Data-Driven Decision-Making: It’s a Catch-Up Game
- By Linda L. Briggs
- 07/23/06
Feeling the pressing need for
sophisticated business intelligence
solutions and processes, higher
ed looks for models—wherever
they exist.
Good data is powerful. At Rensselaer Polytechnic Institute in upstate New York, Data Warehouse
Program Manager Ora Fish was astounded when her data warehousing team was roundly applauded for presenting
data on graduate admissions—data derived from the data warehouse—to top-level officials. “The cabinet
had never had [solid] numbers before,” Fish realized. Suddenly, she says, “they didn’t have to argue about whose
numbers were right. They could talk about the issues: how we’re doing in graduate admissions.”
Having an abundance of data residing in individual silos across campus, but little decision-ready information,
is a typical scenario at many institutions. “We’ve always said the campus was ‘data-rich,’ even when we had manual
collection of that data. But we couldn’t get access to it,” explains Michael Zastrocky, a VP with research firm
Gartner. Zastrocky has more than 30 years of experience in higher education, and has followed
the campus data situation for decades. “Then we moved to relational databases, and people said, ‘We’re still
having trouble getting access to the right information.’ That’s been a major issue on campus.”
One problem: The terms data warehousing (DW) and business intelligence (BI) refer to very different things,
although the two often go hand-in-hand. Data warehousing describes an architecture for storing data; business
intelligence involves analytics run against that data in order to discover patterns and glean information to analyze
past performance and predict future events. Typically, a data warehouse is constructed first, in itself a complex
process that requires significant planning and institution-wide coordination. BI tools are then used on the warehoused
data to produce reports, dashboard displays, and other windows into the data.
While many higher ed institutions now
have extensive data residing in enterprise
resource planning (ERP) solutions, the
typical US university owns what only
amounts to “an infant data warehouse,”
says David Wells, director of education
with TDWI. While ERP solutions
can offer detailed reports, and ERP vendors
are moving to include more analytics
solutions with their products, they can’t
offer the kinds of nuanced ways to slice,
dice, and analyze data that a data-warehouse-
and-BI solution can. Yet higher ed
lags behind private industry in its adoption
of BI. “A few of the leaders in higher
ed have decent data warehouses,” Wells
says, but points out that most schools are
still focused simply on maturing their data
warehouse and data integration capabilities,
“with analytics farther out on the
horizon.” There are exceptions, however.
RPI and UI: Institutions Leading the Charge
Rensselaer Polytechnic Institute, for one,
began to roll out a data warehousing solution
in 2001, when it decided that the
school’s new ERP solution couldn’t provide
all the analytics and reporting needed.
The warehouse was constructed
across a three-year timeline; RPI also
selected a series of BI tools from Brio
Software (later acquired by Hyperion
Solutions; www.hyperion.com) for analytics.
RPI’s DW and BI solution is one of
the most advanced in US higher ed, and
Fish often speaks about it at conferences.
Several years ago at the University of
Illinois (which, along with RPI, TDWI’s
Wells cites as a leader in data warehousing),
administrators decided to build an
enterprise-wide data warehouse. At the
same time, the university moved to a new
ERP system, replacing a number of legacy
data systems across UI’s three campuses.
The new, institution-wide system
raised the question: How to get meaningful
data back out of the ERP system?
To meet this need, over a several-year
process UI has constructed a mature data
warehouse that includes all the core business
data from the ERP. In addition, the
university uses enterprise decision support
tools from Business Objects to access and analyze
warehoused data.
UI’s case is different from most others
in higher ed not only because the initiative
is so large and advanced, but because
the institution’s data warehouse is truly
institution-wide—a feat difficult to
accomplish in higher education (see “The
Challenge of BI in Higher Ed”).
Still, UI’s Director of Data Warehousing
Andrea Ballinger strongly suggests that
the university’s concurrent rollout of the
ERP system and the data warehouse is not
an approach she would recommend to
other schools. In fact, she says, trying to
do both at the same time was perhaps the
biggest challenge the school faced in its
data warehouse deployment. But the
deployment was also challenging, she
adds, “because we were trying to do
something that to our knowledge, no
other university had done before. There
are many data warehouses out there, but
we knew of no other in higher education
that started from zero, with a focus on
enterprise-wide data warehousing. We
wanted to do it right from the beginning,
and end up with one fully integrated
enterprise-wide warehouse.”
Bridging IT and Business
"The key to a successful BI program is really about the successful marriage between IT and the business side of the institution."
Ora Fish, Rensselaer Polytechnic Institute
Key to building a data warehouse is bridging
the long-standing gap between IT and
business. More so than with many other
technology solutions to business problems, data warehousing tests the bridges
between IT and the rest of the campus.
At UI, for instance, the bridge between
IT and business is largely handled by
three “functional area coordinators.”
These subject-matter experts focus on
three key data areas: students, finance,
and human resources. They act as liaisons
between their areas of specialty and the
data warehousing team. According to
Aaron Walz, business architect for the
Decision Support team, these individuals
“translate what the customers are saying,
putting that into a language that the technical
staff can understand.” Yes, any significant
IT initiative needs the business
side of the house on board. But in such a
case, says RPI’s Fish, “IT really needs to
understand what questions business
needs to ask—and at a much higher level,
because now you’re talking analytics.”
That means that the people who know
how to produce reports—often highly
technical IT staff within the DW group—
need a deep and broad understanding of
what drives the institute. In the world of
data warehousing, Fish says, “it’s really
about the successful marriage between IT
and the business side. That’s the key to a
successful BI program.” And, don’t imagine
that creating a data warehouse is a
project with an end date, or that it’s
a quick fix for internal reporting
problems, she adds. Maintaining a
data warehouse is an ongoing
process that requires constant attention,
user training, and marketing.
“It’s not a project with beginning
and end dates,” Fish explains.
“That’s something that we’ve
learned here: It’s a new service that
we’re providing the campus.”
Even at UI, with its mature data
warehouse, the process is ongoing, says
Walz. “Ideally, in building an enterprise
data warehouse, you don’t stop there; you
build data marts and other structures to
give people simplified ways to get at the
data. We’ve started on that process, but
it’s by no means complete.”
Build It and They Will Come?
Part of the data warehouse challenge is
the constant effort to explain its usefulness
to users. “‘Build it and they will
come’ is a very ugly myth,” Walz says
bluntly. When the university erected its
new systemwide ERP structure, some of
the standard mainframe reports were
duplicated, but not all. Even so, Walz
says, users “still had to be convinced that
coming to the warehouse was worth their
time and that they needed it. You have to
provide something that is very targeted to
what they’re trying to do, so they can see
that it’s helpful.”
Both Rensselaer Polytechnic Institute (above)
and the University of Illinois are leaders in
academic data warehousing, but their business
intelligence challenges were markedly different.
His group puts ongoing effort into promoting
the warehouse, including training
sessions and periodic messages to everyone
with a warehouse account, encouraging
them to use it. The Decision Support
team also analyzes who’s using the warehouse
and how, and uses that information
to drive marketing efforts. “We also make
presentations to different groups on campus,
such as Human Resources and business managers. We tell them, ‘Hey, here’s
what you can do with the data warehouse.’”
Walz says. “If it takes too much
time, or if it’s too complicated to get
access, they just don’t use it.”
Measuring the Return
Quantifying a return on investment (ROI)
for a data warehouse or BI investment can
be challenging in the private sector; in the
complex world of a not-for-profit public
institution like a university, it can be close
to impossible. At UI, Ballinger says she is
just beginning a project to collect returnon-
value (ROV) figures that place a dollar
amount on the school’s hefty investment
in its data warehouse. (ROV attempts to
look beyond ROI and take a broader,
systemwide view of the contribution of
IT.) In looking for measurable returns,
Ballinger is seeking answers to this basic
question: What has access to the data
warehouse done to improve the efficiency
and effectiveness of the units using it?
What Can You Do with Data Warehousing and BI?
WHAT SORTS OF DATA (that you’re not getting now
from ERP or elsewhere in IT) can a BI solution provide?
Here are three examples of of nuanced
information that a DW/BI solution can produce:
- Is our online ed competing with our traditional ed?
The University of Illinois is spread across three campuses
(Chicago, Springfield, and Urbana-Champaign).
Springfield focuses on online and distance ed for
liberal arts. When the Springfield provost asked, “Are
we cannibalizing our existing business by offering
courses online?” the campus used information in the
warehouse to create reports that found they weren’t—
students were supplementing their schedules with
additional hours of online courses.
- How attractive are we? One way that Rensselaer
Polytechnic Institute (NY) uses its data warehouse is
to track admissions closely by answering questions
like: “Are we attracting the types of students we
want? How well are we translating student inquiries
into applications? How about converting those applications
into admitting students?” When a new strategy
for attracting students was introduced recently, the
results were immediately available on a daily basis
from the data warehouse via analytic dashboards.
- Researching research. Rensselaer is a large
research university, so it finds the data warehouse
useful for monitoring research, with questions like:
“Where, exactly, is our research money coming from?
Who’s performing the research—are they working
through the university or the research center?” Unlike
in the past, “It’s very easy now to slice and dice, and
everyone agrees on the numbers,” says RPI Data
Warehouse Program Manager Ora Fish.
For example, a purchasing agent renegotiating
a licensing plan based on bulk
purchases might
be able to save
money, based on
data revealing
usage across the
university rather than in one department.
Ballinger says that her plan is to work at
gradually gathering those sorts of dollar
amounts into a cohesive picture. “We
have a boatload of data about who’s
using the data warehouse when,” she
says, “but we don’t know what they’re
doing differently from when they didn’t
have the warehouse.” Because UI’s data
warehousing project is such a large effort,
Ballinger says it’s critical that she have
that data in order to continue to make the
case (or not) for the data warehouse.
Show Results Early
Setting up a data warehouse and BI program
is a large and expensive undertaking
that will require plenty of teamwork,
meetings, and time from those campus
“customers” who will be using the system.
It’s therefore important to show
results as quickly as possible. “We
demonstrated the value immediately;
within six months,” Fish says. “That’s
important; you need to show payback in
a relatively short length of time.”
That’s why Fish recommends that
schools build an analytical dashboard
along with the warehouse. Dashboards
are tools that focus on specific metrics
within a project, in order to give a quick,
graphical assessment. By giving the leadership
that sort of instant, high-level view
of various areas, she says, RPI was able to
demonstrate a nearly immediate return to
top management. It was hugely satisfying,
she adds, to pull data in from the
warehouse and provide the first looks at
an analytical dashboard. “People’s eyes
were just popping out,” she recalls. “‘Oh,
is that what we’re doing?’ they said. ‘I
thought we were going up, but the numbers
don’t support that!’”
The Challenge of BI in Higher Ed
LAUNCHING A DATA WAREHOUSING and business intelligence initiative is tough enough in private
industry. But higher ed institutions face some extra hurdles in rolling out a data warehouse, according
to David Wells, director of education with TDWI, a training
and educational institute for business and IT professionals. Among the challenges Wells and others cite:
- Gathering consensus. Unlike a typical large business, a university isn’t a single enterprise. It can be
politically difficult if not impossible to get various entities across campuses to agree on basic issues,
such as what the end-purpose of the warehouse is, what data to share, and who should be in charge.
“I think it’s a more challenging business case to make,” Wells says, partly because institutions typically
divide immediately at the top into academic and administrative sectors. Because a data warehouse
has to integrate across organizational boundaries, he says, the arguments that might sell administrators
usually don’t resonate with the academic community, and vice versa. “It takes a real believer driving
from the top to make it happen.”
- Define, define. Because of the lack of a centralized, top-down structure, issues like defining the elements
in the warehouse can be surprisingly difficult. “A major challenge
is bringing the campus together in common definitions,” says Ora Fish,
data warehouse program manager at Rensselaer Polytechnic Institute
in Troy, New York. “Take a question like, ‘How many faculty do we have?’
It sounds very straightforward. Well, you have to define what faculty is,
and then the answer depends upon who is counting.” The challenge
with a data warehouse, Fish says, is to get everyone to agree on a single
definition—in this case, for instance, a definition of faculty—then
set up attributes that allow the data to be viewed from a variety of
angles and nuances.
- Hiding in the ‘shadows.’ Another issue that’s not unique to higher ed
but may be exacerbated there, says Wells: the degree to which the
operational data exists in “shadow systems.” Because colleges and universities
have a strong tendency to build departmental systems that don’t depend on central computing,
the same kinds of data, in entirely different forms, can be found in pockets across a typical university.
“The data can have varying degrees of quality and completeness,” Wells says. “That can be a challenge.”
- America’s top model? Unlike business sectors such as insurance, finance, or health care, where
established models exist for creating a data warehouse, there’s no such model in higher education.
“There don’t seem to be any good industry models for higher ed,” says Wells—perhaps, he suggests,
because higher education in general lags behind corporate culture in embracing data warehousing
and BI. In describing her school’s highly complex data warehousing rollout, Andrea Ballinger, director
of data warehousing at the University of Illinois, confirms, “We couldn’t find models—and we
couldn’t find anything in the corporate world that would fit, either.”
- Use what you choose. Colleges typically decide issues by consensus and committee—not the easiest
way to build cross-departmental, institution-wide solutions. Also, mandating use of a single product can
be tough in a climate where academic freedom rules. The University of Illinois addresses this by offering a
variety of tools for accessing the data warehouse, so that no one is forced to use a specific product.
On the other hand, higher education has advantages that some publicly traded companies, hounded
by quarterly earnings reports, might appreciate. To wit: a focus away from pure profitability. “To do
anything new and innovative in this type of culture is more challenging, yes,” says RPI’s Fish. “But yet,
we’re different—we’ve been here more than 150 years. We’re not driven by short-term profits. We
don’t have to make our quarterly numbers. We have a very settled and strong culture.”
Where to Begin?
With departmental information spread
across the institution in silos, and with the
significant challenges that building a
data warehouse represents, the question
then becomes
one of where
to begin. It
helps to decide
precisely what
your “enterprise” will be, say the pros,
since at a large university, that can be
many different things. UI’s Walz recommends
starting with questions like:
“‘What is the enterprise at a university
level? Is it the campus? Each college? The
university layer that spans all campuses?’
There are different issues that you have to
face.” Also, he recommends shifting the
initial focus away from the technology
and onto the users. “It’s easy to look at
technology and tools. Instead, ask: ‘Who
are the customers we intend to serve? Are
they administrators, or lower-level staff?’
Understand the audience. Do you want
people to be able to write their own
reports? What needs are you trying to fill?
What problem are you trying to solve?”
In UI’s case, the data warehouse was
planned from the start to fill the needs of
many user types—a solution that Walz
says is far more difficult. “Our customers
span everyone from administrators to secretaries.
Some folks are using the warehouse
to write their own reports. Some
would prefer reports be handed to them.”
Not surprisingly, such huge endeavors
must be driven by senior administrators.
“Get buy-in and direction from
them,” Fish advises. “Everything else
can fall into place, but they have to sell
it at the top.”
WEBEXTRA::The BI vendor lowdown is here.
Stay tuned for Part II of this series on
data warehousing and BI, coming in
November: We’ll look at specific solutions
in use at various institutions.
