Seize the Day!

In 2006, my then-CIO tasked me with learning about ‘this data warehousing thing’.  I fired up Google, typed in ‘Data Warehouse Higher Education’ and got a hit – Higher Education Data Warehousing Forum.  “Sounds pretty on the nose to me”, I thought and thus began a relationship with an organization that has changed my life and changed the institutions I worked for, all for the better.

In those early days I would read every post with a thirst for the knowledge they offered.  The community was, and still is, extraordinarily generous.  I remember long, detailed responses from people like me who were trying to figure out how to apply dimensional data modeling concepts to our weird world of higher education. I could see new vistas opening up and I wanted to be part of it.

Fast forward 10 years.  The call for board nominations goes out on the Forum, as it did two days ago, and one of my colleagues asks me “are you interested?”  I thought about the intervening years, during which I launched two BI initiatives.    I thought about how finding HEDW led me to a wealth of knowledge that laid the foundations for a career in data and reporting, in putting the power of information into the hands of those who need it (or trying to anyway).  I thought about the interesting colleagues I met at the conferences over the years, and I said ‘Yes!’  It was time to give back and to make sure that this valuable, volunteer-led organization continued to thrive.

Is it time for you to give back?  Serving on the HEDW Board is interesting and challenging work, filled with a sense of purpose and a lot of laughter with amazing people.  If you are at all interested, please consider running for a board seat.  Help us shape the future of data-driven decision making at universities around the world as we adapt to the rapidly changing landscape we inhabit.

HEDW Board FAQ

2018 HEDW Conference

Registration is Open for the 2018 Conference

The 2018 HEDW Conference is April 8-11, at Oregon State University in Corvallis, Oregon. The HEDW Board in conjunction with Oregon State have put together a premier event for our members.

Choose from two full-day training sessions

The conference begins Sunday, April 8 with two outstanding (optional-extra fee) training session choices, Information Dashboard Design with Nick Desbartas, Practical Reporting, Inc. and The Modern Data Ecosystem with Joe Caserta. Click on one of the training session links for more information.

Renown Keynote Speaker

Stephen Few, Perceptual Edge, will open the 2018 HEDW Conference on Monday, April 9 at 8 am with a keynote presentation, Data is Not the Solution.

“…Stephen Few will show how data professions, as routinely performed, are creating more confusion than understanding, more problems than solutions. He will go on to describe the shift in focus that must occur for data professionals and their organizations to produce the understanding that is needed to create a better world. In the world of higher education, this shift in focus is critical.”

More information about Stephen’s presentation, visit the Keynote Speaker page.

Full three-day Conference Schedule

This year HEDW has an excellent selection of presentations and panel discussions but will also have space for the Birds of a Feather sessions. In the coming weeks, the conference schedule will be populated with the days/times of the presentations. For more information about the presentations and presenters, open the links on the right hand side of the conference page.

Three Steps to the 2018 HEDW Conference

Detailed information for the conference registration, lodging and ground transportation are located on the 2018 HEDW conference registration page.

We look forward to seeing you in April,
Michael Hansen, Executive Director/CDO
Institutional Analytics & Reporting
Oregon State University

Conference Reflections: Cathy Lester Grants

Thinking about attending the annual HEDW Conference?

A few short years ago, when I was new to my job on our University Data Warehouse team, I had the opportunity to attend my first conference. The sessions, the training, the keynote were all on target; but for me, the networking was the most invaluable part of the conference.

Don’t let the cost keep you from this awesome event!

The HEDW Board endeavors to keep the conference costs low, but we know many institutions have budget constraints.  Consequently the HEDW Forum offers two grants to cover the costs of conference attendance in memory of Cathy Lester, one of its founding members. The grants are aimed at institutions with limited data warehousing budgets that could add to the conference and/or benefit from attending. One attendance grant and one speaker’s grant are made annually.

See what last year’s grant winners had to say about the opportunity to attend:

Because of this grant, I was able to share data visualization best practices with my fellow attendees. It was so meaningful for me to receive such positive feedback about the session throughout the conference, and to connect with other data visualization lovers.

In addition to speaking at the conference, this grant also allowed me to experience this conference for the first time. I was able to sit in on talks from institutions around the country and hear solutions for the very issues I had been wrestling with.

I continue to implement the things I learned at the conference in my day-to-day tasks.

Of course, one of my favorite things about being able to attend this conference was the opportunity to network with data warehousing professionals from such a diverse group of institutions. I learned nearly as much from chatting with other attendees over meals and coffee breaks as I did from the structured sessions throughout the conference.

I’m so grateful to the committee for providing me with the opportunity to attend, speak at, and learn from this year’s HEDW Conference!

The conference was fabulous, it was one of the best run conference I have been to in my career!

You can find information on applying for the HEDW Grants here:

https://hedw.org/conference-home/2018-oregon-state-university/2018-hedw-conference-grants/

 

Sherri Flaks

IT Director, Enterprise Data & Analytics

Johns Hopkins University & Medicine

The Dean’s Information Challenge: From Data to Dashboard

(First in a series of a published EDUCAUSE article November, 2016)

by Jeff Meteyer, University of Rochester

Summary:  Data drives dashboard construction. At the University of Rochester, our dashboard development teams sought assistance and created processes to ensure consistent results and widely agreed-on metric definitions. These experiences inspired data governance and communication partnerships across various systems.

This case study focuses on issues of interest to deans and the challenges of developing a dashboard system.

The Dean’s Perspective

New and existing deans face daunting pressures in managing their areas of responsibility; among their required tasks are to:

  • Develop student outcomes strategy
  • Understand the unit’s financial position, as well as its operational and academic standings
  • Interpret information from an avalanche of data from various (potentially non-integrated) sources
  • Ensure the effective and efficient use of resources related to finances, people and space
  • Engage and support cost/value discussions around tuition and student job market success
  • Solicit research dollars in targeted areas of opportunity
  • Make strategic faculty hiring decisions and develop faculty retention strategies

Deans are interested in existing performance metrics and how they portray both current and historical performance. But, in attempting to address such questions, deans might find themselves in information silos — sometimes without realizing it. Sometimes these “silos” result from institutional evolution, as described below.

When deans rise through the ranks of a particular institutional branch, they might be unaware of “tribal” knowledge from other parts of the institution. Further, they often rely on subject matter experts to help lead the way in terms of definition and performance. However, some measurements, such as student metrics, do not behave according to a predictable data-driven model across an entire university. In addition, some deans might hail from other institutions with particular dashboard cultures and expectation levels. These deans might believe it is easier to bring this “vision” to their new establishment, only to find that the new institution’s data might not easily suit or fully populate the former institution’s dashboard model. Deans might find themselves without an existing dashboard model, so the activity of designing one becomes an additional task to manage.

Deans have reporting priorities beyond the student area, including research ranks, investment trends, salary expenditures, and submission/acceptance ratios for research projects. Deans might focus on the workload ratio of various faculty members or principal investigators in relation to research time, student course load, and mentoring. Further, tenure tracking and overall demographic measures play a role in both recruiting new talent and retaining existing performers.

Financial reporting priorities may include operational metrics, which indicate where a dean’s unit stands in relation to its plan or budget and (ideally) in relation to previous timeframes. Forecasting future fiscal cycles is challenging in the absence of historical data and an understanding of how that data relates to prediction models.

When data definitions vary, exceptions or anomalies can arise in the metrics using those varying definitions. Many people in an institution’s higher ranks have gone through evolving organizations where successive leaders invented new ways of thinking and addressing organizational shortfalls to establish their influence on the institution’s practices. However, deans and administrators do not want to risk misinterpretation or misstep due to nebulous data definitions; they want to be relatively certain, given the data integrity and interpretation, that the decisions they make align with their strategy.

The recent rise in data governance teams — which address issues such as data definition, data security, and determining usage at various institutional levels — is helping to establish process-driven decision-making, reducing the challenges of data sharing and interpretation previously experienced by deans and others.

Data Challenges

When designing dashboards as information portals, it is important to ask what the appropriate metric is and how it will be used for decision making. A visualization or reporting team sometimes illustrates what can be done, but their design intentions might actually muddy the dean’s decision-making process. The goal should be to define the metric in as granular a way as possible so that the resulting illustration of that metric helps users determine a course of action.

For example, a five-year trend in research spending (in aggregate) might give users a sense of what to expect for forecasting, but if the goals are at the sublevel (such as capital expense reduction), the data sets should be represented so that users can easily identify distinct signals or trends without having to redesign a report. Setting thresholds and alert logic is important; data sets arrive quickly, and the buildup lets user see the signal and act accordingly. If data does not refresh often enough, the signal may be lost or delayed, resulting in missed opportunity.

Development Process

At the University of Rochester, our reporting and analytics team worked in tandem with then-Dean Robert Clark and the dean’s support staff to outline a vision for developing institutional dashboards. The dean offered a level of creative freedom, letting the team illustrate the dashboard design and as a group define the resulting metrics.

Figure 4. An example institutional dashboard

The team developed various proofs of concept showing the information that could be gleaned from the data warehouse and other systems across the decentralized institution. Multiple review sessions ensued, in which team members explained why they thought a particular visual accurately represented a metric’s performance over time. Our mantra throughout the design process was: keep the message simple, while also ensuring that a single visualization could answer multiple questions. The challenge was in defining which level of information provided sufficient actionable support, versus having a visualization that drilled down a “discovery wormhole.”

Defining data elements — and finding agreement on those definitions among contributing parties — proved to be an ongoing challenge. Although we reached agreement for the initial round of institutional dashboards, the process pointed to the need for data governance. We are investigating tools such as iData’s Cookbook, which can help an organization capture and publish data definitions that require a structured approval process, as being the authoritative source. A recognized source containing definitions for terms, metrics, and reports can help the team maintain consistency in the design and development of reporting and analytics.

Our data analysis team learned the importance of understanding how data definitions matured over time and the different ways data was collected and classified in the information systems. This understanding sometimes led to the development of bridges between data transformations to allow presentation of a continuous data story. Special events, such as ERP or other system replacements, drove the need to consider which data definitions and transformations are needed and how a data conversion strategy reaches beyond the source system to reporting systems. Weeks of data modeling and cleanup ensued; our goal was to assure users that we could tie the ends together and portray multiyear windows of trends.

Results

After nine months, we arrived at an agreed-to set of institutional dashboards that worked to illustrate student, faculty, and research relationships and performance over time. We designed the visualizations to allow multiple questions to be asked and answered, and included various parameters that users could manipulate to discover results. We also addressed the granular versus aggregated views to allow “drill-throughs” when security rules permitted access.

Dashboarding might seem like another reporting project. However, we encountered many variations in grey requirements and acceptance of visualization “art” for the final deliverables — what worked as a pie chart one week worked better as a bar chart the following week — and thus helped us to exceed many of our resource forecasts. The bridge between data graphing and the “story to be told” can be vast, and the environments and issues at the time can influence how a visualization represents data. At the same time, adhering to standard, defined metrics over time helps solidify and standardize how the visualization can illustrate performance variations. Further, adhering to strong data definitions reduces the translation risk for data extending across multiple instances.

In our case, data cleanliness issues and deciding how to portray data certainly exceeded our original two-week time estimate. However, our hope is that we have now put processes in place to reduce the development timeline to a more manageable timeframe, driven by agile scheduling, reusable components, and standards.

Future Plans

A dashboard effort can lead to analysis paralysis if the signaling features are not strong enough to point out anomalies. A visualization can become stale — like a roadside billboard you learn to ignore. It’s important to employ processes to provide data-driven refreshes that show new situations, as well as the effect of decisions made based on previously illustrated data. This correlating between data and action helps prove the worth of tracking information and displaying results to confirm how a strategy fared.

Given human tendencies toward instant gratification, dashboards must produce quick responses to actions at a level granular enough to show correlation. As deans continue to work under various pressures, their ability to measure performance rapidly and discover details that serve the goals are critical dashboard deliverables. As their strategic partner, it is imperative that IT teams help deans address data cleanliness, definition, and refresh concerns.

IT Leadership Is Key

For colleges and universities, lack of data is not usually a problem. Getting information to academic decision makers, however, is not so easy. The problems involved are not merely technical issues such as data integration; to create dashboards and reports that help deans and administrators make decisions, IT leaders must help their institutions develop data governance, build agreement around data definitions, understand data stewardship, and communicate across silos to reach agreement on dashboard goals.

The lessons learned from these four case studies can be a roadmap for other institutions as they work to provide academic leaders the information they need to make effective and fruitful decisions.

 

© 2016 Mike Wolf, Martha Taimuty, Monal Patel, and Jeffrey Meteyer. The text of this EDUCAUSE Review online article is licensed under The text of this article is licensed under Creative Commons BY-NC-ND 4.0.

Teaching Business Intelligence at Cornell University

by Jeff Christen, Cornell University

For the Fall 2016 semester, I had the opportunity to teach a pilot, four credit, masters level course through Cornell University’s Computer and Information Science department. The course was entitled, Business Intelligence Systems and its goal was to give students a solid foundation and understanding of BI concepts including dimensional data modeling, ETL design and data visualizations. In addition, the course reinforced their technical learning through hands-on experience with industry standard tools.

Course Content

The primary text for the course was The Data Warehouse Toolkit Third Edition by Ralph Kimball & Margy Ross, supplemented with some on-line articles. The first third of the semester focused on dimensional modeling using the Kimball methodology. The second third of the semester covered SQL, ETL design & development, and basic data visualization concepts through hands on tutorials and individual assignments using the virtual BI lab. The final third of the semester was largely focused on the team projects with some special topic discussions.

Virtual BI Lab

The core of the hands-on learning experience was a virtual lab environment where each student had their own Amazon Web Services Workspace (virtual desktop) with Oracle SQL Developer, the WhereScape RED ETL tool and Tableau desktop installed. In addition to their Workspace, each student had access to the class database which was an AWS Oracle RDS with many sample data sets plus a database account for each student with access to create tables, views, etc.  The students had everything they needed to explore data sets, create their own dimensional models and populate a data warehouse using WhereScape RED and finally build various visualizations with Tableau.  The use of the AWS Workspaces allowed the student to access their environment from any device and not worry about the installation of the various software products and database drivers. The AWS Workspaces lab also allowed us to work through examples interactively during class. This also came in handy for office hours when a student could simply open their laptop and show me where they were struggling with their assignment.

Team Project

In addition to several individual assignments focused on core skills, the students also participated in a team project.  Project teams consisted of 4-5 students and each was one of three BI projects using real world Cornell business challenges and associated data sets.  There were multiple teams per project which helped illustrate that there are multiple ways to implement a BI solution.

Project teams had access to a snapshot of the necessary source data via the virtual lab.  The teams were also given a description of the business goals for their project and they had the opportunity to question a business representative about their project.  The students had milestone deliverables throughout the project to help keep them on track by providing deadlines and feedback on their progress.  Example milestones included: requirements tracking, logical dimensional data model, and ETL source to target mapping document.

The final deliverable was a working Proof of Concept, fully documented, along with a 20-minute presentation on their work to their client and classmates.  Each project team also created an online portfolio of their project work using ePortfolio by Digication.  The project ePortfolio sites served two purposes;  1) They serve as a nice reference for the client who may choose to implement one of the student Proof of Concepts and  2) The students may simply add their ePortfolio url to their resume to showcase their project work.

Here are two sample team ePortfolio sites:

MPS Projects for Spring semester

As part of Cornell’s Masters of Professional Studies – Information Science program, students must complete a semester long team project (http://infosci.cornell.edu/academics/mps/mps-project) MPS students work in teams of 2-3 and receive credits for the project. Two of the project options available to students this Spring semester are from the Fall BI course.  The MPS student teams will work much more closely with the Cornell business clients to further refine requirements from the Proof of Concepts created by the Fall class teams with the goal of implementing a production BI solution for their client.  For the MPS projects, the students will not be working in the virtual lab environment, but in Cornell’s DW/BI environment where they will be introduced to Cornell’s DW/BI data modeling and coding standards and processes.  The student teams will meet weekly with members of Cornell’s Office of Data Architecture and Analytics (ODAA) where they will receive feedback on their work in the form of model reviews, code review and project discussions.  The two BI MPS projects are co-sponsored by the business clients and ODAA. This should result in deliverables that not only meet the business requirements but are sustainable and supportable by ODAA after the students leave.

Feedback

Student feedback on the Fall BI Systems course has been very positive.  Students found the blend of Kimball Dimensional Modeling theory, mixed with a lot of hands on work, to be an effective means of learning the concepts and complexities of data. Thirty-two students successfully completed the course, and Cornell plans to offer the course again for Fall semester 2017.

2017 HEDW Survey of Top 10 Issues

by Hank Childers, University of Arizona

We recently surveyed HEDW’s membership to determine the top 10 topics of interest.  There were over 300 responses.  The survey was similar to the one we did the year before, which allows us to compare the results.  Aaron Walz of Purdue University and I will present an analysis of the results at the 2017 HEDW Conference at the University of Arizona.  But here’s a sneak preview of some of the headlines.

Data Governance Is #1 for the second straight year

57% of respondents placed this in their Top 10.  The pattern seems clear and compelling.  But what’s the message?  Is this about data quality, which was independently ranked #3?  Or about data definitions, which was independently ranked #4?  Is this about the need to have data governance, or about how to go about actually doing it successfully?  Considering that the cluster of data governance, data quality, and data definitions occupies three of the top four spots, this is a very strong signal!  It seems like a ripe topic for presentations this coming April.

Data Governance is only Top 10 topic named by more than 50% of respondents

As strong as the signal around data governance seems to be, it’s interesting to note that it’s the only topic to meet the 50% threshold.  This speaks to the differences among us.  Perhaps this is related to different institutions being at different points in their evolution, or perhaps different priorities operating at different institutions.  Or perhaps half the institutions have mastered data governance, and the rest of us haven’t!

Student Success Climbs to #2

This is not a surprise, since it was #3 last year.  It is the only issue in the “Higher Education Issues & Opportunities” category to make the Top 10.  And that was true last year as well.  It was selected by 47% of the respondents, so it’s close to that 50% threshold.  And given the visible attention being paid in general to student success by our institutions this is not a surprise.  The next highest ranked item in that category was Learning Analytics at #16.

No People topics in the Top 10

No topics in the category called “People, including hard skills, soft skills, & marketplace” made it to the Top 10.  The highest was Data Modeling at #15.  Maybe we don’t like people.

Top 12 is the new Top 10

Similar to what happened last year, there was so little difference between #10 (30%), #11(29%), and #12 (28%), that we expanded the list from 10 to 12, based on the natural break in the data.  (#13 was 23%.)

blogtop102016dec

Mobile comes in dead last

Mobile came in as #51 out of 51, with 5% of the respondents naming it as an issue.  Not that long ago it would likely have been much higher up.  Things do change.

Speaking of maturity

We will soon be contacting the membership inviting HEDW members to participate in a survey based on the BI Maturity Model.  We did this survey three years ago, and we judge it a good time to do it again.  It should be very interesting to see what has changed and what hasn’t on an overall basis, and also for those individual institutions who participated in the earlier survey to assess their progress.  Aaron Walz and I will also report on our analysis of these results at the HEDW meeting in April 2017.  More to come on this.

 

(November 30, 2017; hankc@email.arizona.edu)

Welcome to the HEDW Blog!

Greetings, HEDW friends!

Fall is finally upon us, by the calendar and the weather outside, and the return of bustle to our campuses.  This is also the start of a new planning season for HEDW, which kicked off with our annual Board meeting a couple weeks ago in Tucson, the beautiful site of our 2017 conference.

This is HEDW’s 14th year, and over the past years we’ve been thrilled to see more and more folks from institutions around the U.S. and beyond join our community.  This year and beyond, I hope that we all can build on the great relationships already established and forge new connections as we collaborate, commiserate, and share experiences in higher education data warehousing and analytics.

With that in mind, the Board has chosen a thematic goal for this year to “Consciously experiment with alternate means of engaging the membership”.  In support of this, we’ll be exploring ways to keep the enthusiasm and collaboration we see at our annual conferences strong all year long.  In particular, we’re:

  • Exploring webinar technology for use in structured presentations, group discussions, and other ways.
  • Planning regular blog entries to share thoughts and wonderings with the community – this is the first!

And, of course, our core activities continue:

  • Gearing up for our 2017 conference, in sunny Tucson, hosted by the University of Arizona, April 23rd – 26th, 2017
  • Ongoing research work, including our Research Top 10 survey, and continuing maturity work
  • Online discussion forums on hedw.org.

You can find the roster of current Board members on the HEDW site at https://hedw.org/about-us/executive-board/.  Please feel free to reach out to us via the contact link on the same page, or in the new “Ask the Board” forum thread.

I’m so glad to be part of this vibrant community, and to have met and learned from – in person or virtually – so many of you.  I hope to see you online in the forums and at the conference in Tucson!

— Amy

Amy A. Miller, HEDW President 2016-2017
IT Director, Enterprise Information & Analytics, University of Pennsylvania