Embedding Analytics For Competitive Advantage - Presentation Transcript
Embedding Analytics for
Competitive Advantage
Three Market‐Driven Approaches for
Software Development Companies
A White Paper by Visual Numerics, Inc.
July 2008
Visual Numerics, Inc.
2500 Wilcrest Drive, Suite 200
Houston, TX 77042
USA
www.vni.com
TABLE OF CONTENTS
Embedding Analytics for Competitive Advantage ...................................... 4
Analytical Building Blocks ........................................................................ 4
Alternatives for Embedding Analytics......................................................... 5
In‐House Analytics ................................................................................... 5
Open Source Analytics ............................................................................. 6
Commercial Options ................................................................................ 6
IMSL® Numerical Library Introduction ........................................................ 7
Case Studies ................................................................................................ 8
Solving Specific Numerical Challenges..................................................... 9
Putting Analytics in the Hands of Users................................................... 9
Creating a Custom Solution ................................................................... 10
Other Approaches..................................................................................... 10
Summary ................................................................................................... 10
Embedding Analytics for Competitive Advantage
As customer demand for advanced analytics grows, enterprise software vendors are finding
ways to add new business intelligence (BI) capabilities to their current offerings to stay
competitive.
With embedded analytics, vendors can empower their users to do more with their applications,
enhancing the “stickiness” of their software. For example, nCode International1, a developer of
data analysis software, data acquisition systems and durability solutions, recently added Visual
Numerics statistical tools to their ICE‐flow Glyphworks software. These additional tools enable
customers to analyze and understand durability testing and product usage data from within
nCode’s solutions, as opposed to moving the data to a separate application for analysis.
For businesses, having more analytics embedded in the software solutions that their employees
use on a daily basis delivers significant value. Employees get the decision‐making tools they
need to make good business decisions, when they need them. Organizations that have
employees making better decisions faster gain a significant competitive advantage over
organizations without similar abilities.
This white paper is intended for software vendors that want to make analytics an embedded
part of their products. It will discuss analytical building blocks, alternatives for building
solutions, and through case study examples, it will outline three approaches taken by vendors
to embed analytics into their software applications.
Analytical Building Blocks
In a cover story discussing analytics in business, “Math will Rock Your World2” (Business Week,
January 2006), the author writes about “imposing the efficiencies of math” on finance, science,
engineering and business problems. This article and others that followed outlined the many
ways organizations and industries could leverage numerical analysis to solve problems and
make decisions analytically.
As discussed in the Business Week article, the building blocks for analytical applications are
mathematical and statistical techniques. Math functions such as interpolation and
optimization, and statistics functions such as correlation, regression or random number
generation help organizations understand data, simulate situations by asking “what‐if”
questions, and make fact‐based decisions.
A key decision for software vendors that want to make analytics part of their application is how
to embed the mathematics and statistics that enables analytics. Three alternatives are
discussed in the next section.
1
http://www.ncode.com/productDetails.aspx?display=details&product_id=67
2
http://www.businessweek.com/magazine/content/06_04/b3968001.htm
Page 4
Alternatives for Embedding Analytics
Software vendors have many alternatives for adding analytic functionality to their applications.
Three commonly considered alternatives are:
• Build math and/or statistical functions for their application using in‐house development
resources
• Use “free” open source options
• Leverage a commercial option
The following sections will explore some advantages and disadvantages of each.
InHouse Analytics
Many software vendors first consider developing analytics functionality in‐house. Because of
the nature of their business, software vendors employ intelligent programmers who can write
the code for math and statistical algorithms. Publications such as the Numerical Recipes3
reference books contain complete commented source code in several languages for hundreds
of algorithms. Leveraging Numerical Recipes or a similar source of algorithms can be as easy as
copying the source into an application and compiling the code. Once complete, the in‐house
developed algorithm is owned by the software vendor and is an embedded part of the
application.
For novel or cutting‐edge algorithms, the development process is considerably more in‐depth
than simply consulting existing references of source code. Working from a journal article
outlining the mathematics involved in an analytical solution can be a considerable challenge for
even seasoned software developers if the goal is to create commercially viable and usable
general source code. In addition, such development is typically serial in nature meaning that
adding more resources will not shorten the development cycle. If time‐to‐market is a
consideration, then the time required to develop in‐house solutions may be a costly drawback.
While in‐house development may seem easy, there are both direct and ongoing costs to
consider. There is the initial, direct cost of the development time to write, test and document
the algorithm. For simple algorithms, this might be a quick process. For complex algorithms or
custom solutions, development can take much longer – months or even years in some cases. In
addition, there are ongoing costs of maintaining algorithms, including upgrading code for new
hardware or software platforms, potentially migrating code to new environments, and making
updates as necessary. Maintenance costs can easily run twice the cost and time of initial
development.
Aside from development costs, an important consideration when weighing whether to develop
analytical algorithms in‐house is how to best deploy resources. Asking the question “should our
internal development staff be used to write math or statistical functions or should they work on
3
http://www.nr.com/
Page 5
our core application” can help a software development company decide if it is better to create
in‐house analytics or leverage open source or a commercially available option.
Open Source Analytics
To avoid the cost of writing and supporting algorithms developed in‐house, software vendors
might look to one of the thousands of open source options, such as the GNU Scientific Library4.
The commonly cited attraction of open source is, of course, that it is free and supported by the
general development community.
For software vendors looking to add analytics to their applications, open source has potential
drawbacks. Since many open source options come with third‐party obligations or viral type
license agreements, some customers will simply not accept software that contains open source
code to avoid dealing with multiple companies and organizations for support. Additionally,
utilizing open source codes in a commercial application may require the entire application’s
source to be made available, which is not an option for many software vendors.
For developers, not all open source projects are easy‐to‐use. Many are sparsely documented,
occasionally in foreign languages, and contain few, if any, examples. Other projects are
available only in source code form, which means the onus is on the user to compile necessary
binaries. When an open source project designed for Linux is required on Windows, this
compiling step can be a daunting task. Projects rarely have resources available to address
technical support questions. Users are left to email busy developers or visit unresponsive
forums. Finally, open source projects often stagnate after initial activity. Once the original
developers get it working for their specific problem, there is no impetus for continuing
development.
While there are many examples of successful open source projects, software vendors should
consider the risks before pursuing open source options for their applications.
Commercial Options
In computer science, a library is a collection of subprograms used to develop software. Libraries
contain \"helper\" code and data that provide services to independent programs. Visual
Numerics is one of several vendors that offer a commercial library option for embedding
analytics. Visual Numerics IMSL Numerical Libraries5 provide numerical computation code in
the form of math and statistical algorithms.
The intent of a commercial option like the IMSL Libraries is to empower software developers to
add analytical capabilities to their application, but avoid re‐inventing the wheel through writing
complex math or statistical routines, or introducing risk by using untested open source software
in their application.
4
http://www.gnu.org/software/gsl/
5
http://www.vni.com/products/imsl/
Page 6
A commercial option also offers extensive numerical algorithm development expertise. Visual
Numerics focuses on the development, sales and maintenance of analytical algorithms,
supporting a wide range of languages and platforms, and has years of experience working with
thousands of customers. A commercial option like the IMSL Libraries therefore provides a wide
breadth of available algorithms, significant depth of capabilities, and extensive testing by the
vendor and a large and diverse user community.
Table 1 summarizes the key advantages and disadvantages of these alternatives for embedding
analytics into a software application.
Table 1. Embedded Analytic Alternatives Summary
Key Advantages Key Disadvantages
Source code examples Time to code, test,
are often available document and maintain
Analytic functionality algorithms
In‐house developed is owned by Opportunity cost of
analytics the software vendor developers writing
algorithms instead of core
software application
Many algorithms are Third‐party obligations and
freely available viral‐type license
Open source agreements are common
analytics Documentation, testing
and support are often
minimal
Depth and breadth of Cost to license
available algorithms
Reduce development
Commercial time through the use of
options pre‐built algorithms
Reduce risk through
commercially tested and
widely used algorithms
IMSL® Numerical Library Introduction
The IMSL Libraries are a collection of mathematical and statistical algorithms that developers
can integrate into their applications. IMSL Libraries cover a wide range of functions ranging
from linear algebra and curve fitting to numerical optimization in the math domain and time
series analysis and data mining functions in the statistical domain.
IMSL Numerical Libraries are implemented in the widely used computer programming
languages of C, Java, C#/.NET, and Fortran. Software developers embed algorithms from these
Page 7
libraries into their software applications using their preferred programming language. The first
IMSL Library for the Fortran language was released in 1971, followed by a C language version
originally called C/Base in 1991, a Java language version in 2002 and most recently a C#
language version in 2004. Over time, the libraries have grown in supported languages,
supported platforms, and in depth and breadth of available math and statistics algorithms.
Table 2 outlines general math and statistics areas found in the IMSL Numerical Libraries. These
areas illustrate a broad and full‐featured spectrum of critical analytic capabilities that can be
used to build advanced numerical software applications in many fields.
Table 2. IMSL Numerical Library Math & Statistics Areas
Mathematics Statistics
Matrix Operations Basic Statistics
Linear Algebra Time Series & Forecasting
Eigensystems Nonparametric Tests
Interpolation & Approximation Correlation & Covariance
Quadrature Data Mining
Differential Equations Regression
Nonlinear Equations Analysis of Variance
Optimization Transforms
Special Functions Goodness of Fit
Finance & Bond Calculations Distribution Functions
Random Number Generation
Neural Networks
The items listed in the table represent entire areas of functionality with numerous algorithms
within each area. Within the IMSL Libraries, the actual count of available math and statistical
algorithms runs into the thousands, giving developers many options to mix and match
algorithms as needed to create unique and competitively engineered analytical applications.
How to use numerical algorithms in a software application depends on the type of application,
the needs of the users of the application, and the desires of the application vendor. Some
software vendors might add a few algorithms that provide specific functionality; others might
want to offer a broad selection of numerical analysis functions; while others might need highly
customized functionality. The next section outlines three recent approaches by software
vendors to embed analytics into their software applications.
Case Studies
The following sections highlight examples of customers that have embedded IMSL Library
functions into their software applications.
Page 8
Solving Specific Numerical Challenges
Many companies need analytics to enable their software application to solve a particular type
of problem. An example is Moore Nanotechnology Systems6, a leading manufacturer of ultra‐
precision machining systems. Customers in defense, aerospace, consumer electronics, and
ophthalmic industries use Moore Nanotech’s ultra‐precision machining equipment to create 3D
freeform surfaces such as optical lenses, mirrors and inserts. With their machining equipment,
Moore provides NanoCAM, a full 3D CAM software that enables customers to define the
precise surface to be machined using their ultra‐precision tools.
IMSL Library functions are embedded into the NanoCAM application. Math algorithms,
including spline definition and interpolation functions, enable the accurate definition of 3D
freeform surfaces. Accurate definition is critical for ensuring quality output from the
manufacturing process – typically in nanometer range. The IMSL Library functions embedded
into NanoCAM also provide the fast computation capability required by the application.
Manufacturers who need to be completely precise with the surfaces commonly interpolate
point data sets ranging from 300,000 to 1,000,000 points. The ability of IMSL Library functions
to process these large datasets in a fraction of a second speeds up the design and analysis
process.
For Moore Nanotech developers, in‐depth analysis functions from the IMSL Libraries were
easily embedded in the application, saving months of development time versus trying to write
the functions from scratch.
Putting Analytics in the Hands of Users
For many software applications, vendors can deliver value by providing a complete set of
analysis tools for customers. Customers can then use these tools in unique ways on their own
data and make business decisions to create competitive advantages.
An example is Teradata7, a provider of enterprise data warehousing and business intelligence
solutions. To provide users with a wide range of embedded statistical functions for in‐database
analysis, Teradata embedded multiple IMSL Library probability distribution functions and
inverses and random number generators into a new product, the Teradata Statistical Library
Add‐in, Powered by IMSL, part of Teradata Warehouse Miner.
These statistical functions are provided as embedded user defined functions (UDFs) accessible
through Teradata Warehouse Miner or standard SQL Queries. Teradata Warehouse Miner’s
UDF capabilities work like an index from which users can select prepackaged functions which
are linked to the detailed data. From a desktop environment, business users can click, drag and
drop the needed analytic function from a list of UDFs and run it against the appropriate
detailed data in the warehouse without wasting time by manually coding the function or
moving data between systems. Combined with other analytic functions in Teradata Warehouse
6
http://www.nanotechsys.com/
7
http://www.teradata.com/
Page 9
Miner, the Statistical Library functions, Powered by IMSL can be used in a variety of
applications, including classification, forecasting and others.
Creating a Custom Solution
Some software applications can benefit from customized analytics. An example is Teksouth8, a
privately held firm headquartered in Birmingham, Alabama. The company has served both the
commercial markets and Federal government for more than a generation. Teksouth products
and services include data warehouse development, implementation, and support; web‐based
decision support tools; secure high‐speed data transport tools, as well as system engineering,
integration, and program management.
To enable private sector and government organizations to more accurately project and budget
spending changes, and help companies address accountability and regulatory compliance
issues, Teksouth, working with Visual Numerics Consulting Services, developed an innovative,
customized neural network forecasting application for its customers based on neural network
forecasting functionality in the IMSL Library.
The application extracts information from historical cost data, applying the information to
forecast future costs. The analytic engine has the flexibility to prioritize historical data and
exclude anomalous data to improve its predictive capabilities. One of the most powerful
features of this solution is its ability to mimic the brain's own problem‐solving process by
applying knowledge gained from historical data to new problems, fine‐tuning its forecasting
accuracy even more over time.
Many organizations evolve their predictive analytic capabilities over time. Visual Numerics
Consulting Services can help organizations like Teksouth get to the next stage of analytical
maturity by identifying and then helping to implement methods to bring more sophisticated
analysis into the organization.
Other Approaches
The previous section highlighted three examples of how software vendors can embed advanced
analytics into their applications. With more than 3000 functions across four native development
languages on dozens of hardware platforms and operating systems, the ways that advanced
analytics with the IMSL Numerical Libraries can be embedded into software are virtually
limitless.
Summary
Software vendors can help their customers solve problems and make decisions analytically by
embedding analytic functionality into their applications. While there are alternatives to
embedding commercially‐developed and supported analytical functions: building in‐house or
8
http://www.teksouth.com
Page 10
leveraging open source; commercial libraries deliver the optimum approach for many vendors
because of their:
• Breadth and depth of numerical algorithms
• Broad platform and language support
• Proven reliability
• Technical and consulting support
Analytics can be embedded in many ways. Depending on the software vendor, application and
customer needs, the best approach will vary. Regardless of approach, the end goal remains the
same – to give customers the ability to create competitive advantage by making decisions
analytically.
Visual Numerics is an established provider of commercial libraries. Math and statistical
functions in the IMSL Libraries provide analytic capabilities to many software vendors today and
offer unique value, including:
Fast time to market – Software companies like nCode International and Moore
Nanotechnology brought products to market significantly faster using pre‐built IMSL
Libraries as opposed to developing functionality in‐house.
Execution & Efficiency – Analyzing data at the source is often more efficient than
moving it outside the database. Teradata provides IMSL Library statistical functions to
users as embedded user defined functions (UDFs) accessible through Teradata
Warehouse Miner or standard SQL Queries. IMSL Libraries, available in C, Fortran, .NET
and Java make it easy to embed analytics in the same native language in which an
application is written.
Customization – Working with Visual Numerics Consulting Services, Teksouth was able
to customize advanced neural network technology to develop a unique budget
forecasting application. The ability to customize algorithms or simply mix and match
algorithms from the thousands available, IMSL Libraries enable developers to create
unique and competitively engineered analytical applications.
Finally, IMSL Libraries ensure numerical consistency across different languages and platforms.
Developers do not have to hard code libraries to fit each application environment. This
advantage is particularly helpful when software companies need to upgrade or migrate their
applications to new hardware or software platforms.
To learn more:
• Visual Numerics IMSL Numerical Libraries9 product information
• Visual Numerics OEM/ISV partner10 information
9
http://www.vni.com/products/imsl/index.php
10
http://www.vni.com/company/partners/oemIsvPartners.php
Page 11
0 comments
Post a comment