So you’ve decided it’s time to implement a business
intelligence (BI) system in your organization, but you’re suddenly faced with
an intimidating variety of options. There are the older, well-established BI vendors and the newer entrants. There are
multi-vendor data warehouse/BI solutions and one-stop-shop packages. There are
the on-premises solutions as well as the cloud solutions. And of course they
all claim to be powerful, flexible, easy to use and affordable.
Hopefully, this article will help you sort through some of
the most important issues in making a BI software selection.
Cost
In almost every situation, the cost of the BI system will be
a major factor, and will probably limit your search to particular segments of
the market. The single most important aspect of cost, however, is to make sure
you understand the true total cost of ownership (TCO) of the solutions you are
considering. In typical BI implementations, the cost of the software is
actually a small fraction of what you will need to pay. This is because most BI
solutions will require expensive integration and report creation services. The
cost of these consulting services almost always eclipse the software licensing
fees. Furthermore, the services required to maintain the reports and analytic
applications (e.g., modifying and extending them to meet the ever-changing
needs of the business users) and create new ones must also be taken into
consideration. You can assume that you will always need more of these services
than you expect.
Note that selecting a cloud-hosted BI solution doesn’t alter
this equation. Cloud BI is just like typical business intelligence in terms of
the types of data integration and report creation/maintenance you will need.
The lack of large up-front hardware and software investments in cloud BI can be
attractive initially, but since the service fees will dramatically exceed the
hardware and software costs of anyway, this should be kept in perspective.
Furthermore, the serious hardware horsepower (namely, CPU and RAM) that
enterprise-scale BI applications demand end up being far more expensive over
time than actually buying the hardware yourself.
Startup Time
Many BI projects are so complex and take so much time to get
off the ground that, in the end, they never launch at all. It is obvious that
the shorter the timeframe involved to go from BI package selection to actual
use in the organization, the greater the chances that the investment will
deliver a positive ROI to the business.
One of the best ways to cut through the BI marketing hype is
to ascertain, in advance of your selection, how long it will take to get the BI
solution up and running in your organization. Take into consideration
purchasing and installing hardware and software, data integration, data
preparation and report/dashboard creation. Ask the BI vendor how long it
typically takes to go from project start to launch – if it looks like weeks (or
months) of work will be required before your users will be able to run a single
report on your actual data, look elsewhere. If the vendor can’t commit to
getting at least one solid report or dashboard running over your own actual data
before you even commit to purchasing the system from them, it’s probably
because they would have to spend weeks or months on development before they can
reach that point. Because there are now BI solutions that can accomplish this
in days (sometimes in hours), you know that solutions requiring longer
development projects just to get started are probably using old technology. You
can do better.
Data Integration and Preparation
One of the keys to a successful BI implementation is how the
corporate data stores are integrated with the BI solution and how the data is
prepared for analytical use. Typically, Business Intelligence software
solutions require the presence of a “data warehouse,” a centralized database
filled with all the business’s data. Do not underestimate the expense and
difficulty of creating an effective data warehouse for those BI solutions which
require one! Historically speaking, data warehouse design and implementation
usually generate lots of income for integrators, but are the source of great
angst for their customers. If you keep in mind that there are excellent enterprise
BI solutions available that allow you to completely sidestep these issues, you
will be doing a lot towards reaching a successful BI selection!
Another element involved in the preparation of data for use
by many BI systems is OLAP (Online Analytical Processing). This technology is
more than two decades old and was designed to improve BI query performance over
large datasets. Like the data warehouse behind it, OLAP is very lengthy and
costly to implement. Most importantly, there is really no need for it anymore. Modern
BI technology can process massive amounts of data without OLAP, at a fraction
of the time, complexity and cost. If OLAP is recommended or required by any BI solution
you are considering, you should probably take it off your list.
Self-Service Report, Dashboard and Analytic Application
Creation
A major dividing line in the BI field is between those BI
solutions which require experts (e.g., integration firms, professional services
contractors) to create and modify reports, dashboard and analytic applications
and those solutions which allow regular business users to do this themselves. Since
BI needs are constantly changing, this is a huge issue.
The advantages of conventional BI systems, namely those
maintained by dedicated (or outsourced) IT staff, is that the full power of the
BI technology is always available. Because the IT professionals are proficient
in every aspect of the software, they can create almost any conceivable report
or application. Furthermore, when end users don’t have to spend the time
learning the software and then maintaining their own reports, they have more
time to spend on their primary job functions.
On the other hand, by selecting a BI solution that allows
any Excel-savvy business user to create (or at least modify/customize) reports,
dashboards, queries and analytics applications him/herself, you have eliminated
the greatest bottleneck (and expense) in the world of BI. The time, cost and
miscommunication involved in depending on outside experts to create and modify
BI applications almost ensures that your business users will be frustrated as
they try to get the system to continue meeting their needs. The greatest competitive
advantage of BI is generated when business users can get any answer or run any
analysis whenever they want.
This decision is clearly an important one, and BI buyers
have to decide how much IT dependence they can tolerate versus how important it
is for end users to be able to create and maintain their own reports and
dashboards.
User Interface
A slick user interface is appealing and BI vendors love to
sell themselves with their advanced reporting and dashboard screens. While the graphical
visualization of data is important, of course, one of the biggest mistakes you
can make is to judge a BI solution based on the fancy sample dashboards and
reports the vendor shows you. The truth is that graphical visualization
software components are commonplace and easy to develop or implement from any
number of software vendors. The real challenge is to quickly create and customizing
reports and dashboards to your own needs. If the BI solution you’re looking at
can’t show you full-fledged reports and dashboards over your own data within
just a few days, you should probably look elsewhere.
Technology
The most mature BI technologies in the market are based on
OLAP, described earlier in this article. While it served its purpose well for
many years, the costs, compromises and inefficiencies involved with OLAP really
make it a poor choice when seeking a new BI solution in the twenty-first
century.
A newer technology which has been getting a lot of attention
in the past few years is called in-memory BI, or in-memory database (IMDB).
IMDBs offer an alternative to resource-intensive data warehouse and OLAP
projects. An IMDB is a database management system designed to provide very high
performance so long as sufficient memory (RAM) is available to hold the entire data
set being processed. Contrast this with conventional Relational Database
Management Systems (RDBMs), which rely instead on much-slower disk-based media.
While IMDBs have become popular during recent years due to the large
addressable memory spaces made possible by now-ubiquitous 64-bit business
computers and operating systems, they still present a hard ceiling when dealing
with large datasets.
Unlike hard disk-based database solutions, for which it is
easy to continuously add more storage at low cost, memory-based solutions
require more and more relatively expensive memory to grow. While 64-bit PCs
theoretically provide a very high maximum memory threshold, in practicality,
deploying the required volumes of memory becomes prohibitive. This is because,
with the in-memory approach, the entire data set must be loaded into memory at
once. When the size of the data (even after compression) exceeds the amount of
RAM, in-memory BI solutions become unusable. Companies with rapidly-growing
data volumes will find that in-memory solutions will soon reach limits which
make them impractical. This becomes substantially more obvious when there are
multiple users accessing the same in-memory store.
Conclusion
It is clear that there are some complex issues involved in
selecting the ideal BI solution for your organization. Perhaps more than in
other fields of software, it behooves you to carefully investigate the many
dimensions involved in making this selection. With this awareness, you will be
better equipped to evaluate both the conventional BI systems available and the
newer solutions, based on more updated technologies.
