able to “see” the business environment. Today, we have smart sensors that perform these functions and much more. Streaming transactional data, such as, customer check-ins’, card swipes, live social feeds and news. This information is being generated across the vastness of humanity.
Wait, but how does the data flow from one place to another?
Such a system should have access to all data as quickly as possible to be agile and responsive. In terms of implementation, there should be channels for the “data flow.” A direct metaphor in the technology today is an Enterprise Service Bus (ESB). An ESB uses a messaging framework to communicate between disparate systems in the enterprise. You can use this ESB metaphor to form an “event stream” that connects disparate systems.
How are we going to process this flow of data?
Here comes the scalable analytics engine into play, now powered by High Performance Computing (HPC) components and with many deployment choices. Different kinds of smart algorithms run on this layer, which are easy to deploy and manage. Additional information, if required, can be provided in the form of business rules. The rules and algorithms together can be used to automate some of the basic decisions making thus operationalizing your analytics processes.
Where are we going to store this data?
True, but the Enterprise Data Warehouse (EDW) today is moving away from centralized models, towards federated or/and cloud-oriented architectures. Hadoop and NO-SQL are the buzz of town. Scale, scale, scale! Store all you want.
How can I interact with such a system?
Nothing is good without the good old dashboards, which have evolved into cockpits over time. Turn a knob here, steer a little to the left and there you are. Play with data, visualize a three dimensional graph spread across space, interact with it and tweak configuration to render the same data in a completely different visual! Show it to everyone, and you have five more friends who share the same taste in analytics as you.
This is all very cool but why do we need to use intelligent systems in business?
The value in a business lies in how responsive it is - how quickly can the business react? Intelligent systems are designed to help your business become more agile. You can deploy more sensors to capture more data, deploy an ESB to transmit data in real time, scale the analytics layer to process data faster and use rich interfaces to make decisions on the fly. Intelligent systems aim to reduce all the three latencies – capture, analysis and decision (as shown in the Figure below). Lesser is the latency, higher is the value and hence the ROI.
Sounds good in theory, can you show us an example at work?
Let’s look at an online retailer’s website. The website generates a bunch of user data – clicks, items they select, items they purchase, items they abandon, reviews and feedback they provide. Typically, this data is stored in a data warehouse in the form of huge weblogs. There are generally multiple steps involved: log the data to a temporary database and move it later by firing an ETL script to the EDW. Later pull the data from the EDW for analysis, process and generate the daily results. By the time the results are generated for the website to get back to the customer, the users has either made up his mind to purchase elsewhere or decided against purchasing. Either ways, the opportunity is lost.
This is exactly where an intelligent system is extremely effective. As data is being generated, it flows through the ESB into an analytics engine. Dynamic models process this data to generate user scores and classifications in real-time (milliseconds). Pre-defined rules pick up the scores and classification, generate offers and recommendations, provide eligible discount, enable cross-selling and result in instant conversion. It just doesn’t stop there; these systems also optimize the supply-chain and deliver goods in minimal time. Overall, you can sell more, have more happy customers and generate a significantly higher ROI.
