BI Best Practices
Adopting best practices in business intelligence give you a better
chance of success with your own BI deployment. There are certain
imperatives that apply regardless of what type of business you have,
what type of BI solution
you select or how big your BI implementation is.
Prepare Well Ahead
Customer-facing programs and
monitoring of critical business processes constantly demands
fresh data. Regardless of what BI
tools you deploy, once users recognise the value of reliable,
consistent data they make demands for data latency to be reduced
Reaching real-time BI requires implemention of real-time technologies.
The key is to plan ahead of demand and stay ahead. Making
the right BI technology choices up front of demand is the only
way you will keep ahead of business user requests.
Planning a 'services'
approach for real-time BI infrastructure up front means that
data loads can be automatically managed by continuously monitoring
all warehouse processes and respond rapidly to issues - such as
data loading failures and long query response times. Bureau automates
all processes all the time, regardless of how often they run.
Know Where Real Time Data Is Really Needed.
It is easy for business users to demand real time data, but often
the value given does not validate the additional investment to deliver
data beyond daily or 2 hourly updates.
- Real-time data feeds
require more infrastructure and are more difficult to manage.
- Real-time processes must be constantly monitored. Issues arise
within business hours, rather than during batch laods done overnight.
This requires fast response support teams, putting further pressure
on support staff.
- Additional hardware and capacity is required to store the data.
Generally, each real-time feed requires two servers, one to run
the load and a second to back it up.
- Real-time data feeds from some source systems can be expensive
and even impossible to implement.
Justification for real time data is a critical part of your BI
Educate the Business
Many business users either don't have the insight or the time to
explore what is possible with real-time BI. A common mistake is
for business users to want new systems to replicate what the old
ones did - just without the bugs or with a few additional cells.
By developing prototypes of potential new BI systems that would
manage hub operations better, users realise that technology can
help them improve the way they manage their business. The biggest
challenge becomes finding the resources to support the ideas that
Align Decision-making and Business Processes
There are three sources of latency in real-time BI:
- Time required to extract data from source systems
- Tme required to analyze the data
- Tme required to act upon the data
The first two are easily dealt with using real-time technologies.
The third, the most problematic is getting people and processes
to change. If you fail to ensure that downstream decision-making
and business processes efficiently utilize real-time data, the value
of having real time data decreases.
Co-existing Strategic and Tactical Decision Support
Traditionally, data warehouses have focused on supporting strategic
decisions. Operational decision making was supported by operational
Real-time technologies moves decision making to the tactical level,
in an 'operational BI' model. In a BI environment, strategic and
tactical decision support co-exist in the same warehouse environment.
However, the performance requirements for each differ:
- Strategic decision support - analysis of large
amounts of data at less frequent periods, but requiring high processing
- Tactical decision support - repeated access
to only a limited amount of data, in real time or very frequent
intervals. Queries are small, requiring less processing.
These differing requirements must be scoped to set priorities -
for example, a data-mining query should have a lower priority than
a tactical query. Each class of query must be sized for capacity
planning and priority in the queue. Continuous monitoring alerts
support IT if the queue gets out of control.
The Right People For The Right Jobs
Too often, BI resources are drawn from traditional IT pools. Most
of these people have experience with transactional systems and databases
supporting transactional systems. The design and operation of a
data warehouse to support BI is fundamentally different and requires
not only a different skill set, but a different mind set. Some can
appreciate the difference and are able to apply their experience
with transaction-oriented, real-time systems to the BI environment.
BI requires more understanding of how the business works. Building
Standard Business Value libraries and Master Data sets needs a contextual
basis. Good technical skills does not necessarily translate into
good BI skills.
Eventually, the data
warehouse personnel, who work most closely with business users,
develop considerable business knowledge. In the interim, IT resources
and business users must work closely together to ensure this divide
Automation of ETL Processes
Extracting data from
transactional systems and feeding it in real time to a data
warehouse is a finely tuned process. A process that needs to
be as automated as possible. Human intervention should only be required
when an alert is triggered from the monitoring system.
The process must also be be a 'services' model - flexible and reusable.
Manage Storage Retention Periods
data warehousing, data changes more frequently than with traditional
warehousing. To store all changed records over extended periods
results in massive data volumes and significantly impacts computing
resources and data access.
Just as query requirements are carefully analysed and scoped, data
availability must also be carefully assessed. The business must
carefully decide which changes to store in the warehouse and which
changes should be overwritten. For example, an airline constantly
overwrites a flights ETA, as there is no business value in tracking
changes. However, all data about customers, bookings, and seat inventory
are preserved at every point in time, as this data supports a a
wide variety of business uses.
To reduce complexity further, views can be created that only include
active records, shielding users who need current data from complex
Storing the enterprise
warehouse data in 3rd normal form supports and encourages enterprise-wide
use. Data is easily maintained, redundancies are eliminated, shared
data definitions are agreed and an enterprise-wide view of the data
is provided. This the the most likely model to provide data that
meets the requirement of 'a single version of the truth'.
Business users who are confident in enterprise
wide data are less protective about wanting to retain their
local data silos.
Next: BI Excellence
Back To Top
BI Strategy Index | BI
Vision | PM Objectives
| Drivers Of BI | Barriers
To BI | BI Lifecycle | BI
Strategy Document | BI Long Term
| BI Evolution | BI
Best Practices | BI Excellence Survey
| BI Scorecard | BI
& ERP | BI & SMBs | BI
and Cloud Computing
NEW! - Get Your Copy of the New
BI Strategy Guide HERE!
Your Choice of Print Or Instant Download Ebook