Transcription

Paper 036-2011No Do Overs: Tips for a Smooth EBI ImplementationTuesdi Helbig, Western Kentucky University, Bowling Green, KY, USAChris James, Western Kentucky University, Bowling Green, KY, USAGina Huff, Western Kentucky University, Bowling Green, KY, USAABSTRACTLike many technology-based projects, you get only one chance at implementing the SAS Enterprise BI Server at yourorganization. Coupled with a lack of knowledge about business intelligence, having to get it right the first time canmake implementation a daunting task. In August 2010, the Office of Institutional Research at Western KentuckyUniversity unveiled our portal using SAS Enterprise BI Server. While we had almost 50 years of combined SASprogramming experience, understanding how to make SAS Enterprise BI Server work at our institution wassomewhat overwhelming. Now that we are on the other side of a successful implementation, it is easy to see thesteps we took that made the project successful and the missteps we would correct if we could do it over again. Forthose considering or implementing SAS 9.2 Enterprise BI Server, we would like to share our tips for making SASEnterprise BI Server implementation a smooth process.INTRODUCTIONFor over 30 years, the staff in the Office of Institutional Research at Western Kentucky University have used SAS fordata analysis and reporting. Wanting to deliver information via the web, WKU installed the SAS 9.2 EnterpriseBusiness Intelligence Server. Following a steep learning curve and some educated guessing, IR developed a dataportal for use by academic administrators at the University. In August 2010, IR unveiled the IR-Decision SupportSystem to its campus.Before we embarked on our adventure, IR staff talked to other institutions that had already implemented the SAS Enterprise Business Intelligence Server. We discovered that each organization implemented EBI in a slightlydifferent manner. While the other institutions were very helpful, we knew nothing about the process so it was difficultto ask useful questions. We did what we thought was best for us, learning along the way. Having implemented oursystem, we felt others might benefit from the lessons we learned during our journey.BEFORE YOU STARTBefore you begin the process, there are a number of steps to take to prepare for the implementation. Here are oursuggestions.GET THE RIGHT PEOPLE ON YOUR TEAMIn How the Mighty Fall: And Why Some Companies Never Give In, Jim Collins (2009, pg. 56-57, 159-160) says oneof the most important traits of organizations that succeed is having the right people in the key seats. Not having theright people and/or having the wrong people on your team can easily derail your project. WKU was fortunate enoughto have the right mix of talent and skills on our staff to make the project a success. We had the technical expertise tolearn how to use and expand upon the system; we had extensive knowledge of the data; we had long ago cleanedand warehoused our longitudinal data; and we had a good idea what our users wanted. In addition, we had talentedpeople in IT, who helped us install and maintain the servers, as well as capture additional data for the system. Wealso had an enormous desire to make the project succeed.INVOLVE THE PEOPLE AFFECTED BY BIFirst, management has to be onboard. WKU had SAS present the Enterprise Business Intelligence Server to ourCouncil of Academic Deans and Academic Affairs Executive Staff. Their support was critical to WKU acquiring theproducts. Enrollment Management and Graduate Studies were particularly supportive because they saw the benefitsto their areas immediately. The deans also prodded us to get the system up and running because they were excitedabout getting their hands on the data.

No Do Overs: Tips for a Smooth EBI Implementation, continuedNext, involve anyone whose workload will be affected by the project. IR relies on IT to warehouse our transactionaldata and support the servers necessary for the project. Having them onboard from the beginning helped us quicklyresolve any data, warehousing, and server issues. We have met with IT many times during the process to discussservers, data needs, roll out strategies, help desk considerations, best practices for implementation, and how thesystem fits with established systems. Having the support of IT made the project exponentially better than it couldhave been if we had tried to do everything on our own.Finally, have power users and casual users test the system. We received some of our best ideas from our betatesters. We found simple things, such as adding a variable to flag certain groups of students (Honors, STEM majors,etc.) made the system much more useful. In response to our graphs, our beta testers quickly informed us that theywould rather create their own graphs if we would just provide the data. We also found out which data they did notcare about and were able to streamline our portal to include only features our users found useful.USE THE HELP SAS PROVIDESFrom project management to tech support, SAS can make the process run very smoothly. SAS has guided manyorganizations through the process, so why not make use of their expertise. WKU took advantage of the helpprovided by SAS sales staff, our project managers, our consultants, tech support, and other SAS employees we wereable to work with throughout the process. Having the expertise at SAS to rely on provided a map for the journey—wejust had to figure out how to read it.PLAN TO EXPANDNot knowing what we were getting into, WKU underestimated the potential user base and data volume during thesizing phase of the project. WKU actually had the 9.1.3 version ready to launch when we decided our sizing wouldnot accommodate our growing number of users and data volume. While it delayed our implementation by a fewmonths, we went through the sizing process once again. Eventually we bought the appropriate size servers,promoted to 9.2, and launched the system. Had we not re-evaluated our initial sizing, our system would have failedas soon as we launched it.ANTICIPATE AND RESOLVE POTENTIAL CONFLICTSUnderstandably, faculty, staff, and students at WKU get IR and IT confused, often thinking we are the same unit. Infact, both units write programs to provide data to users on campus. IR thought it best to talk to IT about the newsystem, what it would and would not do, and make sure we had any conflicts resolved before we got too far along onthe project. Before we rolled the new system out, IR also met with IT about formally distinguishing the responsibilitiesof each unit. Clients across campus sometimes make the same data requests of both units, and then use the resultsthat portray their unit in the best light. More often, clients have asked us to reconcile our numbers, which is alwaysdifferent because IR uses frozen data and IT uses transactional data.Through our discussions, we further delineated that IR would handle all reporting of “official” data, particularly as itrelates to students, courses, degrees, faculty, etc. IR would also handle all requests for trend data, as we takesnapshots of our data each semester, while IT uses live data, which change constantly. IT remained responsible foraddressing all requests for data to support day-to-day operations of the University. Sorting out the distinctionsbetween the responsibilities of IR and IT proved to be invaluable when we rolled out our system. In almost every BItraining session, we have been asked to explain the responsibilities of each unit.NOW YOU ARE OFF AND RUNNINGNow that you have laid the groundwork for a smooth implementation, it is time to get to work. Here are the next stepswe suggest.PICK A CATCHY NAME FOR YOUR SYSTEMWhile a name with “Business Intelligence” or “Portal” in the title would have been helpful, those options were notavailable to us at WKU. WKU already has a single sign-on portal, which we did not want clients to confuse with oursystem. Additionally, some faculty are offended by a suggestion that education is anything like business, so wewanted to steer clear of putting “business” in the title. While it sounded good at the time, we have since reconsideredthe name of our system. We settled on a somewhat lackluster title that describes what the system does, but does not2

No Do Overs: Tips for a Smooth EBI Implementation, continuedimmediately capture the user‟s attention—the IR-Decision Support System. We will eventually name our systemsomething catchy and will have to change our system and documentation at that point.GET THE APPROPRIATE TRAINING—THE SOONER, THE BETTERDepending on your particular needs, intensive training on the Business Intelligence products will be beneficial. Thetraining we found most helpful was the SAS Platform Administrator Fast Track. While we had experience in datacleaning, data warehousing and SAS programming, we had no experience administering a business intelligenceplatform. We also had staff participate in the training for SAS Data Integration Studio and Applied Analytics UsingSAS Enterprise Miner.TAKE ADVANTAGE OF CONSULTANTSWKU had our consultant tackle complex projects for us. When you are new to developing business objects, it isbeneficial to watch the process first before trying to build something yourself. This learning opportunity afforded usthe ability to take on increasingly complex projects of our own as well as the ability to maintain the projects started byour SAS consultant. Another important benefit of using SAS consultants is that they see the project from acompletely different view, which can lead to finding a better way to organize and structure your project.DARE TO BE CREATIVEDo not just use the system as-is. Implement a new theme across your web applications and use graphics, HTMLsnippets, and JavaScript to make the portal appealing to your organization. SAS made deploying a new theme mucheasier in 9.2 than in previous versions. A new theme allows you to customize colors and graphics included in thesystem. Additionally, custom HTML snippets and JavaScript provide added visual appeal to the standard look of theInformation Delivery Portal.INVEST TIME IN DESIGNDo not put too many reports on the system. No one wants to sift through pages of reports to find what they want.The beauty of OLAP cubes is the flexibility to use one base table to create tables using any of the hierarchies in theinformation map. However, you have to organize the system to best suit your organization‟s needs.IR chose to rely more on OLAP cubes for a number of reasons. While stored processes are useful for producingprint-ready reports, you have to guess at the format that will work for most people. Additionally, reports that usestored processes can only be modified by the IR staff. With OLAP cubes, users can move and filter the variableswithin the tables to meet their specific needs. No longer do we have to guess if departments would rather see theirmajors by ethnicity, then gender or by gender, then ethnicity. They can choose for themselves without talking to us.There are times when stored processes are the best tool. Based on the typical data requests we receive, we createda number of print-ready reports that can report the data for the university, college, department or major. We havealso spent the time to make the reports presentable using ODS Layout. For an example of our report, see Figure 1.In addition to deciding the types of reports to use, WKU decided to create tabs for each of the major content areas weuse in our reports. Thus, we have tabs for applicants, college readiness, courses, degrees, enrollment, facultyworkload, persistence, program review, and survey results. We also have unit specific tabs for colleges and otherunits, which only those units can see. The unit-specific tabs provide dashboards specific to that unit. The tabs alsouse reports and tables from content areas, but isolate the data so that only the data for the specific unit is shown(e.g., honors students, graduate students, majors in the College of Business). Additionally, unit-requested reportshave been added to some unit tabs. A typical page on the WKU IR-Decision Support System is shown in Figure 2.3

No Do Overs: Tips for a Smooth EBI Implementation, continuedFigure 1. Example of Stored Process Results Using ODS Layout4

No Do Overs: Tips for a Smooth EBI Implementation, continuedFigure 2. Typical Page on WKU IR-Decision Support SystemPROVIDE DOCUMENTATIONWe assumed our users were unfamiliar with the data on the system. Therefore, we provided users withdocumentation in a number of ways. First, we included a Documentation selection in the drop-down box of the Tableof Contents of reports built using OLAP cubes. (See Figure 3.)Figure 3. Example of Documentation Selection in Table of Contents Drop-Down BoxSelecting the Documentation report brings up a webpage that explains exactly what the data tables show and whichmeasures and variables are available for use in the report. The documentation also includes definitions of thevariables, as well as ranges for numeric variables. A sample of the documentation is shown in Figure 4.5

No Do Overs: Tips for a Smooth EBI Implementation, continuedFigure 4. Sample of Documentation for Tables Using OLAP CubesPREPARE THOUGHTFUL TRAINING MATERIALSWKU took the time to develop both a quick Navigation Guide and an extensive Users‟ Manual. We demonstrated thesystem for IT prior to rolling it out. Based on their experience, they suggested we make sure we prepared a simplisticand thorough Users‟ Manual and let users walk away with it in their hands after training. We have made both theUsers‟ Manual and Navigation Guide available on the portal. We also send hard copies home with each person whocomes to training. Our Users‟ Guide walks users through the portal step-by-step, providing tips and addressingpossible problems. It also provides exercises and solutions at the end.AFTER IMPLEMENTATIONOnce you have implemented the system, there are a number of things to keep in mind. Here is what we foundhelpful.ACT SWIFTLY ON USER FEEDBACKIncorporate valid user suggestions quickly. Respond to each suggestion. When users see that you have heard andacted upon their suggestions, they continue to provide feedback as they use the system. Even if you have to let themknow that their suggestion is on your to-do list or will not be implemented in the near future, they at least know youhave heard them and are taking their input seriously.KEEP A LOG OF ALL SUGGESTIONSWKU used the roll out sessions to get feedback on our system. As we presented the system to each college andunit, we wrote down all of the suggestions users made. We keep a running list of suggestions and regularly evaluatethe list by assigning each suggestion a priority and due date. Additionally, we have included a place on the portal forusers to provide comments (See Figure 5). Once a user submits a comment, it goes into our database. The userscan also see a running list of the comments they have provided to us.6

No Do Overs: Tips for a Smooth EBI Implementation, continuedFigure 5. Comments Box on WKU Portal PageCONCLUSIONWhile our journey has just begun, IR expects our portal to become even more successful as time goes on. Once wehave developed and launched the tables and reports requested by our users, we plan to begin using predictiveanalytics to create early warning systems to flag students at risk of not succeeding. While it is nice to know whereyou have been, it is often more important to see where you are going.REFERENCESCollins, Jim. (2009). How the Mighty Fall: And Why Some Companies Never Give In. New York: HarperCollinsPublishers, Inc.RECOMMENDED READINGCherland, Ryan. (2010). “Delivering Decision Support Information to Departmental Units.” SAS Global Forum 2010Proceedings, Paper 128-2010. Available at s10/1282010.pdfCollado, Evangeline & Borden, M. Paige. (2007). “To Cube or Not to Cube: Is OLAP the Better Choice?” SAS GlobalForum 2007 Proceedings, Paper 013-2007. Available at pdfKincheloe, Faron. (2010). “Lessons Learned from a BI Installation.” SAS Global Forum 2010 Proceedings, Paper043-2010. Available at s10/043-2010.pdfReda, Said. (2010). “Executive Dashboard: A Driver for Institutional Change and a Means to BroadenUnderstanding.” SAS Global Forum 2010 Proceedings, Paper 131-2010. Available ngs10/131-2010.pdfThamhain, Hans. J. (2005). Management of Technology: Managing Effectively in Technology-IntensiveOrganizations. New York: John Wiley & Sons, Inc.Yorkos, Ali. Y. & Murray, Maureen. (2010). “BI Development for Education by Our Institutional Research DepartmentOR How BI Tools Made Our Motto „IR Is Data‟ a Fact.” SAS Global Forum 2010 Proceedings, Paper 1302010. Available at s10/130-2010.pdf7

No Do Overs: Tips for a Smooth EBI Implementation, continuedACKNOWLEDGEMENTSWKU‟s Office of Institutional Research would like to thank our colleagues in IT who have helped us make this projecta possibility. Without their continued support, this project would not have been such a success.CONTACT INFORMATIONYour comments and questions are valued and encouraged. Contact the authors at:Tuesdi Helbig, Ph.D.Institutional ResearchWestern Kentucky University1906 College Heights Blvd. #11011Bowling Green, KY 42101-1011Work Phone: (270) 745-3250Fax: (270) 745-5442Email: [email protected] JamesInstitutional ResearchWestern Kentucky University1906 College Heights Blvd. #11011Bowling Green, KY 42101-1011Work Phone: (270) 745-3250Fax: (270) 745-5442Email: [email protected] HuffInstitutional ResearchWestern Kentucky University1906 College Heights Blvd. #11011Bowling Green, KY 42101-1011Work Phone: (270) 745-3250Fax: (270) 745-5442Email: [email protected] and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SASInstitute Inc. in the USA and other countries. indicates USA registration.Other brand and product names are trademarks of their respective companies.8

The beauty of OLAP cubes is the flexibility to use one base table to create tables using any of the hierarchies in the information map. However, you have to organize the system to best suit your organization‟s needs. IR chose to rely more on OLAP cubes for a number of