UNIT V FUTURE OF BUSINESS INTELLIGENCE 9

Future of business intelligence – Emerging Technologies, Machine

Learning, Predicting the Future, BI Search & Text Analytics –
Advanced Visualization – Rich Report, Future beyond Technology.

The Future of Business Intelligence

The future of business intelligence centers on making BI relevant for
everyone, not only for information workers and internal employees, but also
beyond corporate boundaries, to extend the reach of BI to custom-ers and
suppliers. As the Successful BI case studies have demonstrated, when best
practices are applied, BI usage can explode beyond the paltry 25% of
employees today to a much more prevalent business tool. It will take cultural
shifts, new ways of thinking, and continued techni-cal innovation. Business
intelligence has the power to change people’s way of working, to enable
businesses to compete more effectively and efficiently, and to help
nonprofits stretch their dollars further. All of this is possible based on
insights available at the click of a mouse, push of a button, or touch of a
screen.

As discussed throughout this book, much of the key to successful

business intelligence has to do with the people, processes, and culture. Don’t
rely on technical innovation alone to solve the biggest barriers to BI success,
but by all means, do get excited about the innovations that will make BI
easier and more prevalent. BI as a technology has changed dramatically
since its inception in the early 1990s. This chapter focuses on the most
recent technical innovations with examples of how custom-ers are taking
advantage of them. In the final section, I leave you with some words of
wisdom to inspire you to think about how business intel-ligence can become
your company’s next killer application.
Emerging Technologies
As part of the Successful BI Survey, respondents were asked to choose items
from a list of emerging technologies that they believe will help their
companies achieve greater success. Figure 1 shows which items
Figure 1 The majority of survey respondents believe web-based BI, dash-boards,
alerting, and predictive analytics will allow greater success.(Results Based On 513 Responses)

are considered most important in helping companies achieve greater success.

Web-based business intelligence and dashboards were rated the highest, with
predictive analytics and alerting also at the top. Surprising to me, Microsoft
Office Integration, BI Search, and Mobile BI were selected by only a small
percentage of survey respondents.

The view according to business users, however, is slightly differ-ent, as

shown in Figure 2. Business users account for only 10% of the survey
respondents. Those who describe themselves as hybrid business-IT
personnel account for 23% of respondents. I have specifi-cally excluded IT
personnel and hybrids from Figure 2, to show the gap in perceived
importance of certain technologies. When viewing responses only for
business users, the importance of Microsoft Office integration moves to the
top of the list, while alerting moves down.

Some of these differences can be explained by gaps in understand-ing of

the feature benefits, but also by a respondent’s point of view. For example,
IT professionals have been burned in the past by the thousands of
disconnected spreadsheets and the ensuing data chaos. As Microsoft Office
integration with BI has improved dramatically in 2007, IT pro-fessionals
may not realize that spreadsheet-based analysis can now be ―safely‖ enabled
and can be something to be embraced for knowledge workers familiar not
just with Microsoft Excel, but also with Word, Outlook, and PowerPoint. In
a similar fashion, if you are a BlackBerry user, you may rate Mobile BI high.
While web-based business intelli-gence may have been introduced in the late
Figure 2 For business users, Microsoft Office integration with BI moves up in
importance. (Results based on 51 responses.)

1990s, these solutions only reached the rich functionality of their desktop
predecessors in the 2005 time frame. A number of companies are not yet on
the latest releases, though, and still use client/server BI deployments.
Depending upon where a survey respondent is in their web-based BI
deployment will influence how this capability was rated.

At The Data Warehousing Institute’s (TDWI) Executive Summit in

February 2007, I participated in a panel on the role of emerging tech-
nologies in extending the reach and impact of business intelligence.
Attendance was restricted to BI directors and executive sponsors who
influence their company’s BI strategy. Attendees could vote on a limited
number of items that they thought would have the biggest impact in the next
few years. The most highly ranked item: performance management and
predictive analytics. The things that got few to no votes were BI search,
dashboards, and rich Internet applications, contrary to what I believe will
have the biggest impact. As we delved into what these tech-nologies mean,
and in some cases, demonstrated them, the perceptions changed
significantly. In this way, we sometimes don’t know the impact any of these
capabilities will have until the technology has become more mature and the
industry understands it better. If you think about the way breakthroughs like
the iPod and YouTube have revolutionized their markets, when they first
were introduced, they were met with a mixture of fascination and confusion,
without a clear understanding of where they would lead. Recent BI
innovations must go through a simi-lar process of the industry first
understanding their potential, accepting or rejecting them, and then either
embracing or adopting the innova-tions in a limited fashion. Another
fundamental difference in evaluat-ing impact is whether the impact is
measured by the value of one user or decision or aggregated by multiple
users and decisions. Recall from Chapter 12 (see Figure 12-4), thousands of
individual decisions can have as big, if not bigger, an impact on a company
than a single decision. There seems to be a natural tendency to rate
capabilities that have a single big impact as being more important or more
likely to help achieve greater BI success.

Figure 3 provides a framework for evaluating changes in BI technology

to determine which new and emerging capabilities will prove most valuable
to your company, how mature they are, and when to monitor them or when
to embrace and actively deploy them (adapted from TDWI’s Technology
Evaluation Framework). The X axis provides an indication of how mature
the technology is, and the Y axis gives an indication of which technology
will make BI pervasive. Recall from the section on BI users as a percentage
of employees in Chapter 4 that the average usage of BI within a company is
currently at 25%, and even if budget were available and the deployment
were wildly successful, survey respondents felt the use rate would extend
only to 54% of employees.

Evaluate & Test Embrace

Mainstream
Embedded BI Web-Based
Performance
Management BI Suites

Rich Reportlets MS Office

Integration
B Rea
I ch
BI Search Dashboards

Text Analytics Predictive Analytics

Appliances Scorecards
Mobile BI

BI Software Ell Advanced

as a Service Visualization
Alerting

Niche
Monitor & Understand Adopt Where Appropriate

Emerging Maturity Robust

Figure 3 BI technology evaluation

The Y axis, then, indicates the degree to which an enabling technology will
take BI’s reach closer to 100% of employees. Business impact and BI
prevalence are not linearly correlated, however. One enabling technology,
such as predictive analytics, may yield a big value for a single decision, say,
a $4 million savings by better marketing campaign management. Another
enabling technology such as BI embedded in operational processes may
affect thousands of users, each of whom makes dozens of decisions on a
daily basis; the monetary value of these individual decisions may be small
when measured in isolation, but enormous when taken in aggregate. The size
and shading of the bubbles in Figure 3 give an indication of which items
have a bigger single value. The bigger the bubble and darker the shading, the
bigger the impact on a single decision or person.
For each innovation, consider both the technical maturity and the
business impact to decide how to proceed:

■
Embrace Items in the upper-right quadrant show innovations that are
mature and that should be embraced as they will help speed user adoption
across multiple user segments.
■
Adopt Where Appropriate Items in the lower-right quadrant show
innovations that are mature but that may serve only specific segments of
users. Mobile BI is an example of this; the technology is more mature than
BI search, for example, but benefits only those users who have
smartphones such as a BlackBerry.
■
Evaluate and Test Items in the upper-left quadrant are relatively new but
will have a profound impact on user adoption. BI Search is a good
example of this. The technology is very new and not well under-stood. A
number of usability and performance issues still need to be worked out,
but the potential impact on user adoption is enormous.
■
Monitor and Understand Items in the lower-left quadrant are so new that
they may be riskier investments. Items here are less proven and have less
market adoption.

Figure 14-3 portrays broad industry maturity of these capabilities and

the degree to which most vendors offer the capabilities. For clar-ity, I have
selected only certain innovations; it is not meant to be an exhaustive list of
all things going on in the industry (for this, check the BIScorecard blog). All
items are in the context of business intel-ligence as a technology. So while
performance management is certainly a mature concept and technology, the
integration of performance man-agement with business intelligence is still a
work in progress, leaving this item positioned slightly behind web-based BI
and Microsoft Office integration on the maturity spectrum.

Predicting the Future

Data mining, statistical analysis, and predictive analytics are nothing new.
These technologies are well established and are used in a number of
different applications such as fraud detection, customer scoring, risk
analysis, and campaign management. What’s changed is how they have
become integrated in the BI platform. Traditionally, predictive analytics has
been a backroom task performed by a limited few statisticians who would
take a snapshot of the data (either from a data warehouse or from a purpose-
built extract from the source system), build a model, test a model, finalize it,
and then somehow disseminate the results. While the expertise to build such
models remains a unique skill set, the indus-try recognizes that the results of
the analysis should be more broadly shared, not as a stand-alone application,
but rather, as an integral part of the BI solution. This does not mean that
predictive analytics software will become ―mainstream,‖ but rather that the
results of such analyses can be readily incorporated into everyday reports
and decision making. The analysis, then, is what needs to become
mainstream.

Predictive analytic tools from different vendors do continue to differ

significantly in how they work and in what information is stored in the
database versus calculated and presented in a report or incorporated into an
operational process.
At Corporate Express, for example, predictive analytics are being used
1
to improve customers’ online shopping experience.

Market basket analysis helps retailers understand which prod-ucts sell

together and provide product recommendations. In the past, Corporate
Express provided these recommendations by logical product pairings. So if a
customer ordered a stapler, the online store would rec-ommend a staple
remover as the marketing team had marked this as a complementary
2
product.
In analyzing the data, though, it turned out that what was most often
purchased with a stapler was not a staple remover, but rather a ruler, tape
dispenser, and a wastepaper basket—items that indicate a purchase for a new
employee. With the manually associated product recommendations, there
was no significant impact on sales. Leveraging MicroStrategy and SPSS,
Corporate Express tested a new market basket option. The model analyzes
past shopping carts and produces
recommendations to ensure the greatest lift. As a result, the average order
size for market basket pairings doubled (versus those orders with no
pairings), and the market basket application is expected to generate an
incremental gross profit of more than $2 million in 2007.

Dow Chemical also has begun extending the reach of predictive

analytics with SAS’s JMP product (pronounced ―jump‖), a solution that
combines visual analysis with statistics. As discussed in Chapter 12, Dow
uses BusinessObjects and Cognos PowerPlay as enterprise report-ing and
analysis standards. Through these tools and the data in the data warehouse,
Dow began looking at the high cost of railroad shipments: $400 million
3
annually across North America.

A team of statistical experts studied the variables that most affected

these costs and pulled data from the data warehouse and external data
sources into SAS JMP. By benchmarking current payments versus industry
norms, the analysis showed Dow was overpaying by 20%, or $80 million. In
entering new contracts, the purchasing department now uses the software to
predict appropriate rates, enabling them to negoti-ate more aggressively.

For both Corporate Express and Dow Chemical, the move to predic-tive
analytics has been evolutionary. The underlying information archi-tecture
and a culture of fact-based decision making had to first reach a level of
maturity and data quality before predictive analytics could be embraced.
While both companies have been doing statistical analysis for decades, the
degree to which predictive analytics has now been incorporated into daily
processes (online store at Corporate Express and purchasing negotiations at
Dow Chemical) reflects the degree to which predictive analytics has shifted
from the backroom to the front line, with the most casual of users deriving
value from such analytics.

BI Search & Text Analytics

BI Search offers a number of promising benefits to business intelligence:

■
Simple user interface.
■
A more complete set of information to support decision making, with the
integration of structured (quantitative) and unstructured content (textual).
Structured data refers to the numerical values typically cap-tured in the
operational systems and subsequently stored in a data warehouse.
Unstructured content refers to information stored in tex-tual comment
fields, documents, annual reports, websites, and so on.
■
Users can find what they need through search, rather than through
navigating a long list of reports.

Text analytics is closely related to search in that unstructured infor-

mation or text can be transformed into quantitative data. For example, it
allows for searching of information in a comment field to see how many
times a customer praised a particular product. Text analytics is the numerical
analysis of textual information.
Despite all the improvements in data warehousing and BI front-end
tools, users continue to feel overwhelmed with reports yet undersatis-fied
with meaningful information. They don’t know what’s available or where.
Similar reports are created over and over because users don’t know which
reports already exist or how, for example, the report ―Product Sales‖ differs
from ―Product Sales YTD.‖ Some of the most valuable information is
hidden in textual data.
A BI Search interface promises to change the way users access
information. Picture a Google interface to BI. Without any training in a BI
tool, users can enter a phrase such as ―Recent sales for customer A‖ and then
be presented with either a list of predefined reports or, in some cases, a
newly generated query. The added benefit is that in addi-tion to displaying
reports coming from the BI server, the search engine will also list textual
information that may be relevant—a customer letter, sales call notes,
headline news. When search capabilities are combined with text analytics, a
report may include numerical data that scans the comment field to indicate
number of complaints with number of posi-tive comments. Never before has
such unstructured data been so nicely accessible with structured or
quantitative data.
If the integration of search and BI is successful, it is yet another
innovation that will make BI accessible and usable by every employee in an
organization. According to Tony Byrne, founder/president of CMS Watch, a
technology evaluation firm focusing on enterprise search and content
management systems, search as a technology has existed for more than 50
4
years. Consumer search (Google and Yahoo, for exam-ple) as a technology
emerged with the Internet in the mid-1990s. In many respects, the success of
consumer search has helped spur hype around enterprise search, in which
companies deploy search technology internally to search myriad document
repositories. Text analytics has existed for 25 years but with usage in limited
sectors, particularly, the government. The convergence of search with
business intelligence first emerged in 2006. Google is not the only enterprise
search solution that BI vendors support but it is one that has the most
consumer recognition and thus has helped business users to understand the
possibilities. To illustrate the point, note that BI search was selected by only
27% of the Successful BI Survey respondents as a capability that would help
foster greater success (see earlier Figure 14-1). Yet in discussing these
technol-ogies with individual executives who don’t currently use business
intel-ligence, a Google-like interface to BI generated the most enthusiasm.

The incorporation of text analytics with traditional business intel-ligence

is still in its infancy. I place both BI search and text analytics close to the
Monitor and Understand quadrant but in the Evaluate and Test quadrant in
Figure 3. Again, both technologies independent of BI have existed for
decades; it is that convergence with BI that is new. While the convergence is
still relatively immature, the promise it brings for BI to reach more users and
in the value of incorporating textual data is enormous.
 The number of customers taking advantage of the BI Search and text
analytics integration is only a handful. BlueCross BlueShield (BCBS) of
5
Tennessee (TN) is an early adopter of these capabilities. BCBS of TN is a
not-for-profit provider of health insurance. In 2006, it paid $17 billion in
6
benefits for its 2 million commercial members. Managing claims and
negotiating rates with providers is critical in ensuring BCBS can meet its
obligations to the members it insures. While the insurer has had a mature
7
business intelligence deployment for ten years, Frank Brooks, the senior
manager of data resource management and chief data archi-tect, recognized
that there was value in bringing the text data stored in comment fields from
8
call center notes together with information in the data warehouse. Given
how new the technology is, Brooks asked their BI vendor, Cognos, along
with IBM (who produces the search solution OmniFind) and SAS (who
offers text analytics solution Text Miner) to work together to develop several
prototypes and show the business users the concept of bringing BI,
enterprise search, and text analytics together. With this capability, a business
user can enter the key word ―diabetes‖ in the OmniFind search box and be
presented with a ranked list of things such as:

■
Cognos reports and OLAP cubes that show claims paid for diabetic
treatments
■
Call center notes that involve diabetes
■
New research on improving care for diabetes patients

The business was enthusiastic. There has been a high degree of col-
laboration between BCBS of TN and its information technology partners
to understand the new capabilities, develop the right infrastructure, and
optimize the indexes to provide the best search performance.

Consistent with the evaluation framework in Figure 3, under-standing

new technologies requires a significant amount of evaluation and testing.
BCBS of TN evaluated the capabilities for more than a year before
developing plans for implementing in production.

Advanced Visualization
All leading BI tools have basic visualization capabilities: you can take
tabular data and turn it into a bar chart, trend line, and so on. They also
support some kind of conditional formatting of data: display positive
numbers in green, display negative numbers in red, and enlarge those with
the worst variances. Advanced visualization goes beyond a simple chart such
as a bar or line chart to include things as:
■
Spark lines, a highly condensed trend line the size of a word.
■
Bullet graphs, a construct by Stephen Few that includes a target indi-cator
within the bar chart.
9
■ Small multiples, which are series of small, similar graphics or charts. As
they use the same scale and are positioned side-by-side, they facili-tate
visual discovery by letting users make comparisons at a glance.
■
Heat maps that display two variables as different intensifying colors.
■
Decomposition trees, a visualization that displays each drill-down akin to
an ever-expanding organization chart.
■
Geographic maps that display things such as sales figures in a map form,
using color to highlight sales performance. By mousing over a particular
country, region, or state, you can see the individual data values.

Advanced visualization software and capabilities also help you apply

best practices in data visualization, even for basic visualizations. As an
example, many reports today are designed as a dense page of numbers. The
dense page of numbers may not help facilitate insight, but they are what
users are accustomed to. With visualization software, a dense set of num-
bers can quickly be converted to a more meaningful display.

Figure 4 Advanced visualization with Tableau Software

 Figure 4 shows several charts created in Tableau Software. By
displaying multiple graphs side-by-side, as ―small multiples,‖ you can more
easily see which product category—technology in this case—is the top-
selling product category for all regions. The dimensions (time and region)
and scale within each graph are the same, allowing for a rapid comparison.
Sales seem to have dipped in 2003, particularly in the eastern region. By
toggling the quick filters (shown on the right in Figure 4), it’s possible to
focus on the individual customer segments to see that technology sales to
cor-porations are on a steady decline, whereas consumer and small business
segments show strength.
Creating this kind of display with standard BI software is theo-retically
possible, but one that would take many, many more steps. As well, if I am
uncertain as to the best way to display the information, advanced
visualization software can make suggestions. The capability to create easily
such advanced visualizations is generally not available in BI suites. Users
must rely on specialty products. In a theme similar to predictive analytics
and search, visualization software has existed for years; the change is in its
convergence with business intelligence such that advanced visualizations are
appearing in dashboards and reports. In this regard, the emphasis for BI tools
is changing from a focus of simply ―getting to the data‖ to ―what insights
can I discover from the data and how can the most information be displayed
in the smallest space.‖
Rich Reportlets
Report-based interactivity is a ho-hum term that warrants a better name.
―Active reports,‖ ―on-report formatting,‖ and ―navigable reports‖ are similar
terms that also don’t fully capture the value of this capabil-ity. I suspect poor
terminology and lack of awareness also explains why survey respondents
ranked this capability on the low end of importance for emerging
technologies (see Figure 1). So after much thought and brainstorming with
some colleagues, I will refer to this capability as ―rich reportlets.‖ The
difference in power and appeal with rich report-lets versus, say, green-bar
paper reports and much of what is currently deployed over the Web, is
comparable to the difference between a black Ford Model T and a red
Mercedes sports coupe.
Rich reportlets are powered by Web 2.0 technologies to create rich
Internet applications (RIA). When BI suites were first re-architected for the
Web, report consumers could only view a static page. Given how static a
display this was, more sophisticated users would export the data to Excel for
analysis. Less sophisticated users would submit requests to IT or to the BI
team to modify the report design. The Web in this case is only a delivery
vehicle for data; it does not facilitate user adoption
and insight. With rich reportlets, someone accesses a report over the Web
but in a much more interactive and appealing way. At a simple click, data
can be re-sorted, filtered, or graphed, without having to launch a
complicated report editor. With the use of either Adobe Flex or Macromedia
Flash, these reports come to life in ways that make business intelligence fun.
I have seen, for example, a bubble chart that displays bubbles dancing across
the screen as the time axis marches onward. Such animation makes BI
appealing as well as insightful as users see the trend in action. In this regard,
the term ―report‖ doesn’t do justice to the capability that is more akin to a
mini application.
This type of interactivity affects all BI users, whether casual or power
users. The appeal makes BI more engaging, and while some technologists
may scoff at the importance of this, when other barri-ers to adoption exist,
appeal matters. A lot! The ability to interact with the data in a simple and
intuitive way facilitates greater insight at the hands of the decision maker.
The report consumer is not forced to delay this insight until a power user can
modify the report. Lastly, the cost of ownership is lowered because a single
reportlet can be ―tweaked‖ to that decision maker’s needs, without IT having
to maintain thousands of individualized reports.

The Future Beyond Technology

Technical innovation is only one aspect that will help increase BI’s
prevalence. In discussing future plans with many of the case study com-
panies, much of their concern was not about technology, but rather, in
finding new ways to use BI to address common business problems. For the
more large-scale deployments, some expressed concern about man-aging the
risk of making any kind of major change to such a business critical, complex
application. With success, of course, comes greater demands on the systems
and the people. Ensuring an effective way of prioritizing competing requests
warrants constant attention. One busi-ness leader expressed frustration at his
department’s inability to make wise investments, while witnessing other
departments, working in more unison and getting more value from business
intelligence. Yet he remains optimistic that his business will get there and
that BI will be the first thing people look at, even before email. ―To have one
screen I can get to with a single click, that shows sales, margin, price,
opportunities in graphical form, with drill down—that would be magic!‖ His
comments remind me that the technology is sometimes the easy part; getting
the organization aligned is harder. Even the most successful BI companies,
then, continue to have their battles.