Oracle Customer Data Hubs Chief Defends CDH Model

By eweek  |  Posted 2005-03-11

Oracle Customer Data Hubs Chief Defends CDH Model

How much did you sell yesterday? How much was in inventory? If it wasnt in inventory, how much will it cost to build? What did you spend yesterday? Who is your best supplier of PCs? Whos got the lowest cost? Who can deliver most quickly?

Oracle Corp. wants to hook you up with all the answers via its most recent technology vision, based around data hubs that feed into one huge, centralized database. The Redwood Shores, Calif., company announced the first of this hub series, the Customer Data Hub, 14 months ago.

Its an ambitious vision, encompassing everything from stopping terrorists at the border to hooking pancake restaurant customers up with guaranteed response times to complaints.

But at least one critic, The CDI (customer data integration) Institutes Aaron Zornes, says that Oracles hub vision is mostly fluff.

Whatever technology is out there, Zornes says, is mostly built on precursor technology such as Oracles data engine—known as TCA (Trading Community Architecture)—or OCO (Oracle Customers Online).

Oracles CDH hasnt been proved scalable, Zornes contends, nor is much of anybody using it except Type A shops that can build it themselves. Moreover, with no formal training until recently, it doesnt even qualify as a product, Zornes maintains.

To read more about Aaron Zornes doubts about Oracles customer data hubs, click here for the column by Lisa Vaas.

Ouch, ouch and double-ouch. Database Editor Lisa Vaas recently got Oracle Senior Director of Customer Data Hubs Peter Heller on the line to dig a little deeper into where exactly Oracles hub technology is at, as well as to answer some of Zornes charges.

One of the complaints about Oracles approach to CDI is its too broad: that Oracle is missing the crucial vertical industry insight of a product such as the financial industry has with DWL Customer, for example. Whats your response?

The way we model customers allows us to model for what we call business-to-business interactions. It allows us to do business-to-consumer, business-to-employee, and business-to-business-to-consumer. The fact that were structured in this robust way means the data model applies to many industries. Customers in 17 industries are now using this Customer Data hub.

When we speak to customers about solving customer data problems, everybody has questions about fields: "Do you have this field, that field?" Its an easy answer: You can always add another field.

We have the metadata on top: It doesnt require programmers to add a field to the data model. You can just add a field through the mechanism of the system. Whats important is the way customers are related to each other, the way they form these structures—b2c, etc—thats the heart and soul of it. Thats the hard part, the part that people value.

Is anybody actually using this stuff out of the box, though? Or are they all building it themselves from TCA and OCO, as Zornes suggests?

[Zornes] misunderstands a few things. Weve tried to make it clear. What were getting at here, the heart and soul of this, the oldest element of this product thats been in place is TCA, the data model.

Its the richness of that data model thats the value statement, the most valuable part of the architecture. On top of the data model and subsequent to our early release of it, we released [OCO]. It was a viewer that looks at all the customers in the data model. Its very cleverly laid out. Its very cool.

We consider it a product. Its on the price list, and were selling it. You just dont do that willy-nilly. The stepback is the make-or-buy conversation. There are about 40 vendors that make bits or pieces: integration, data quality, enrichment, synchronization or data models, themselves.

What Oracles offering is more of these pieces pre-built, pre-integrated and supported by a major vendor, than anybody else. When it comes to product features vs. integration features, for product features, what theyre implementing is the data model, DQM [Data Quality Manager], OCO and APIs to the data model.

Thats the package. Those are the things that make up the heart and soul of this product. IHOP [Corp.] will tell you thats what they have, and Church Pension Group and other customers.

The integration portion is always going to be custom. Thats a level playing field for all vendors across the board. I dont have that as a fact, but when it comes to hooking systems up, thats a custom effort. Its the event model, in the system itself, when changes are made and detected, that is the product feature.

We have 53 inbound APIs. Thats part of the product. That doesnt mean everybody uses 53. They might use five, or one. But the product is considered to be an open platform. You can access it, send data to it through a standards-based infrastructure. Theres no dependency on Oracle Applications Server. Many customers implemented choices of middleware—they had made prior choices, and we considered that a good thing.

Theres always an element of customization. It takes a couple of years to implement this kind of thing. The fact that we released initial components in May of 2000 … the fact that we had early customers doesnt eliminate the fact that … weve enhanced the whole offering for three to four releases.

Not everybodys going to use all the features. Thats nothing false about what were saying just because theyre not using all the features. Customers are on different product paths.

Next Page: Does CDH work best in all-Oracle shops?

CDH in All

-Oracle Shops">

Zornes says that CDH works best in all-Oracle shops, which kind of negates the core idea of having these hubs tie together disparate systems, doesnt it?

[Zornes] said that if you were an Oracle customer, theres a propensity to choose our solution. We said, "Well, so?" Thats OK, theres nothing wrong with that.

This is part of my view on what he said: What were talking about is these mixed environments. The idea is that companies have customer information fragmented around the company in multiple systems.

If one of those systems happens to be an Oracle system, we get to bid on the opportunity to be that customer master, the central master.

He somehow was discounting the fact that because we have a customer master in our application system that somehow were not a player in the [CDI] market. Our view is that Oracle is positioned beautifully for the CDI marketplace.

We have two sides to our business: our applications side, selling ERP [enterprise resource planning] and CRM [customer relationship management] systems, which requires a customer master, and since most companies have fragmented ERP systems, you have two choices: [one is to] unify around E-Business Suite.

If youre not inclined to do that, you dont have to unify around our suite, but you can unify around our data model. You dont have to implement it in the context of E-Business Suite. You can certainly implement Oracle CDH independent of [Oracle] transaction systems.

Second, with all our technology customers, one of Oracles core messages has always been about IT consolidation and simplification. Thats what people come to Oracle to do: They have 50 databases, and they want to collapse that down to one or two.

Those companies who are trying to save money on pure IT spending find this a very interesting subject. It gives them the opportunity to retire older systems, possibly. If you can have a project that protects customer data, you have the opportunity to, quote, do some consolidation.

Whats the difference between Oracles data hub and its data warehouse technology?

A data warehouse is passive and historical. A data hub is active and current. In the data hub, you actually update data. Youre creating data. We want people to update data records, change relationships, update info. Its something you dont do in a data warehouse. Its a downstream system.

Another important, important fact is that a data warehouse is unidirectional. All source systems send information to this specific place. You might clean up data as you try to normalize it and get rollups as it goes into the warehouse, but what have you done to correct data in its source? Nothing.

Data warehouse users, they make their best efforts to clean up data, but the data hub is trying to correct data at its source. When your call center agent is sitting in front of a screen and you call in and say youre Lisa Vaas, they dont have 50 records to choose from. Thats the net impact of not having a bidirectional process. Thats the distinction: unidirectional as opposed to data hubs being continuously synchronized.

How does it compare with what NCR Teradata is offering in terms of real-time updating?

What I hear about most is they are trying to quote operationalize the data warehouse, trying to provide access to the common user as opposed to an analytical group. Thats the crime in all [business intelligence]: Youre trying to get it into a community where everybody can use it, but theres no money left over after building for the data analysis group.

Thats the coexistence. These dont preclude one another. I emphasize how the data hubs and data warehouses need to coexist. Its the data hubs trying to clean up data prior to getting the data into data warehouses. With this mechanism, you get faster analytics. You dont have to go through the cleaning step as youre getting data into the warehouse.

Next Page: Oracles data hub is said to lack analytics: Discuss.

Oracles Data Hub Lacks


Its been said that Oracles data hub does not offer analytics. Whats up with that? What were trying to do, our No. 1 purpose, is to create a data quality program and process. Before you can do any analytics, anything meaningful, you have to have a single view of the customer. With 50 views of the customer, you cant add them up and know whats going on with that customer.

The natural assumption is that the reason youre doing this is to attain better analytics. But my research and efforts have shown these are two distinct initiatives, sometimes owned by two different departmental entities: those who care about integrity, privacy, security and taxonomy, and those who care about data quality.

Those groups need systems to support their efforts. Then when that data is in a good spot, youre in a position to find true customer value, and you can layer on top of analytics systems.

Its the chicken & egg scenario: What justifies what? It might be hard to justify data quality if you dont already believe in it. A way to justify it is by showing you can deliver better data relevance.

What about the scalability of CDH? Were talking about whether this solution can help enterprises that have millions of customers.

Dell has 200 million customers. They are our high-water mark. They require 2 million updates per day. We went through a benchmark to facilitate their environment, and we shared that with [Zornes]. Then weve got Cisco, which has 4 million customers. Weve got telco companies, also in the millions. Also small-end customers.

To read the latest in Oracles move to buy Retek, click here.

There are two solutions were solving: Some have smaller counts of customers but more data sources. Some have over 200 data sources. Its not surprising … for any company of any size, you ask how many customer lists they have. What ends up happening is a lot of stuff in systems, … somebody got involved in a special program for six months. They began accumulating data, but then the system falls out of favor or is no longer relevant to the business.

Check out eWEEK.coms for the latest database news, reviews and analysis.

Rocket Fuel