Information
about the customer is the foundation of Customer Relationship Management.
Every step in the CRM process, every principle of the CRM approach,
requires tangible data about customers. To apply CRM to the entire enterprise,
information must be consistent, reliable, and in a usable format.
Piecing
together information about customers from multiple systems for every
analysis or marketing campaign creates inefficiencies and bottlenecks.
Centrally storing a single, comprehensive view of customers can significantly
reduce demands on IT and improve the efficiency of marketing and customer
communication processes.
In my last
article, I described the importance of crafting a customer strategy
prior to focusing on CRM technology. Once the strategy is in place,
CRM cannot take place effectively without an investment in technology
to access customer information. The foundation for CRM efforts is the
customer data warehouse.
Yet studies
on data warehousing that have been released over the past two months
raise questions as to the success of data warehouse projects. A Data
Warehousing Institute study of 1,600 companies released in March found
that 41 percent of respondents said they are experiencing difficulties
with their implementation and 42 percent found that the data warehouse
implementation met their expectations.
Another
finding in this study caused greater concern - only 13% of respondents
track data warehousing return on investment and less than 40% said they
plan to begin tracking ROI. As more companies seek to measure the return
on their CRM spending, the pressure to implement data warehouses on
time and within budget will increase.
If data
warehouse implementation projects are expensive and have a high risk
of going wrong, why would I recommend that any firm undertake such a
project now? A data warehouse is essential for a business to have consistently
clean, reliable customer level data that is continuously accessible
enterprise-wide. The key is to plan and manage the warehouse implementation
with risk and return in mind. Below, I outline ten steps to implement
a data warehouse effectively.
Implementing
a Data Warehouse
The first
three steps are combined into a "define the project" phase.
This phase should take six to twelve weeks to complete, depending upon
the complexity of a firm's customer data and the number of technical
and business users interviewed to develop the business requirements.
The time to complete depends upon the amount of time the project team
and business users can devote to interviews. The deliverables for this
phase of building a data warehouse are a working project team, a clear
list of business requirements, and a conceptual data model.
Assemble
the Team
The success
of a data warehouse depends on a technical and business partnership
that endures long after the construction of the warehouse is completed.
The project implementation will require a team well versed in technology,
analysis and marketing and may include current and future users of customer
data. The team must be large enough to include the appropriate knowledge,
and small enough to meet frequently and make decisions quickly.
Gather
Business Requirements
The first
order of business for the project team is to collect system requirements
specific to each business unit that will use the warehouse. Business
requirements are simple statements of what the system is expected to
do. In fact, each statement implicitly begins with "the system
shall . . ." and describes a discrete detail of a business need.
For example,
after interviewing the sales department, the sales requirements may
include the statement, "provide summary reporting of the top 10
products sold each day/week/ month," along with twenty or even
fifty additional requirements for the sales function. Including every
customer point of contact in the requirements gathering process ensures
long-term support for the data warehouse.
Define
Technical Requirements
How each
business unit plans to use the data warehouse will shape decisions about
hardware and software, architecture and interfaces. Questions like who
will use the system, what level of technical or analytical expertise
does the user group possess, and what types of remote connections will
be needed require consideration during this step.
At the
end of this first phase, the project team can complete a conceptual
data model which illustrates what tables will exist in the warehouse
and how the tables will typically be related to each other. For example,
in a banking environment, the conceptual data model may show a table
for each product type (checking account, credit card account.) These
account tables will be related to the customer table as each customer
may have one or more accounts.
The next
three steps in building a data warehouse make up the "understand
the data" phase. This phase may take another six to ten weeks to
complete, depending upon the volume and complexity of the customer data
as well as what information is currently available about the data. The
deliverables for this phase are collectively referred to as metadata,
literally "data about the data."
Identify
Data Requirements
In this
step of the process, the project team converts the business and functional
requirements gathered in the first phase into a logical data model.
A logical data model is a picture showing all the tables in the data
warehouse, all the data elements in each table and the relationships
among the tables. There are several software packages available that
simplify producing logical data models. The time consuming process in
this phase is the review of the data model with the business users interviewed
in the first phase. Although this may be a tedious step, it is important
to gain approval and understanding from the business and technical users
of the warehouse.
Create Data Maps
The data
mapping step of the implementation charts the movement of data from
the source system, through data processing and conditioning, to its
location in the data warehouse. This step also describes how data is
extracted, transformed, and loaded (the process known as ETL) into the
data warehouse. The result of this step will be a physical data model,
showing how each element will be stored in the data warehouse.
Develop
the Data Dictionary
Finally,
the project team will develop a data dictionary - the reference guide
for designers, builders and users of the data warehouse. Data definitions
are presented for each element detailing exactly what the field represents
using definitions that are common across systems, channels and time.
The final
four steps in the process cover the "execute the plan" phase.
In most projects, the two planning phases take about two-thirds of the
project time, while executing the plan takes about a third. The time
to complete this phase will depend upon how rapidly the firm is able
to make decisions regarding technology and how willing the firm is to
put the appropriate resources toward the data warehouse build. The deliverable
for this phase is a populated, accessible data warehouse.
Determine
Whether to Use Outside Support
Companies
have a variety of support options to choose from for the development
and maintenance of their data warehouse. Consultants can handle design,
development, software/hardware selection and implementation and can
coordinate vendor relationships. Service bureaus can house your warehouse
on their hardware on a contract basis, although access to your data
may be limited. Your firm needs to assess the support they need from
third parties based on the size and scalability of your warehouse needs,
your available resources in the short and long term and your budget.
Decide
on Software and Hardware
Each firm
will need to choose hardware and software from a myriad of combinations
to meet the needs uncovered in the planning phases. Software packages
to review include database servers, ETL tools, and processors. Database
servers support data transport, querying and table management; many
are provided by large, well-known companies. Tools for extracting, transforming,
and loading (ETL) data transfer data from source systems to the data
warehouse and may clean or reformat data on the way. A firm may also
require a processor to condition data for analysis or to household data
for marketing.
Perform
Individual Warehouse Build
During
the first several builds of the warehouse, expect to fine-tune the production
process. The production process is the automated building of the warehouse
on a regular (nightly/weekly/monthly) basis. In this step, the project
team should celebrate successes and communicate setbacks to a wider
audience in preparation for regular warehouse communications.
Develop
a Production Calendar
After the
first several builds of the warehouse, the project team can create a
production calendar to let users know when and with what frequency data
will be available for querying and reporting. The production calendar
should also be widely shared.
Completing
the customer warehouse merely places customer data into a box. Next,
your firm must select and implement the tools that will allow you to
extract and use customer information for analysis and marketing. The
completed warehouse is the foundation for CRM, so celebrate its successful
implementation and begin designing and building the rest of the CRM
infrastructure to realize the return on your investment.
