IN THE UNITED STATES PATENT AND TRADEMARK OFFICE
`
`In re Patent of: Hanaman et al.
`U.S. Patent No.: 6,963,826
`Issue Date:
`Nov. 8, 2005
`Appl. Serial No.: 10/668,476
`Filing Date:
`Sept. 22, 2013
`Title:
`PERFORMANCE OPTIMIZER SYSTEM AND METHOD
`
`
`
` Attorney Docket No.: 30651-0047IP1
`
`DECLARATION OF SCOTT SMITH
`
`1.
`
`My name is Ronald Scott Smith of comScore, Inc., 11950 Democracy Drive,
`
`Suite 600, Reston, VA Sunnyvale, CA. I have been asked to offer technical opinions with
`
`respect to prior art references cited in this Inter Partes Review (“IPR”). I base these opinions on
`
`my work regarding database management systems and data warehousing. My current curriculum
`
`vita is attached.
`
`2.
`
`I earned my B.S. degree (1987) in Economics from James Madison, in
`
`Harrisonburg, Va. I am a Sybase Certified DBA as well as a Sybase Certified Performance and
`
`Tuning Specialist.
`
`3.
`
`I have been working in database management systems and data warehousing for
`
`over twenty-five years. Specifically, during my career, I have held a variety of positions related
`
`to the development of database and information management systems.
`
`4.
`
`I currently hold the position of Vice President, Data Warehousing at comScore,
`
`Inc., where I am responsible for the Enterprise Data Warehouse (EDW) environment, including
`
`both operations and development activities and manage the development and implementation of
`
`new/enhanced data products in the EDW environment. In this position, I am also responsible for
`
`the SAP Sybase IQ 15.3 Multiplex Environment, utilizing 30 Dell r710/r810 servers (724 cores)
`
`and for the design and operations of the corporate SAN environments, including both the EMC
`
`Page 1 of 27
`
`COMSCORE 1003
`
`

`

`(VMAX/Clarion) and Violin Memory SAN environments. Since 2000, I have worked to
`
`architect an extensible and reliable EDW platform, which has grown to 150Tb over the past 10
`
`years.
`
`5.
`
`From 1996 to 2000, I worked as a Senior Consultant for Sybase Inc., which is a
`
`leading provider of enterprise software and services, including database technology. In this
`
`position, I developed a data migration process to extract, scrub/transform and load legacy data
`
`into mission critical systems. I also performed numerous development activities, including
`
`database design, SQL performance enhancements, stored procedure development, stress testing
`
`and provided guidance on business process re-engineering and implementation. I restructured an
`
`existing Data Mart to reduce data growth, and designed and implemented 3 Data Marts using
`
`PowerDesigner and Sybase IQ. I developed a custom framework for capacity analysis and
`
`performed the analysis for an existing data warehouse. In addition, I presented a technical
`
`discussion on Dimensional Data Modeling to district management and consultants.
`
`6.
`
`From 1994 to 1996, I worked as a Database Specialist for InfoPro Incorporated.
`
`In this capacity, I served as a Senior DBA of production Sybase SQL servers and created a
`
`custom migration process that included 3rd party data schemas and applications. I also designed
`
`and implemented a PowerBuilder application framework, which was used to deliver 4
`
`PowerBuilder applications. PowerBuilder is a development environment from Sybase Inc. for
`
`developing database applications.
`
`7.
`
`From 1990 to 1994, I worked as Software Developer at SAIC Corp. In this
`
`capacity I designed and coded Integrated Contracts Management System modules using
`
`PowerBuilder and designed and coded data propagation routines that pushed data from regional
`
`sites to a central reporting site using ORACLE database triggers and stored procedures.
`
`Page 2 of 27
`
`

`

`
`
`8.
`
`During 1987 to 1990, I worked in various positions as a Programmer Analyst or
`
`Systems Analyst. During this time, I created a data element dictionary and proposed data
`
`structures for a multi-platform FOCUS application and developed an in-house database
`
`application for tracking software and hardware.
`
`9.
`
`I am familiar with the content of U.S. Patent No. 6,963,826 (the “‘826 patent”).
`
`Additionally, I have reviewed the following: Kimball, R. and Merx, R., The Data Webhouse
`
`Toolkit -- Building Web-enabled Data Warehouse, New York, Wiley Computer Publishing, 2000
`
`(“Kimball”) and Karuna P. Joshi, Anupam Joshi, Yelena Yesha, and Raghu Krishnapuram. 1999.
`
`Warehousing and mining Web logs. In Proceedings of the 2nd international workshop on Web
`
`information and data management (WIDM '99), Cyrus Shahabi (Ed.). ACM, New York, NY,
`
`USA, 63-68 (“Joshi”). Counsel has informed me that I should consider these materials through
`
`the lens of one of ordinary skill in the art related to the ‘826 patent at the time of the invention. I
`
`believe one of ordinary skill as of Sept. 22, 2003 (the filing date of the ‘826 patent) would have
`
`had a bachelor’s degree in computer science from an accredited university, or equivalent work
`
`experience or training, and a knowledge of database design, computer programming, and
`
`networking technologies. I base this on my own personal experience, including my knowledge
`
`of colleagues and others at the time.
`
`10.
`
`I am employed by comScore, Inc., but my compensation is not based on the
`
`substance or outcome of my opinions.
`
`11. My findings, as explained below, are based on my education, experience, and
`
`background in the fields discussed above.
`
`12.
`
`This declaration is organized as follows:
`
`Page 3 of 27
`
`

`

`I.
`
`II.
`
`III.
`
`IV.
`
`Brief Overview of the ‘826 Patent (page 4)
`
`Kimball and Combinations Involving Kimball (page 7)
`
`Terminology (page 19)
`
`Conclusion (page 26)
`
`I.
`
`13.
`
`Brief Overview of the ‘826 Patent
`
`The ‘826 patent is directed to modeling and warehousing usage information and
`
`transactional information, as well as statistical analysis information derived from applying a
`
`statistical methodology to the usage information and transactional information. The ‘826 Patent,
`
`col. 4, lines 36-43. The ‘826 patent discusses performing data modeling on this data and storing
`
`the data in a data warehouse. One of skill in the art as of Sept. 22, 2003 would understand that a
`
`data warehouse generally includes one or more databases that hold an enterprise’s data and may
`
`do so in a way that is designed to aid in both complex analysis of data and decision support.
`
`Often, a data warehouse’s primary purpose is to store the enterprise’s data, and present the data
`
`in a way that allows members of the enterprise to make business decisions. For example, the
`
`data warehouse may store data from the enterprise’s marketing and sales system in a way that
`
`allows management to view reports related to that data, such as annual or quarterly sales
`
`comparisons or comparisons of marketing spend versus sales.
`
`14.
`
`The data to be warehoused is typically extracted from the source systems and
`
`processed in a staging area before it is loaded into the databases of the data warehouse. The
`
`processing in the staging area includes, for instance, cleaning the data (e.g., correcting
`
`misspellings, resolving conflicts in the data, handling missing data elements), deleting data not
`
`used in the data warehouse, and converting the data into a format acceptable for the data
`
`Page 4 of 27
`
`

`

`warehouse. Once the processing in the staging area is completed, the data is loaded into the data
`
`warehouse.
`
`15.
`
`End user applications can then access the data warehouse to retrieve and present
`
`data from the warehouse to a user. For example, an end user application can query the data
`
`warehouse for particular data, and then present data yielded from their query in a report, graph,
`
`or some higher form of analysis to the user. The end user application can be, for example, an ad-
`
`hoc query tool or a data mining or modeling application.
`
`16. When designing a data warehouse, data modeling is typically employed to create
`
`a data model for the data warehouse. The data model can define the structure of the data within
`
`the data warehouse by defining the entities, the data elements, their formats, and the relationships
`
`between them. When an information system uses one or more databases, for example, data
`
`modeling may entail defining the database structure to be used by the one or more databases to
`
`hold the data. For instance, when relational databases are used to implement the data warehouse,
`
`data modeling may entail defining the tables in the relational database used to store the data.
`
`17. The ‘826 patent’s usage and transactional information is derived from a sales
`
`representative’s use of a customer relationship management (CRM) or sales force automation
`
`(SFA) application. See, for example, The ‘826 Patent, col. 1, lines 10-11; col. 4, lines 36-57.
`
`
`
`18. A sales representative uses the CRM/SFA application available through the
`
`representative’s computer to manage, access, and collect transactional information, such as
`
`information related to sales contacts, sales calls, or fulfillment. The ‘826 Patent, col. 1, lines 59-
`
`63; col. 2, lines 24-26; col. 2, lines 45-50, col. 12, 61-62; col. 15, lines 25-26. In addition,
`
`Page 5 of 27
`
`

`

`software on a sales representative’s computer observes and records usage information about the
`
`representative’s use of the computer, including the representative’s use of the CRM/SFA
`
`application. The ‘826 Patent, col. 4, lines 58-61; col. 10, lines 25-35. For example, the software
`
`may monitor and track “the viewing of a particular display screen” or “the length of time a
`
`particular display screen has been viewed.” The ‘826 Patent, col. 4, lines 33-44.
`
`19. This information is uploaded, aggregated with other data, and stored in a data
`
`warehouse. The ‘826 Patent, col. 9, lines 33-46; col. 11, lines 58-63; col. 12, line 34 to col. 13,
`
`line 8. Statistical analysis is performed on the usage and transactional information stored in the
`
`data warehouse. The ‘826 Patent, col. 9, lines 46-55; col. 15, line 49 to col. 16, line 55. To
`
`perform the statistical analysis, data is “extracted from the data warehouse and provided to a
`
`statistical engine where statistical analysis routines are performed on the data.” The ‘826 Patent,
`
`col. 15, lines 54-56. The results of the analysis are “uploaded back into the data warehouse for
`
`presentation purposes.” The ‘826 Patent, col. 15, lines 56-58.
`
`20. The data warehouse employs a data model that resulted from data modeling on
`
`the usage information, the transactional information, and the results of the statistical analysis.
`
`The ‘826 Patent, col. 13, line 25 to col. 15, line 38; col. 17, lines 47-54. For example,
`
`dimensional modeling was performed on the usage information, the transactional information,
`
`and the results of the statistical analysis to create the data model shown in figures 6A-6I and 8A.
`
`The ‘826 Patent, col. 17, lines 47-54. Figures 6A-6I shows aspects of the data model related to
`
`the usage information and transactional information. The ‘826 Patent, col. 17, lines 47-54; figs.
`
`6A-6I. Figure 8A shows aspects of the data model related to the results of the statistical analysis.
`
`The ‘826 Patent, col. 17, lines 47-54; fig. 8A.
`
`21. The data stored in the data warehouse is employed to prepare reports and tables
`
`Page 6 of 27
`
`

`

`that provide “insights for management and upper management with respect to the effectiveness
`
`of CRM/SFA systems and related data sources.” The ‘826 Patent, col. 18, lines 4-23. Figures
`
`9B-9Q illustrate examples of reports or tables that are generated. The ‘826 Patent, col. 18, lines
`
`18-23.
`
`II.
`
`Kimball and Combinations Involving Kimball
`
`A.
`
`Kimball
`
`22.
`
`In general, Kimball describes a data warehouse that stores data representing
`
`actions taken by a user at a Web browser when interacting with an enterprise’s Website
`
`along with other data for the enterprise, such as sales data and customer communication
`
`data. Pages 33-38; 129-168 (describing clickstream data); pages 170-171 (describing
`
`sales data); pages 171-172 (describing customer communication data).
`
`23. More specifically, Kimball is directed to a data warehouse designed to store
`
`“clickstream” data along with other data for the enterprise, such as sales data and customer
`
`communication data. Pages 4-6, 33-38; 129-168 (describing clickstream data); pages 170-171
`
`(describing sales data); pages 171-172 (describing customer communication data), and 385. The
`
`clickstream data generally refers to the “composite body of actions taken by a user at a Web
`
`browser” when interacting with an enterprise’s Website and “can include both the actual clicks
`
`(browser requests) and the server responses to those requests.” Page 359. In one case, the
`
`“clickstream exists tangibly in the form of Web server logs,” but the clickstream data may also
`
`exist tangibly and be collected in other ways, as described further below. Page 359. Kimball
`
`refers to the data warehouse designed to store clickstream data as a “Webhouse.” Page 385.
`
`24. Data mining techniques, which employ statistical methods, are applied to
`
`the data in the Webhouse to generate information about meaningful patterns in the data,
`
`which is then stored back into the data warehouse. Pages 15, 33-38; 251-267, 346-347.
`
`Page 7 of 27
`
`

`

`Data modeling, in particular dimensional modeling, is used to determine the data
`
`structures used that store all of this information in the data warehouse. Pages 129-185.
`
`The results of the data mining, as well as other information from the Webhouse, is
`
`presented to decision makers to help them make meaningful business decisions. Pages
`
`14-15 (describing presenting warehouse data), 31 (describing delivering data mining
`
`results), 33-38; 69-89 (describing decision making), 201-249 (describing warehouse
`
`interfaces).
`
`25.
`
`In particular, fig. 1.2 of Kimball (reproduced below) illustrates one example of “a
`
`commercial system supporting a public Web server and an associated data Webhouse.” Page 31.
`
`“At the top of the figure, we see the remote user, connected to the Web through an Internet
`
`service provider (ISP).” Page 33. The Web server is shown at the bottom left, along with an
`
`Application and Business Transaction server. The data Webhouse is shown on the bottom right
`
`in the gray box.
`
`
`
`Page 8 of 27
`
`

`

`
`
`26.
`
`The remote user employs a Web browser to send HTTP requests for web pages to
`
`the Web server. See Page 33; see also, pages 91-95. The web pages are user interfaces to be
`
`shown on the display device of the remote user’s system. In response to the HTTP requests, the
`
`Web server sends the requested web pages to the Web browser, which retrieves any additional
`
`resources needed to display the web pages, and uses the same to display the web pages. See
`
`Page 33; see also pages 91-95.
`
`Page 9 of 27
`
`

`

`27.
`
`Kimball describes several techniques for capturing information about the remote
`
`user’s interaction with the Web browser and the Web server. In the example shown in fig. 1.2,
`
`the Web server “log[s] client interactions into one or more log files.” Page 97. Specifically,
`
`“[e]ach time the Web server responds to an HTTP request, an entry is made in the Web server’s
`
`log file.” Page 99. The information logged can include, for instance, the URL of the webpage
`
`requested, the time it was requested, the IP address of the remote user’s device, the referrer (a
`
`URL of the referring server), and any cookie included with the request. Pages 98-112. The
`
`cookie may include a unique identifier for the Web browser used by the user. Pages 105-109,
`
`122-123.
`
`28.
`
`In another example, a null logging server may be used to monitor and capture
`
`information about the Webpages sent to the remote user from the Web server and displayed by
`
`the remote user’s Web browser. Pages 123-125. “The null logging server is a Web server whose
`
`primary mission is not to deliver content, but to accept log data.” Page 124. An <img> tag is
`
`embedded in the Webpages sent to the remote user. Page 124. When the Webpages are
`
`rendered in the remote user’s browser, the <img> tag causes a request to be sent to the null
`
`logging server. Page 124. In response, the null logging server sends a null image (e.g., a one-
`
`pixel transparent image) that does not affect the appearance of the web page. The null logging
`
`server also logs the request in the null logging server’s log. Page 124. The request includes a
`
`query string that contains data about the Webpage, and also includes a cookie if it was previously
`
`set on the remote user’s Web browser. Page 124. The query string and cookie are recorded in
`
`the log. Page 124.
`
`29.
`
`As another example, the Web browser on the remote user’s computer can monitor
`
`and capture information about the user’s interaction with the Website and send it to the Web
`
`Page 10 of 27
`
`

`

`server. See pages 345-346. For example, the Web browser may “measure dwell time accurately
`
`(e.g., excluding the time a browser window was obscured by other windows), and return XML-
`
`coded information to the Website.” Page 345.
`
`30.
`
`As yet another example, “sandboxed tracking applications” can be used [w]ithout
`
`any modification to current browsers” to capture “a great deal of feedback.” Page 346. One of
`
`skill in the art, as of Sept. 22, 2003, would understand this sandboxed tracking application to be a
`
`standalone tracking application that can be run on the remote user’s computer to monitor and
`
`capture information about the remote user’s interaction with the Web browser.
`
`31.
`
`The collected information may be used to derive further information about the
`
`remote user’s interaction with the Web browser and the Web server. See, for example, pages 23-
`
`28, 59-64, and 160-161. For example, the collected information may be used to derive the entry
`
`point into the website and the pages visited by the user at the website (and the order in which
`
`they were visited). Page 59-64; 160-161. The entry point of into the website and the pages
`
`visited by the user represents pages transmitted online by the Web server to the Web browser,
`
`which would display those web pages during its normal operation. See Page 33; see also, pages
`
`91-95. Accordingly, I believe one of skill in the art, as of Sept. 22, 2003, would understand that
`
`the entry point and pages visited both individually correspond to webpages transmitted online to,
`
`and viewed by, the remote user.
`
`32.
`
`The derived information may also include the dwell time of the user on a given
`
`webpage, which represents the length of time the webpage was viewed by the remote user. Page
`
`59-64; 160-161. The webpage for which dwell time is measured may correspond, for example,
`
`to a webpage used to process purchases and therefore accepts data entered by the remote user.
`
`Page 11 of 27
`
`

`

`See, for example, page 170 (describing that “the Web server receives a properly filled out page
`
`back from the customer containing a sale”).
`
`33.
`
`The derived information also can include, for example, the number of times the
`
`remote user visited the website, the change in weekly frequency of website access, the average
`
`number of pages the remote user visited per session, or the average dwell time of the remote user
`
`per session, among other information. Pages 263-264; 129-168.
`
`34.
`
`Kimball’s system also captures transactional information. Pages 11, 33-39, 169-
`
`172, 351-352. In particular, one of the “capabilities of the Web server is to take orders for
`
`products from the company, or to perform some other kinds of meaningful business
`
`transactions.” Page 33. The Web server does this in cooperation with the Application and
`
`Business Transaction server. See Page 33. “The job of the business transaction server is to
`
`record the business transactions.” Page 33. “When the Web server receives a properly filled out
`
`page back from the customer containing a sale, . . . the Web server invokes a transaction on the
`
`[Application and Business Transaction server].” Page 170. In addition to information about
`
`sales or orders, customer communications are captured. Pages 171-172. “These
`
`communications include mailings and telephone interactions such as sales calls, support calls,
`
`and inquiry calls.” Pages 171-172. The transactional information represents information about
`
`the sales, orders, or communications collected by the web server, transaction server, or other
`
`system. Page 33; 170-171. The collected information about the sales, orders, or
`
`communications is information related to conducting business or negotiations.
`
`35.
`
`Kimball’s system aggregates the clickstream information and the transaction
`
`information. For example, a clickstream postprocessor collects the clickstream and transaction
`
`information from the Application and Business Transaction server and the Web server and places
`
`Page 12 of 27
`
`

`

`the information in the data Webhouse. Pages 187-197. As illustrated in figure 8.1, there may be
`
`multiple Web servers and application servers from which the clickstream processor collects the
`
`information. Page 187; see also pages 24-25 (describing the collection of data by different
`
`physical servers; see also page 114 (describing the use of multiple Web servers to implement the
`
`Website). In addition, the clickstream processor may collect the data from sources other than the
`
`Web server. As described by Kimball, the clickstream information may be aggregated from
`
`referring partners, ISPs, or Web watcher services. Page 24-25. The collected information is
`
`extracted, transformed, and loaded into the data Webhouse. Pages 185-197; pages 36-37.
`
`36.
`
`Kimball describes performing data modeling on the collected clickstream and
`
`transactional information to determine the data structures that are used to store this information
`
`in the data Webhouse. Pages 129-185. In particular, Kimball describes using dimensional
`
`modeling to model this information. Pages 129-185. Dimensional modeling is a “methodology
`
`for modeling data that starts from a set of base measurement events and constructs a table called
`
`the fact table, generally with one record for each discrete measurement.” Page 364. A
`
`measured event is referred to as a fact. See, page 266 (definition of fact). “A fact may be the
`
`sale of a product at a retail cash register, the price of a stock at a point in time, the amount of
`
`insurance coverage entered into a new policy that is being created, the balance of an account, or
`
`the change in your salary as a result of your promotion.” Page 130. This fact table is then
`
`surrounded by a set of dimension tables, describing precisely what is known in the context of
`
`each measurement record. Page 364. Because of the characteristic structure of a dimensional
`
`model, it is often called a star schema. Page 364. Pages 129-168 describe specific dimensional
`
`models for the clickstream information. Pages 169-176 describe specific dimensional models for
`
`the transactional information.
`
`Page 13 of 27
`
`

`

`37.
`
`Kimball’s system applies data mining on the clickstream information and the
`
`transactional information stored in the data Webhouse to provide information about meaningful
`
`patterns in this information. Pages 15, 31, 34, 35, 38, 74, 251-267, 346-347, 362. The data
`
`Webhouse prepares and hands information, such as clickstream and transactional information, to
`
`the data mining tool, which performs data mining on the information and returns the results of
`
`the data mining back to the data Webhouse of storage. Pages 253-265; 347.
`
`38.
`
`The data mining may include, for instance, clustering, classifying, estimating or
`
`predicting. Page 253. Clustering may entail, for example, “looking through a large number of
`
`initially undifferentiated customers and trying to see if they fall into natural groupings.” Page
`
`253; see also page 74. “The input records to this clustering exercise ideally should be high-
`
`quality verbose descriptions of each customer with both demographic and behavioral indicators
`
`attached to each record.” Page 253. The data for these records is provided by the data
`
`Webhouse and may include usage information and transaction information, such as the following
`
`items in the list on pages 263-264:
`
`Date of First Purchase,
`
`Date of Last Purchase,
`
`Average Number of Purchases in Last Year,
`
`Change in Average Number of Purchases vs. Previous Year,
`
`Total Number of Purchases, Lifetime,
`
`Total Value of Purchases, Lifetime,
`
`Number of Times Visited Website,
`
`Change in Weekly Frequency of Website Access, Current Quarter To Previous,
`
`Average Number of Pages Visited Per Session,
`
`Page 14 of 27
`
`

`

`Average Dwell Time Per Session,
`
`Number of Web Product Orders,
`
`Value of Web Product Orders,
`
`Number of Website Visits to Partner Websites, Current Quarter,
`
`Change in Partner Website Visits, Current Quarter to Previous.
`
`
`
`Pages 251-252, 261-265, 347. “Specific tools that can be used for clustering include standard
`
`statistics, memory-based reasoning, neural networks, and decision trees.” Page 256. These tools
`
`all employ mathematical concepts.
`
`39.
`
`“Clustering and data mining techniques can be used to directly recommend
`
`marketing decisions.” Page 74. “Rather than simply clustering customers relative to revenue or
`
`profit, customers can be clustered according to their history, and hence their likelihood, of
`
`responding to certain kinds of promotions.” Page 74. “We use these techniques to decide how
`
`to cross-sell, upsell, and create promotions for each specific customer.” Page 74.
`
`40.
`
`Classifying entails associating a class with a particular piece of data. “An
`
`example of classifying is to examine a candidate customer (for instance) and to assign that
`
`customer to a predetermined cluster or classification.” Page 254. For instance, “customers [may
`
`be classified] as credit worthy or credit unworthy.” Page 254. To perform classification, for
`
`example, “a verbose description of the customer . . . is fed into the classification algorithm.”
`
`Page 254. This description is provided by the data Webhouse and may include usage
`
`information and transaction information, such as the items listed above. Pages 251-252, 261-
`
`265, 347. Based on this information, the “classifier determines which cluster centroid the
`
`candidate customer or patient is nearest to or is most similar to.” Page 254. “Specific tools that
`
`Page 15 of 27
`
`

`

`can be used for classifying include standard statistics, memory-based reasoning, genetic
`
`algorithms, link analysis, decision trees, and neural networks.” Page 254. These tools all
`
`employ mathematical concepts.
`
`41.
`
`Estimating and predicting entail trying to estimate or predict some value based on
`
`historical data. “For example, we may find a set of existing customers that have the same profile
`
`as a candidate customer.” Page 254. “From the set of existing customers we may estimate the
`
`overall indebtedness of the candidate customer.” Page 254. Estimation and prediction can also
`
`drive classification. Page 254. For instance, we may decide that all customers with more than
`
`$100,000 of indebtedness are to be classified as poor credit risks. Page 254. The data for
`
`estimating or predicting is provided by the data Webhouse and may include usage information
`
`and transaction information, such as the items listed above. Pages 251-252, 261-265, 347.
`
`“Specific tools that can be used for estimating and predicting include standard statistics and
`
`neural networks for numerical variables, and all of the techniques described for classifying when
`
`only predicting a discrete outcome.” Page 254. These tools all employ mathematical concepts.
`
`42.
`
`Kimball describes performing data modeling on the results of the data mining to
`
`determine the data structures that are used to store the this information in the data Webhouse. In
`
`particular, as described by Kimball, the data mining tool hands “off the results of the data mining
`
`tool runs to the Webhouse for storage.” Page 262. This may be in the form of “a database to be
`
`stored in the Webhouse” and the database may include “time and customer dimensions.” Page
`
`346. I believe one of skill in the art, as of Sept. 22, 2003, would understand this discussion of a
`
`database with time and customer dimensions, that is handed back to the Webhouse for storage, as
`
`describing that data modeling was performed on the results of the data mining so as to create the
`
`data structures that hold the results of the data mining.
`
`Page 16 of 27
`
`

`

`43.
`
`In addition, pages 129-168 explicitly describe a data model that includes a
`
`customer dimension that has attributes for storing the results of a data mining operation. In
`
`particular, the customer dimension may include a cluster attribute, which describes the
`
`customer’s overall demographic cluster, and a credit profile attribute, which describes the credit
`
`profile of the customer (for example, a poor credit risk). Pages 143-148. As described above, a
`
`customer cluster and credit profile may be the output of a data mining process.
`
`44.
`
`Data from the Webhouse can be presented, for example, in a Web browser on a
`
`user’s screen. As described by Kimball, the data Webhouse “is a Web-enabled data warehouse
`
`devoted to publishing the company's data assets appropriately.” Page 37. To that end, the data
`
`Webhouse “deliver[s] a mixture of query results, top line reports, data mining results, status
`
`updates, support answers, custom greetings, images, and downloadable OLAP cubes.” Page 31.
`
`To access such content, a user is authenticated and “connected to the Webhouse application
`
`server, which is a Web server devoted to applications serving the qualified users.” Page 38.
`
`“The Webhouse application server delivers everything in browser-compatible format.” Page 38.
`
`One of skill in the art, as of Sept. 22, 2003, would understand this description in Kimball as
`
`describing the data warehouse delivering the results of data mining studies, as well as other
`
`information in the data Webhouse, in a browser compatible format so that this would be
`
`presented in a Web browser on a user’s screen. Pages 231-232 show some examples of
`
`interfaces that may be used to deliver content.
`
`45.
`
`As a particular example of data mining results that may be presented, the data
`
`Webhouse provides “[d]ata mining studies on near-term and long-term bases showing the
`
`evolution of customer demographic and behavior clusters.” Page 35. This refers to tracking,
`
`over time, the changes in customer clusters that result from data mining. The changes may be
`
`Page 17 of 27
`
`

`

`tracked in the customer dimension. Page 147. To do so, the customer dimension, or at least a
`
`portion of the customer dimension, may be treated as a slowly changing dimension, with the
`
`change in cluster over time being recorded. Page 147. So that meaningful decisions can be
`
`made based on the evolution of customer demographic and behavior clusters, the changes over
`
`time would need to be presented to a user, for example, by visually representing the information
`
`on the screen of the user’s computer. In view of this, one of skill in the art, as of Sept. 22, 2003,
`
`would understand the delivery of such data mining studies as including the presentation of the
`
`changes over time to an appropriate user.
`
`B.
`
`Kimball in view of Joshi
`
`46.
`
`Joshi describes a system, similar to Kimball’s, that includes examples of
`
`interfaces that may be adapted to Kimball’s data warehouse to present the results of data mining
`
`studies. In particular, Joshi describes a system that analyzes web logs to extract data that
`
`represents a user’s interaction with his or her web browser and a web server. Joshi, page 63.
`
`Data modeling is performed to determine the data model for the data warehouse. Joshi, pages
`
`64-65. The data model includes a dimensional model with fact tables and dimensional tables for
`
`the data extracted from the web logs. Joshi, page 64. The extracted data is stored in a data
`
`warehouse into the appropriate fact tables and dimension tables. Joshi, page 64. Data mining is
`
`applied to this stored data to detect meaningful patterns in the data. Joshi, pages 64-66. The data
`
`model also includes a schema for the data mining r