What trends in data warehousing should a CIO watch out for?
The most important trend is in the area of what we call active enterprise intelligence. A traditional data warehouse focuses only on reporting, analysis and data mining. However, enterprises are using data warehousing not only for strategic decision-making but also operational decision-making. They have started using data warehousing not only for strategy formulation but also for strategy execution. But how do they get the data at the front lines of the organization?
People are interacting with customers, so a personal banker or a call center representative needs to have access to the right information to make the right decision. Thus, you start thinking of not only the big decisions but also the small decisions that need to be taken on a daily basis. This is the idea of active enterprise intelligence where we support both strategic and operational intelligence.
Another trend is what I would call decision services. Organizations have started to work with SOA for Online Predictive Processing (OLPP), but they really have not, in most cases, deployed SOA for decisioning applications. Think in terms of the service provider not just providing transactional services but also decisioning services, risk scoring, and inventory analysis in real time. What you are going to see is decisioning services with synchronous decision services with integration using service brokers to co-ordinate the cooperation and Web services between the transactional and decision services.

In April you launched multi-temperature data warehousing consulting. What is the significance of this?
Multi-temperature data management (MTDM) is an important trend because a huge customer base will lead to large volumes of data. The economics of storing large volumes of data may or may not be attractive. We need to find ways to store more data but at a lower price. The idea of MTDM is to distinguish hot data from cold data. Hot data is data that is accessed very frequently. Cold data, on the other hand, is not accessed as frequently and the performance requirements are not as aggressive.
With MTDM what you want is a completely automated system for managing different temperatures of the data. I can then optimize the price performance of that data. If I can identify certain data as cold and I have advanced optimization techniques for eliminating access to that data for a majority of queries, then I can store the data on fewer disks of higher density to lower my price per gigabyte. But the database has to be very sophisticated to manage the hot versus the cold data because if you access all the data with equal frequency then there would be a performance problem.
Our Teradata Active Systems Management uses priority scheduling and work load management to manage and prioritize the I/Os for different types of data with different performance service levels. We can move the data around so that the hot data has higher performance storage media and the cold data has lower performance storage media so that they optimize the balance of the data across all the storage devices. MTDM has been deployed for large telcos in India as they have a lot of data.

How can Teradata’s Event Based Marketing (EBM) solutions help enterprises?
In your first question on trends I touched upon the trend of decision services. While SOA deployments represent synchronous decision services, EBM is an example of asynchronous decision services. EBM is more than just marketing; it is event-based analytics.
If you look at traditional database marketing, the implementations are focused on how to make junk mail more efficient. This is not really a good way to build relationships. With EBM you can look at and recognize which events are important to a customer relationship and then focus on those events with appropriate responses, offers or value propositions to build relationships. In this way, with less communication, enterprises can derive higher value and change the economics of the customer management game by being more relevant and cost effective.

How does it actually work?
Let us take the example of a bank which sends out mass mails to people to invest in a college savings plan for their children. This would get a very low response rate. With EBM there can be a plan that sets up an event to detect large deposits made into bank accounts. Now large deposits can be customer specific and would vary from one customer to another. You therefore monitor large deposits on a customer-specific basis and then based on the customer assets decide how they can invest in a better way. Thus, the bank gives a message that is relevant to the customers.
Consider this: the way people spend their salary bonus is different from the way they spend a tax refund. Bonus is something that they expect at the end of the year. During this time of the year, if you target them to save money by investing in their children’s education, it will get a high conversion rate. But a tax refund, which is free money, would just be spent on anything but savings as it was not expected from the government. In this situation, offering a new car loan with a certain amount of down payment would make more sense.
What this means is that the offer varies not only with the size of the deposit but also the source of the deposit. The key aspect is that marketing can be done by keeping a focused view of the transaction; you won’t require additional data like customer demographics.

Is there a specific model for EBM or would it vary as per the needs and priorities of customers?
With EBM there are two kinds of models for deployment. One is called the push model. In this, every time a new transaction comes in, it is evaluated, and depending on its relevance it is filtered on certain criteria and action is taken; it is more reactive.
The other is the pull model. Rather than filtering each transaction that comes in, users can be more efficient by batching up the software event detectors and setting up criteria that display, say, all large transactions in the last one hour. Users can perform mini-batch processing of the events that can be effective, especially if there is a large volume of data.
The use of a push or pull model therefore depends on the value of the data and deciding how to get it quickly. Most businesses will probably use the combination of the two, and it would be different for different applications.

How does EBM fit into Teradata’s product offerings?
The core of the Teradata solution is a scalable relational database. You place the transaction-level details in the database. This database provided by us also has a data model which is a blueprint that directs how to put the data in the database. This is a different blueprint or data model, and depends on the industry. Every industry has its own version that is unique to it. The data model characterizes which events are important for that industry. The blueprints tell us which events we need to care about. We also have software event detectors that allow the setting of rules like detecting large deposits or detecting when customers change their buying patterns. Rules can be set specific to organizations.
We provide the data model which is nothing but the blueprint to prioritize important events and organize them in the database; the database itself is for acquiring and storing the data as well as the tools for putting the data into the database. The software event detectors will have the business rules and highlight which events are important and what needs to be done with them. We also provide a library of events or best practices to start with which can then be modified as per customer need.

What kind of RoI can enterprises expect from EBM?
We have found that the response rates or ROI is at least 300-500 percent higher for EBM solution than traditional database marketing or so called target-based marketing. The bottom-line—it is big and at a lower cost.

Which sectors would be the initial adopters of EBM?
In India it would be the financial and telecom sectors. In some markets it would also involve retail, but not in India, because this sector is still in an evolutionary phase here. Another area that is still in a development phase is the airline industry. In the US the airline industry has been an adopter of this technology, but in India it is not happening as of now.