8/2/2019 Search DnB
1/20
ManagerGet a 360 degree view of your customers with
our premier onsite data integration solution
Technical White Paper
Sales & Marketing
Solutions
Customer Integration
8/2/2019 Search DnB
2/20
Table of Contents
The Need for Customer Data Integration 3
Data Integration Challenges 4
The Data Integration Process 5
Deployment Options 8
Introducing Customer Integration Manager: the Premier Business 9Customer Data Integration Solution from D&B
Business Benefits 9
How It Works 10
The Customer Integration Manager Matching Process 12
Technical Benefits 14
Sample Applications 15
Integrated Customer Relationship Management 15
Consolidated Sales Data for Customer Analysis 16
Interactive Customer Lookup During Order Entry 17
Implementation 18
Monthly Update 18
Support 19
Summary 19
All data is fictitious.
2 Customer Integration ManagerTechnical White Paper
8/2/2019 Search DnB
3/20
The Need for Customer Data Integration
CRM. ERP. Supply Chain. Demand Management.
Business Intelligence. Data Warehouse. Look inside any
of today's strategic business technologies and you'll find
the same requirement: to share data freely and accurate-
ly across different business functions. Data sharing is
what lets customer relationship management systems
ensure consistent customer treatment across different
channels. It is what lets enterprise resource planning
systems coordinate manufacturing plans with sales
forecasts and inventory levels with delivery schedules. It
is what lets quality managers track customer complaints
back to specific production units.
In the abstract, data sharing looks easy enough just
connect a few boxes on a white board or assume the
whole company moves to a single all-purpose software
package. But reality is much more complicated.
Different departments, divisions, and subsidiaries will
continue to run separate systems, and even if every
member of the corporate family does adopt the same
integrated product,every supplier and customer will not.
While technology has simplified the physical movement
of data among systems, physical movement is not
enough. Data must be logically integrated as well.
This means your systems must recognize that the cus-
tomer who bought this product through the order entry
system is the same customer who logged that complaint
in the customer service system and the same customer
about to receive those mail promotions from the market-
ing system. The trouble is, each system holds its own
record for that customer in a different format, with dif-
ferent spellings, and possibly even under different
names. You can close your eyes, click your heels three
times, and hope the separate systems go away. Or you
can find a way to identify the relationships among those
records despite the differences.
Finding those relationships is part of a
process called customer data integra-
tion. There is more to customer data
integration than just matching cus-
tomer records, although matching is
at the heart of the process. Customer
data integration also involves estab-
lishing relationships among records
that cannot be matched directly, such
as corporate parents and subsidiaries
with different names or addresses. It
identifies the most accurate version of
information such as a company tele-
phone number or mailing address
that is different in different systems. Most important, it
assigns a permanent identification number to each cus-
tomer. This permits easy sharing of customer data with-
out complicated, on-the-fly matching processes.
Accurate customer data integration is particularly
important for achieving the promises of today's cus-
tomer relationship management projects: lower market-
ing costs, higher retention rates and greater revenue per
3
Customer Integration Managerlinks customer data across systems
Web Site
Accounting Service
Sales
MarketingDatabase
Production
DataWarehouse
Inventory
Customer Integration Manager Technical White Paper
ManagerIntegration
Customer
8/2/2019 Search DnB
4/20
customer, among other tangible benefits. All these goals
are premised on developing a complete picture of each
customer's relationships with your business. This
requires both combining data from different systems
and linking records within the same system that refer to
the same customer. Sometimes multiple records exist by
mistake and can be combined. But there is often a valid
reason for the same customer to have more than one
record, particularly when dealing with businesses rather
than individual consumers. A business customer may
have separate accounts for different departments, loca-
tions or even specific projects. While these accounts
must remain distinct for operational reasons, accurate
customer management and analysis must recognize
that they are related. In fact, the most active customers
are the most likely to have multiple accounts making
consolidation a critical function in managing your
largest clients as well as the smaller ones.
Although improved customer relationships are more
than sufficient reason to undertake a data integration
project, there are other benefits as well. One is the
increasing reliance on consolidated data for processes
other than direct customer contact, such as distribution,
financial planning and purchasing. In fact, consolidated
data on suppliers can sometimes be as valuable as con-
solidated data on customers. Data integration is also the
foundation of data warehouse and business intelligence
systems, which are increasingly used for both strategic
and tactical business decisions. Nothing destroys the
credibility of a data warehouse report more quickly than
an obvious understatement of sales to the company's
largest accounts and, as already noted, inaccurate inte-
gration is most likely to affect precisely those customers.
At the most basic level, customer data integration saves
money by avoiding redundant data entry and cleaning
projects, providing more accurate information, and sup-
porting a complete and accurate view of the customer
that is multifaceted.
Data Integration Challenges
The importance of customer data integration is clear, but
so are the difficulties. The fundamental problem is that
data is captured in many different systems, each with its
own formats, standards and requirements. Data may be
perfectly adequate for its original purpose yet still not
suitable for integration:for example,names and address-es in free form text fields can generate serviceable mail-
ing labels but are difficult to analyze for matching.
Similarly, operational processing is often unaffected by
data entry inconsistencies, such as variations in formats
and abbreviations, that make parsing and matching still
more difficult. Operational users often cram extraneous
information such as customer status codes into name
and address data that makes perfect sense to a human
reader or customized computer program, but confuses
external integration processes. Or an operational system
may lack basic information needed for any matching: for
example, a help desk system might capture only the cus-
tomer's first name and a telephone number to call with
an answer, but no mailing address or account ID.
But data integration is difficult even when the finest
operational systems are in place. Customers themselves
often provide different information to different systems,
4
Although improved customer relation-ships are more than sufficient reason toundertake a data integration project,
there are other benefits as well.
The fundamental problem is that data iscaptured in many different systems, eachwith its own formats, standards andrequirements.
Customer Integration ManagerTechnical White Paper
8/2/2019 Search DnB
5/20
either by mistake or to meet different purposes: a billing
system may correctly hold a different address than a
shipping system. Nor do customers update every com-
pany system when important changes take place: if they
haven't called for customer service since their last move,
the customer service system will have an outdated
address. The system changes needed to support integra-
tion can themselves be a challenge, particularly since
many corporate IT departments have little experience
with customer data integration technologies.
Working with business rather than consumer data adds
yet another level of complexity. Business records include
not just a simple name, address and city/state/ZIP but
personal and company names, titles,departments,build-
ings, mail stops, and other elements. They also contain
industry terms, such as "DBA" for "doing business as",
that must be recognized and interpreted appropriately.
Few companies follow enterprise-wide formats to cap-
ture and hold these elements and even firms with
internal standards must contend with data from exter-
nal sources. Business files may also hold different names
for the same company, such as legal vs. trade names.
Common names may themselves be represented differ-
ently "Kentucky Fried Chicken", "KFC", "Kentucky FC"
and any number of variations. Related firms may have
totally unrelated names, such as Lotus Development
Corporation, a subsidiary of IBM. Multiple locations for
the same business must be somehow brought together.
Today's business trends make customer data integration
more challenging than ever. As more systems are con-
nected, integration techniques must become more effi-
cient to handle the increased volume. Interactive appli-
cations impose strict performance requirements to
ensure data entry and customer service processes are
not delayed. Privacy and security regulations impose
strict limits on how data is shared, making it more
important than ever to ensure accurate consolidation,
precise access control, and audit trails on changes
Privacy concerns also make it more difficult to gather
data directly and thus more important to share as
widely as possible whatever data your company has
already acquired.
The Data Integration Process
Customer data integration may be difficult, but it is
far from impossible. In fact, experience over the past
several decades has provided a firm understanding of
the steps in an effective data integration process:
Input. This is the initial process of gathering and pre-
senting the data to be integrated. Traditionally, large
numbers of records were extracted from source systems,
loaded into files, and run through the integration process
5
Data Integration Process
Parse
Standardize
Group
Match
Enhance
BatchOutput
OnlineInput
OnlineOutput
ValidationData (if used)
SupplementalData
ReferenceRules & Data
BatchInput
Customer Integration Manager Technical White Paper
8/2/2019 Search DnB
6/20
as a group. Results were then posted back to the source
system or used elsewhere. This is referred to as "batch"
processing. Many data warehouses and marketing data-
bases are built this way.
Other systems use "online" processing, where an exter-
nal system presents a single record to the integration
system and waits for the result before continuing. This
often involves interactive processes where a human user
is working on the record. It also includes fully automat-
ed processes, such as processing a credit card transac-
tion, where immediate response is needed even though
there is no direct user involvement. Online processing is
technically more difficult than batch processing, because
the external system must be modified to present the
data and use the results. Online integration must also
run quickly enough to avoid a significant decline in
source system performance.
Matching. This is the heart of the data integration
process, where records are compared to find which are
related. It usually includes three stages:
1. Parsing isolates data elements so they can later be
compared with corresponding elements on other
records. For example, a typical parser would split a name
into title (Mr., Mrs. Ms., etc.), first name, middle initial,
last name, generation (Jr., III) and suffix (Ph.D., L.L.D).
Addresses and other data would be similarly broken into
components. Parsing also identifies missing or question-
able data.
Parsing is necessary because many systems
store name and address data as text lines
rather than separate elements. Often the pars-
er must determine both the nature of the line
and the sequence of the elements within it.
Sophisticated parsers use reference tables that
list how specific words are likely to be used
("Corporation" is probably part of a company name;
"Andrew" is probably a first name) and the likely
sequences of elements within lines of different types.
Reference tables must be tuned for specific applications
such as business or consumer processing and to adjust
for different national data formats.
2. Standardization is the stage in the process thatconverts data elements to standard formats to improve
match accuracy. On consumer records, it might replace
nicknames such as Bob, Bobby and Rob with a formal
name of Robert. For businesses, it might replace differ-
ent versions of a company name with a similar standard.
Like parsing, standardization relies heavily on tables to
make such corrections. Address standardization may sim-
ply apply standard formats or it may extend to valida-
tion and correction using actual postal tables. These
could determine which street names exist in which cities
and what postal code applies to each address. But postal
Parsing
6
Parsed Record
Name: Dan BrandstreetStreet Nbr: 3Street Name: Silven WayCity: ParrippanyState: NJPostal Code: 07054
Input Record
Dan Brandstreet3 Silven WayParrippany NJ07054
Standardization
Input Record
Name: Dan BrandstreetStreet Nbr: 3Street Name: Silven WayCity: ParrippanyState: NJPostal Code: 07054
Standardized Record
Name: Dun & BradstreetStreet Nbr: 3Street Name: Sylvan WayCity: ParsippanyState: NJPostal Code: 07054
Customer Integration ManagerTechnical White Paper
8/2/2019 Search DnB
7/20
validation and coding are generally considered separate
processes from other standardizations and are often
applied outside of the data integration system.
3. Grouping determines which records are compared to
each other. This avoids the inefficiency of comparing
everything to everything else. Most systems build a
group key by extracting portions of data elements such
as state, city and last name. Often several keys are creat-
ed to bring together records that a single method might
miss. For batch matching, records are typically assem-
bled into one large file, sorted or indexed on the key, and
then compared in sequence. For online matching, the
keys are stored permanently on the customer database.
When a record is presented for matching, the online
system generates keys for that record and selects
customer database records with the same keys for
detailed comparison.
Linking. The process of deciding which records should
be considered a match. All matching systems ultimately
rely on string comparisons that is, they assess the sim-
ilarity between the text strings assigned to the different
data elements in the records being compared. Some sys-
tems extract a few characters from each data element to
create a "match code" and treat all records with the same
match code as a match. Some calculate numeric similar-
ity scores for each pair of elements, add the scores, and
treat as a match any pair of records whose total exceeds
a specified threshold. Some assign alphabetic codes to
different types of element matches and specify which
combinations of codes are considered a record match.
Each approach has its proponents, but the code-
combination technique is the most common among
sophisticated matching systems. The advantage of this
method is precise control over how each combination of
element-level matches is treated. The major criticism is
that specifying treatments for tens of thousands of dif-
ferent combinations is a great deal of work. But vendors
of such systems provide pre-built tables to spare users
most of this labor.
An ultimately more significant difference among match-
ing systems is whether they compare input records
against each other or against a separate validation file
Such validation files are compiled from external sources
and contain all entities that might appear on the input
files themselves. Each entity on the validation database
is assigned a fixed ID number. Input records are com-
pared with the validation file using conventional string-
comparison techniques;when a match is found,the ID iscopied from the validation database record to the input
record. Input records that end the process with the same
ID are assumed to match.
Validation-based matching is significantly more effec-
tive than direct comparison because the validation data-
base can contain links between records that are not
physically similar. For consumer data, this might be
someone who has changed addresses or has summer
and winter homes. This not only lets companies unify
customer records that would otherwise remain frag-
mented, but also provides notice that a customer has
moved even if he has not informed the company direct-
ly. Advantages for business matching are greater still
because of the many reasons legal vs. trade names
parents vs. subsidiaries, headquarters vs. branch sites
etc. that related records appear different.
The value of a validation database depends on its cover-
age and the accuracy of its linkages. The major data
compilers can afford to invest in comprehensive valida-
7
An ultimately more significant differenceamong matching systems is whether theycompare input records against each otheror against a separate validation file.
Customer Integration Manager Technical White Paper
8/2/2019 Search DnB
8/20
tion files, knowing they will reuse the data many times
over. Since only an ID number is transferred from the
validation database to the input records, other informa-
tion compiled by the database developer can remain hid-
den or be provided for an extra fee. Validation-based sys-
tems are available for consumer data in the U.S., United
Kingdom and a few other countries. Business matching
files are available for most major developed countries.
Most systems also provide an option for manual
review of suspect matches that is, record pairs that
are similar, but not close enough to be accepted as
matches automatically.
The final output of any linking process is a set of records
with IDs that identify members of the same match group.
In consumer matching,the match group often represents
a single individual or household; in business matching, it
may represent records at the same site or belonging to
the same company. Matching systems often give each
record several IDs, representing different groupings.
Enhancement. Once matches are identified, the
process may attempt to create a more complete cus-
tomer record by combining information from multiple
sources. This could involve comparing conflicting data
from different systems and picking the values that seem
most likely to be correct. For example, birth date may be
populated in one system and blank or obviously wrong
(e.g., 99/99/99) in another. Or the system may combine
information from multiple records, such as the sum of all
account balances. It might also look at an external data
source that has information not collected internally.
Enhancement can be separate from data integration,
particularly when the enhancement data resides at an
external source. But it is often more efficient to do
enhancement within the integration processing stream.
Like other integration processes, enhancement can be
performed in batch or online. With today's communica-
tion technologies, even external data can be accessed by
in-house online systems at acceptable speeds.
Output. The final step in integration is making the out-
put available for use. In some cases, particularly online
systems, this means returning the data back to the
source system. In other cases, such as building a data
warehouse, the information will go somewhere other
than the original source. In still other configurations, the
result is added to a cross reference table that links source
system IDs such as account numbers with a standard
customer ID. This particular approach allows source sys-
tems to share customer data without modifying their
internal processes or data structures. Whatever the
details, the only task of the integration system is to pro-
duce an output either a batch file or online transaction
that links the original input with a standard customer
ID. Other systems can then process this data in any way
that is appropriate.
Deployment Options
Many firms have installed piecemeal integration solu-
tions to meet specific operational requirements, such as
sharing data between two particular systems. But enter-
prise-wide customer data integration usually involves a
dedicated system to gather, process and distribute data
across many different sources. Such systems can reside
at external service bureaus, as part of an in-house opera-
tion, or even in a hybrid configuration.
External service bureaus provide specialized expertise
and have economies of scale that can make them more
cost-effective than in-house processing. Service bureaus
are particularly appropriate for periodic batch updates,
where immediate turnaround is not required. They are
also often used to produce consolidated files, such as
marketing databases, that will reside at the service
bureau rather than the company's internal systems.
8 Customer Integration ManagerTechnical White Paper
8/2/2019 Search DnB
9/20
External processing also simplifies validation-based
matching, since the service bureau has immediate access
to updates to the validation file.
In-house operation has become increasingly common,
particularly in data warehouse projects where the entire
system is maintained internally. In-house operation is
almost required for online applications, where the data
integration processing must take place quickly and be
tightly connected with other corporate systems.
Hybrid solutions let companies do most integration
processing internally, but still access external systems
when needed. The most common hybrid configuration
reads external files for validation or enhancement data.
This lets companies use up-to-the-second versions of
those files without the cost and security issues of copy-
ing them onto in-house systems. High-speed communi-
cations make it possible to use hybrid approaches even
for online integration.
Introducing Customer Integration Manager:
the Premier Business Customer DataIntegration Solution from D&B
Customer Integration Manager provides a single,com-
prehensive solution to your company's business customer
data integration needs. Customer Integration Manager:
Assigns consistent customer IDs to records from all of
your systems, allowing easy integration of customer
information throughout your company
Combines records from your own systems and from
D&B in an in-house validation directory that is easily
accessible for both batch and online processing
Regularly updates the validation database with fresh
D&B data
Remotely searches D&B's own computers for matching
records that have not already been downloaded
Includes D&B's own customer matching engine, a
sophisticated product using world-class technology and
specifically tuned for business matching applications
Performs both batch and online matching
Returns a single best match or provides interactive
users with a set of possible matches to evaluate
Can automatically request further research by D&B
staff when no match is found
Codes records with the D&B D-U-N-S Number, D&B's
universal company ID, opening the door to easy
enhancement with data from D&B's own files and from
many third parties.
Business Benefits
Customer Integration Manager provides your company
with the benefits of sophisticated customer data integration
without the costs of building it yourself. By installing
Customer Integration Manager, your company will:
Gain a consolidated view of customer relationships
that are spread over multiple accounts, sites and trade
names, thereby supporting effective customer relation
ship management programs
Facilitate online access by sales and service staff to
customer information, no matter where the information
is stored
Expand your understanding of each customer by
adding enhancement data from D&B and other
providers
9Customer Integration Manager Technical White Paper
8/2/2019 Search DnB
10/20
Ensure the integrity of customer data in each
corporate system, by identifying and eliminating dis-
crepancies caused by entry errors, outdated information
and incomplete records
Improve the efficiency of internal operations
by helping to coordinate front- and back-office systems
Make better business decisions by providing more
accurate data for marketing studies, sales analyses and
other types of research
Save time and money by acquiring a complete solution
that is easily adapted to your needs, rather than attempt-
ing to develop a comparable system internally
Benefit from the experience of D&B consulting servic-
es, world leaders in managing business customer data
HowIt Works
Customer Integration Manager accepts name and
addresses input from your company systems and returns
a standard ID plus company data from D&B files.
The system is built around the Common Customer Directory.
This starts with a set of D&B records representing the
universe of firms your company is likely to do business
with. Each record refers to a specific business site that
is, a branch or office of a single business at a single loca-
tion. Companies with multiple locations will have at
least one site record for each location. If a company does
business under multiple trade names at the same site,
there will be a separate site record for each name. This
improves matching accuracy since input records carry-
ing any of these names can still be recognized as part of
the same site.
Each D&B site record carries a unique D-U-N-S Number,
a universal site identifier assigned by D&B and used
throughout the world as a standard business ID. Records
for the same site but with different trade name will have
the same D-U-N-S Number. In addition to the site's own
D-U-N-S Number, the site record will carry the D-U-N-S
Numbers of the site's headquarters or corporate parent,
national parent and global parent sites if these exist.
D&B can provide additional data about each site from its
own files, including name, address, telephone number,
revenue, number of employees, and key executives. This
is stored on the site record or in separate tables that are
linked to the site records through the D-U-N-S Numbers.
All records carry a second site number, called the Logical
Site ID or LSID. Unlike D-U-N-S Numbers, LSIDs are
assigned independently at each Customer Integration
Manager installation. Use of LSIDs is explained next.
10
Underlying Technology
Operating System: HP-UX 11, Sun Solaris 2.9,Microsoft Windows 2000; configuration dependson data volume and throughput
Minimum hardware requirements:- Application Server: 2 CPUs @ 750 MHz,
Batch data inputs and outputs can be sent as delimit-ed or fixed record length files.The system can moni-tor a specified directory and automatically initiatebatch processes when new files appear.
Customer Integration ManagerTechnical White Paper
with open-source solutions, Jboss and Jetty.
scalable architecture. It was built on a Struts framework
Customer Integration Manager is built on industry-standard technologies, making it an easy fit for mostcorporate systems environments. Supports platforms
include:
- Database Server: 4 CPUs @ 440 MHz,4 GM RAM, 1 TB Disk.
Database: DB2 (Version 2), Microsoft SQLServer 2000, and Oracle 9i RDBMS.
2 GM RAM, 100 GB Disk
The system is written in J2EE to provide a highly
8/2/2019 Search DnB
11/20
Once the D&B records are loaded, users can match them
against records from the company's own systems. (See
the next section for a detailed explanation of the match-
ing process.) If the input record matches a D&B record,
the input record is coded with the D&B record's
D-U-N-S Number and LSID. This will also happen if the
input record matches a previously-loaded company
record that is itself linked to a D&B record. If the input
record matches a company-provided record without a
D-U-N-S Number, it is given that record's LSID. If there is
no match at all, the system assigns the record a new
LSID. The result is that every record has an LSID while
records linked directly or indirectly to a D&B record have
a D-U-N-S Number as well.
Customer Integration Manager
can find additional D-U-N-S Num-
bers by connecting remotely to
the D&B master file. This
process, called Remote Resolution,
lets Customer Integration
Manager search for records that
were not selected for the local
database or have changed on the
master file but have not yet been
updated locally. If a remote match
is found, the D&B record is copied
to the Common Customer Directory
and its D-U-N-S Number is placed
on the input record;an LSID is
created or copied from an existing
record as appropriate. If no
match is found on the D&B
master file, the record can be
forwarded to D&B's Globa
Resolution Services staff. Thi
group will look for a match that
the computer missed and can
contact a non-matching site to
determine if it is truly a valid business. Since manua
research takes time, the non-matching record is initially
returned without a D-U-N-S Number. If a D&B match is
found later, the D-U-N-S Number will be added and the
LSID adjusted if this links the record to a different site.
Customer Integration ManagerOperation
Common Customer DirectoryCreation
11
Company Systems
Manager
Name & Address
Name & AddressD-U-N-S Numberand D&B data
Customer Integration Managercan findadditionalD-U-N-S Numbers by connectinremotely to the D&B master file.
1
2
Directory
Company InputDun Wordbase 3 Silver Way Parrippany NJDans Pizza 13 Short Ave Parsippany NJ
D&B Input = Initial Common Customer Directory D-U-N-S # LSIDDun & Bradstreet 3 Sylvan Way Parsippany NJ 123456789 1111Dun Worldbase 3 Sylvan Way Parsippany NJ 123456789 1111
Combined Input = New Common Customer Directory D-U-N-S # LSID
Dun & Bradstreet 3 Sylvan Way Parsippany NJ 123456789 1111
Dun Worldbase 3 Sylvan Way Parsippany NJ 123456789 1111
Dun Wordbase 3 Silver Way Parrippany NJ 123456789 1111
Dans Pizza 13 Short Ave Parsippany NJ 2222
(1) Company input matches D&B record; has D-U-N-S Number and LSID
(2) Company input does not match D&B record; has LSID only
Customer Integration Manager Technical White Paper
Customer Integration
CommonCustomer
8/2/2019 Search DnB
12/20
Records can be presented for matching in batch or
online. Online processes can choose the best match
automatically or present interactive users with a list of
candidates, ranked to present the most likely matches
first. Batch processes can also create lists of questionable
matches for users to resolve. Users can force a record to
become part of a specified site, regardless of how closely
it matches records already assigned to that site.
Once processing is complete, the record can be added to
the Common Customer Directory, sent to a flat file,
returned to an online system,or any combination of these.
Adding the records to the Common Customer Directory lets a
company assemble a complete list of its customers.The
system's matching functions can then be used to check
whether new inputs, such as responses to marketing
promotions or requests for technical support, match a
known customer record.
Output formats are customized for
each process and can include the
D-U-N-S Number, LSID, and other items
from the Common Customer Directory.
This database includes several separate
tables, linked by D-U-N-S Number, and
can incorporate data from D&B's master
database. Common Customer Directory
records can also store a user-specified "hard
key," typically a record number or account ID
from the original source system.Company
systems can use this key to find a record that
has already been loaded to the Common
Customer Directory. They can then use
the D-U-N-S Number or LSID on that
record to retrieve additional information
or to find records from all company
systems with the same D-U-N-S Number
or LSID.
The Common Customer Directory also receives monthly
updates from D&B itself. These updates include D&B
records that have been changed and new D&B records
that meet the user's selection criteria. The system can
reassign LSIDs and D-U-N-S Numbers if a match is found
to be inaccurate.
The Customer Integration Manager
Customer Integration Manager is powered by the DUNS
Name Matching API,an implementation of D&B's patented
name matching process.The system provides parsing,
normalizationand linking functions,all optimized for busi-
ness data. Linking uses the "code combination" tech-
nique, generally considered the most effective method
available. This assigns a code for the type of match
between each element pair,and then applies a table that
defines how each set of code combinations is to be treated.
12
Step 1: Compare Records and Generate Match Grade
Input Record Validation Record Similarity Type CodeName Dun Wordbase Dunn Worldbase Moderate B
Street Nbr 3 3 Strong AStreet Name Silver Way Sylvan Way Moderate BCity Parripany Parsippany Strong AState NJ NJ Strong APO Box Blank ZTelephone Nbr 973-605-6000 Blank Z
Match Grade = BABAAZZ
Step 2: Find Corresponding Confidence Code in Confidence Matrix
Confidence Matrix (portion)Match Grade Confidence Code Match Percentage
BABAAZF 8 90%BABAAZZ 8 90%BABABAA 8 89%
Step 3: List All Record Pairs Exceeding a Specified Confidence Code
(In this example: record 111111 matches record 222222; records 333333 and 555555 bothmatch record 444444)
Input Record ID Linked Record ID Confidence Code111111 222222 10
333333 444444 8
555555 444444 8
Customer Integration ManagerTechnical White Paper
Matching Process
Match Engine Linking
8/2/2019 Search DnB
13/20
As described elsewhere in this White Paper, Customer
Integration Manager matches input records against a
validation table, rather than each other. This lets it
identify relationships among records that do not match
each other directly. Details of the process are:
Input. The match engine accepts inputs for business
name, address, city, state, postal code, telephone number
and country code. Country code is used to select coun-
try-specific matching rules and postal code is used to
resolve discrepancies between inputs and D&B valida-
tion records. The other elements are used in the match-
ing process itself.
Parsing. The system uses tables of key words and iden-
tification rules to split the address into street number,
street name and Post Office Box. Parsing tables and rules
are built into the system and not directly accessible by
end-users.
Normalization. The system normalizes the name and
address elements by removing extraneous words (such
as "The") and applying standard forms and contractions.
Normalization tables include standard names for specif-
ic companies as well as general business terms. Address
elements are normalized to improve matching.
Normalized data is a condensed form of the original
input designed for internal comparisons; it does not
replace the original data for display or output.
Grouping. The system creates group keys by extracting
information from the normalized name and address ele-
ments. The name and address keys can be used sepa-
rately or in combination to select records to compare in
the linking process. The system will compare all records
with the same key, regardless of how many there are.
Linking. The system compares the name, street num-
ber,street name, city, state,Post Office Box and telephone
number elements in each pair of records. The compar-
isons use different methods depending on the data type,
but each produces one of four similarity codes
A (strong), B (moderate), F (no similarity) and Z (one or
both is blank). These codes are strung together with the
most significant fields listed first to form a Match Grade
A system table called a Confidence Matrix relates each
Match Grade to a Confidence Code and Match
Percentage. The Confidence Code is a number from 0 to
10, typically interpreted as: 0=no match; 1-4=weak
match; 5-7=limited match needing user validation;
8-10=high quality match. The Confidence Matrix is
provided with the system; because it is based on
extensive empirical research, D&B recommends it
remain unchanged, but it could be modified if necessary
Different runs of the match engine could point to
different Confidence Matrices, although again this is
not recommended.
Output. The output of the matching system depends
on the application at hand. In Customer Integration
Manager it can be one best match or a set of possible
matches from the Common Customer Directory. Output can
include contents of the input record,the D-U-N-S Number
and/or LSID,match results like the Confidence Code or
Match Grade, treatment information such as whether
the record was sent for remote resolution or inserted in
the Common Customer Directory, and enhanced data
from a matching D&B record.The system can generate
different output for matched and unmatched records.
Users control many aspects of the matching process
some within the match engine itself and others in sur-
rounding applications.. In Customer Integration
13
Linking uses the "code combination"technique, generally considered themost effective method available.
Customer Integration Manager Technical White Paper
8/2/2019 Search DnB
14/20
Manager these choices are expressed through
Workflows. Users can determine which types of
records to consider for a match: for example, they may
wish to exclude records that belong to companies known
to be out of business. They can also set the Confidence
Code levels required to accept a match or send it for man-
ual review, how many candidates to return for review,
and which types of records are sent to D&B for remote
resolution. Users willing to override D&B recommenda-
tions can change entries in the Confidence Matrix, allow-
ing them to determine which data elements are consid-
ering in the matching process and how different combi-
nations of element matches are treated. In many cases,
different settings can be applied in different situations
by creating alternate Workflows.
Technical Benefits
The approach taken by Customer Integration Manager
has many advantages:
One table links site records from all sources. The
Common Customer Directory provides all systems with
a single source for links among customer records.This
simplifies cross-system information requests, since
programs must look in only one location.The same directory
is shared by batch and online processes,so the company
need not develop multiple,redundant matching systems.
Minimal impact on source systems. The Common
Customer Directory can be created and managed without
changes to the source systems. Because the Common
Customer Directory holds both source system keys (the
"hard key") and internal linking keys (the D-U-N-S
Number and LSID), source systems can access specific
Common Customer Directory records and find related
records in other systems still without storing any new
data in the source systems themselves.
Company-wide ID scheme. The D-U-N-S Number and
LSID provide consistent IDs to identify the same
customer in all systems. This means that customer data
can be assembled with a simple query rather than a
complicated on-the-fly matching process. Companies
willing to store the D-U-N-S Numbers or LSID on source
system records could join customer data across multiple
systems without using the Common Customer Directory
as an intermediate cross-reference table.
Incorporates D&B data.D&B records in the Common
Customer Directory allow verification-based matching,
which is the most effective way to link business records.
In addition, the corporate family data on D&B records
provides connections that cannot be derived from cus-
tomer records alone. Because company input and D&B
records are coded with the same D-U-N-S Numbers and
LSIDs, it is easy to combine data from both sources.
Accumulates company-specific information. Company-
supplied records can be manually associated with a site
even though they appear unrelated. When similar
records are presented in the future, they will be linked to
the same site automatically. This allows the Site
Reference Database to build a store of information that
improves match accuracy over time. It also lets the sys-
tem associate D&B data with customer records that do
not directly match a D&B record.
Superior matching technology. D&B uses sophisticated
matching technology that is already optimized for busi-
ness data. Users can deploy and operate the system with
a minimum of effort, but still control critical parameters
such as the confidence level required to accept a match
or to flag it for manual review. Online matching
functions are accessed through a Java API that
lets users specify precisely how each transaction
is handled.
14 Customer Integration ManagerTechnical White Paper
8/2/2019 Search DnB
15/20
Remote connections to D&B data and research services
This gives access to the most current D&B data without
constant updates to local files. It also lets the system find
matches against D&B records that are not stored locally,
allowing companies to limit the size of the in-house
database. Automated transfer of questionable records to
D&B for remote resolution permits use of this service
with minimal effort from company systems staff. Customer
Integration Manager supports both batch and online
Data Integration Batch, included with Customer
Integration Manager, gives each client a password-protected
home directory on the D&B server.When clients send files
to the directory via FTP. D&B systems process them
automatically, return the results to the directory
and generate email notifications to D&B staff and clients
Data Integration Toolkit, a set of optional modules
allows online, world wide integration of client systems
with D&B systems for corporate family linkage,
enterprise management, financial management and
marketing management.
Easily extended. Custom database tables can easily be
linked to the Site Reference Database using the
D-U-N-S Number or LSID as a key. These tables can hold
any data needed for company-specific applications.
Sample Applications
The following examples illustrate some of the ways that
Customer Integration Manager can be employed.
Integrated Customer Relationship Management
The lead processing module of a company's sales force
automation system calls Customer Integration
Manager to check if an inquiry is from an existing customer
and uses the result to determine how to respond.This
company has stored the LSID on customer records through-
15
The D&B Database
The D&B database is the world's most compre-
hensive repository of business information. It
covers more than 80 million business sites in214 countries, including 17 million in the United
States. Locations that are part of a larger
organization are linked in a corporate family
tree that identifies relationships between
headquarters and branches of the same corpo-
ration, and between parent corporations and
their subsidiaries.
Each site is assigned a D-U-N-S Number, aunique nine-digit identifier. In addition to a
site's own D-U-N-S Number, its record will carry
the D-U-N-S Numbers of the headquarters or
parent site, the ultimate domestic parent, and
the ultimate global parent. The D-U-N-S
Number is used by more than 50 global, indus-
try and trade associations including the United
States government, European Union and United
Nations.
The D&B database is refreshed more than one
million times each day with data from tele-
phone calls, company Web sites, and business
partners. More than 1,000 data elements are
collected; those available in the Common
Customer Directory include company name
and address, telephone number, and indicators
for whether the site is currently active, out of
business, marketable, and has a valid industrial
classification code available. Customer
Integration Manager can also store other D&B
data including revenue, number of employees,
year started, and executive names and addresses.
Customer Integration Manager Technical White Paper
connections:
8/2/2019 Search DnB
16/20
out its CRM systems, so once Customer Integration
Manager has identified the customer's LSID,the CRM sys-
tems can manage interactions among themselves with-
out further Customer Integration Manager involvement.
1. The lead processing system receives an inquiry from
an unknown person. The system reads the business
name and address and sends these via Java to the
Customer Integration Manager API.
2. Customer Integration Manager parses and standardizesthe input. It finds a match in the Common Customer
Directory and returns the LSID via the Java API.
3. The lead processing system queries the company's
central marketing database, using the LSID as a key. The
query returns a customer status code.
4. The lead processing system uses the customer status
code as part of its business rule to determine how to
respond. In this case, the rule determines this is a high
value customer who should receive a telephone call. The
lead processing system sends a transaction to the call
center, including the LSID as part of the customer identi-
fier. The call center calls the customer.
Consolidated Sales Data for Customer Analysis
A company uses Customer Data Integration to associate
transactions from multiple systems with the proper customers.
The company uses D-U-N-S Numbers to consolidate
information from different branches at the headquarters
level. To ensure that as many records as possible have a
D-U-N-S Number, it uses D&B remote resolution services
for inputs that do not match the Common Customer
Directory.
1. The customer database load process creates a flat file
with records from multiple source systems. It places this in
a directory assigned to receive such files. The Customer
Integration Manager scanning and file transport modules
(SST and FTM) notice the file,transfer it to a work area,
and call Customer Integration Manager's matching module.
2. Customer Integration Manager parses andstandardizes each record and looks for a match in the
Common Customer Directory. Records that match an existing
entry are coded with the site and ultimate parent D-U-N-S
Number and accumulated in a flat file.
3a. Records that do not match a Common Customer
Directory entry are accumulated in a (different) flat file.
When processing is complete,this file is sent to a designated
Integrated Customer Relationship Management
16
Lead Processing
Inquiry from Customer
Phone call to Customer
Call Center
1
234
MarketingDatabase
Customer Integration ManagerTechnical White Paper
Customer Integration
Manager
CommonCustomer
Directory
8/2/2019 Search DnB
17/20
directory.SST and FTM send it to D&B for remote resolution.
If the D&B system finds a match on its master database,
it outputs the original record and the D&B site record. If
there is no match, it sends the record for manual research
and outputs the original record with an appropriate flag.
When the process is complete, the files are placed in a
directory where SST and FTM sends them back to
Customer Integration Manager.
3b. Customer Integration Manager adds the returned records
to the Common Customer Directory,codes them with LSIDs,
and sends the coded records to a flat file.
4.Customer Integration Manager sends the output fileswith coded records to the customer database load area for
additional processing.Most records contain a site and
headquarters D-U-N-S Number that will be used for
consolidation. Records that did not match with D&B are
consolidated using the LSID.
Interactive Customer Lookup During Order Entry
A company's order entry system uses Customer
Integration Manager to determine whether an order is from
an existing customer and if credit is available.This firm
has not changed its operational systems to store the LSID
so it must use the Common Customer directory as a
cross-reference table.
1. The call center receives a telephone call from a cus-
tomer wishing to place an order. The caller knows his
firm has ordered before, but does not have an account
number available. The order entry agent enters the com
pany name and address into the order entry system and
selects a 'customer search' option. The order entry sys
tem sends the name and address to the Customer
Integration Manager API via Java. Customer Integration
Manager parses and standardizes the input and looks for
matches in the CommonCustomer Directory. It finds
several that exceed the specified confidence threshold.
2. Customer Integration Manager sends the nameaddress, telephone number and LSID of each matched record
to the order entry system.The order entry system displays
the records to the agent,who reviews them with the caller
There is no match, so the agent asks for an alternate
address and searches for matches on those.When the rightrecord is located,the order entry system issues a command
for Customer Integration Manager to add the original address
to the Common Customer Directory and link it to the alter
nate address.This will allow the system to identify the com
pany automatically if a future caller uses either address.
3. The order entry system issues a direct SQL query to the
Common Customer Directory,selecting all records with
the specified LSID that represent credit system accounts
There are several such records, relating to different
accounts that the customer has established for different
departments at the same site.
17
Consolidated Sales Data
D&B Remote
1
Multi-sourceData Extract
CustomerDatabase Load
3a3b2
4
CodedRecords
D&BDatabase
Customer Integration Manager Technical White Paper
Customer Integration
Manager
Directory
CommonCustomer
8/2/2019 Search DnB
18/20
4. The order entry system extracts the account numbers
from the returned records and submits them to the cred-
it system. The credit system evaluates these accounts
and returns an approval to the order processing system.
The agent completes the order.
Implementation
Every Customer Integration Manager installation is
tailored to the customer's requirements with help from
a dedicated team of D&B consultants. Most imple-
mentations include the following stages:
Install the Customer Integration software.Preliminarysteps include establishing a home directory, setting data-
base access rights, and installing Java Virtual Machine
(Java 2 SDK Standard Edition version 1.4.1). The user
then runs an automated installation Wizard that will
The Graphical User Interface (GUI) lets the user easilyset up and manage operations and administrative
functions. D&B consultants will provide advice on proper
settings. The GUI controls technical details like directory
locations, and business decisions such as which records are sent
to D&B for remote resolution. Users create
visual workflows to define matching, file processing,
input and output formats, remote resolution, logging and
Set up the Common Customer Directory.D&B consul-tants will help to build the data tables,design special views
and indexes, and determine which records to load. The
initial records will be selected from the D&B master
database. This process also includes matching your
specific records against D&B data, and conducting manual
and remote resolution as needed.Customer Integration
Manager uses a relational database with the
following categories: matching, data append, and
operational. The Common Customer Directory is a set of
relational database tables with predefined data
Develop interfaces with specific applications. Details
will depend on the situation, but tasks will generally
include researching existing data sources and systems,
developing extract programs for batch feeds, modifying
online systems to communicate with the Customer
Integration Manager APIs, and making other changes
needed to use Customer Integration Manager outputs.
Implementation time can vary from a few hours for sim-
ple batch matching to several days for a process includ-
ing remote connections to D&B. Projects requiring
changes to the client's in-house systems may take
longer.
Monthly Update
D&B provides Customer Integration Manager clients
with a monthly update for D&B records in the Common
Customer Directory. This can be sent electronically or by
tape, CD, orother physical media. The update file will include
changed records for the primary Common Customer
Directory table which contains matching data,and complete
replacements for the otherD&B database tables.
Interactive Customer Lookup
18
1OrderProcessing
System
4
Conversation with Customer
Credit System
2
3
Customer Integration ManagerTechnical White Paper
Customer Integration
Manager
Common
CustomerDirectory
prompt the user for the required information.
notifications.
structures for matching.
8/2/2019 Search DnB
19/20
Customer Integration Manager includes a monthly
update utility with scripts to control the process.This script
drops indexes and tables and later recreates them,records
each step in a log file,and notifies an operator via email when
the process is complete or an error occurs.The process can
restart from the point of failure if necessary.The system
stores detailed information on changes to individual
records, with separate files for records that have been
added, updated and deleted.
SupportD&B professional services staff supports Customer
Integration Manager clients during and after
implementation. Technical assistance is provided by email,
telephone, and in person. Clients receive a prompt response
during normal business hours.Training is tailored to each
implementation . D&B usually creates a custom training
system with the client's own data. Most training is
conducted at the client's office.Clients also receive detailed
written documentation.
Summary
Customer Integration Manager provides a comprehensive,
flexible solution for business customer data integration.
It combines sophisticated matching technology, your
organization's internal customer knowledge, and valida-
tion against D&B data, to produce the most thorough
relationship identification available. It supports both
batch and online processes, allowing a single system to
serve all customer data integration requirements. It can
provide a central cross-reference file to link account
numbers in different source systems,or generate a single
customer ID for all systems to use internally. It combines
the efficiency of in-house operation with the accuracy of
up-to-the-minute external validation data. The system is
built on industry-standard databases, hardware and
Java APIs, making it simple to integrate with
existing corporate infrastructures. Perhaps most
important, it links your data with the worldwide
resources of D&B, whose D-U-N-S Number opens the
door to an unmatched range and depth of business
customer information.
19Customer Integration Manager Technical White Paper
Monthly Update Process
D&B adds, changes,deactivates
Monthly Update Utility
Old CommonCustomer Directory
New CommonCustomer Directory
8/2/2019 Search DnB
20/20
103 JFK Parkway, Short Hills, New Jersey 07078Customer Information Management, Marketing Groupcim.dnb.com
Risk Management Solutions
Sales & Marketing Solutions
Supply Management Solutions
E-Commerce Solutions
D&B Solutions
www.dnb.com
Decide with Confidence