FACT SHEET
Richardson, Texas; January 10, 2001

DataGYM Successfully Handles Brazilian Data

Building Database Marketing capability anywhere in the world

When you have less than 12 weeks to get a Database Marketing program initiated in a foreign country with languages, data, names and address lists very different from your Home market, what do you do?

Use DataGYM to reduce chance of building another white elephant, or "re-inventing the wheel" with each new country market:

• Establish a database marketing capability anywhere in the world that can grow with your business
• Fine-tune Home market database marketing process and rollout out in foreign market in 90 days
• Remove or cleanup and recover "dirty" data and addresses to make your process more dependable
• Use analytics and external data from new data sources by recovering data with questionable quality
• Use this equation "New data + old information = triggers for effective communication" in any market, any language, or any foreign name and address list effectively

It started as a simple statement of work from American Express - TRSI to find out if CIANT can deliver a Prospect Database for Brazil using its groundbreaking software called DataGYM, using Portuguese name and address lists. Over following 8 weeks it demonstrated that DataGYM could not only establish a Prospect Database for the Brazil market beating all internal quality benchmarks and report requirements, and database ready for campaign execution.

This was possible due to the specific design and software architecture of CIANT's DataGYM. Some examples of how DataGYM learns and retains its understanding of Portuguese name and address handling is provided in the Appendix. DataGYM is a set of Middleware software tools that enable its users to integrate data from different sources, establish relationships within these data sources and then extract data to drive marketing decision and communication processes.

DataGYM contains over 100 tables defining its runtime operations, business rules and language contexts in the process. This allows DataGYM to handle data in the US, Latin America, Europe, Russia, Asia or Japan with equal ease. The software tools maintain lingual knowledge base that helps it to identify Name and Address components wherever it is and gives it the ability to understand poorly formatted name and address data. DataGYM Database Marketing Suite is designed to deliver solutions around the world.

The DataGYM Database Marketing Suite not only parses and integrates data from different sources into one marketing data repository, it uses probabilistic algorithms to match the different data sources and establish a hierarchy of relationships in the existing data. The resulting database can then be used to determine when and how to communicate with ones prospects and customers.

There are three components in the DataGYM Database Marketing Suite:

DataGYM Consolidation: This component accepts and standardizes all data sources whether sequential or in another database and parses and integrates data from different sources into one marketing data repository.

Facts:

1. Table based parsing engine to parse all critical data elements.
2. Table can handle English to Japanese Kanji data with equal ease.
3. Input data can be validated to prevent garbage-in/garbage out.
4. Address Correction or Address Standardization may be incorporated.

Here are the 5 steps through which Brazil input data was consolidated into the DataGYM Data Repository:

DataGYM Householding - extracts all information from the database that might potentially match an input data/record and saves the relationship knowledge in the database. There are a number of matching algorithms based on probabilistic and lingual references that can be changed and customized based on country of application or on Client's requirements.

Facts:

1. Phonetic use of language to allow spelling distortions and variations.
2. Probabilistic algorithms to determine match probability.
3. First Match/All Match/Best match options.
4. Unlimited number of match hierarchy and unlimited number of match source tables.
5. Nickname tables and "criss-cross" matching allows matching data that might have been mis-assigned.
6. Phonetic Tables can be changed for each market due to language rules.
7. Match data can be given different weights to emphasize particular words.

Here are the 4 steps through which Brazil data was householded to establish relationships between external data and internal data sources.

DataGYM Extraction - With all the relationships establish data can now be extracted from the DataGYM data repository to execute a variety of marketing programs simultaneously.

Facts:

1. Define and maintain an unlimited number of marketing segments on the database.
2. Maintain history of every extraction for response analysis.
3. Use sampling, limit, exclusivity and "dedupe" functions to choose segment survivors.
4. Execute selection based on time or source data.
5. Define specific output formats to send data to different touch points.

Here are the 4 steps through which data from the Brazil database can be used to drive customer touch points or to execute a local marketing campaign.

DataGYM Database Marketing Suite makes the 1-to-1 marketing hopes and dreams a reality, not just in your Home market but also in markets around the world.

Until now, these capabilities existed only in the US, and with "spotty" results in a few local markets around the world. Now DataGYM will not only allow rapid deployment of this capability in any foreign market, it will now enable multinational companies to rollout its best practices from one market to another cost-effectively and rapidly.

Moreover, this patent pending software's web architecture is suitable for "back-office" and Internet-based operations allowing a common view of customers and prospects using any one or more of the web, call center, or "brick and mortar" channels, anywhere in the world.

Like the Database Marketing Suite, the DataGYM software modules can be used for Fraud Management matching purposes "real-time" or can be configured for a variety of web based applications. Examples include Vertical Market Exchanges for credit accreditations and profile preferences; internet interface with wireless and cell phone devices; web site visitor cross-sell; and instant profiling of prospects over web or call center for tailored offerings to maximize Loyalty. In addition, to all traditional data warehousing and data management applications.

For more information, please e-mail info@ciant.com.

 Back to top of page

APPENDIX:

Examples of iterative learning process to handle Address and Name data in Portuguese:

Address Handing

Case 1: The Street Address begins with the street type, followed by street name, street number, apartment

Example: R MANTA 29

From this Address we get R is the street type, MANTA is the street name and 29 is the street number. The apartment is missing.

Case 2: The Street Address has an alpha character not anticipated before

Example: R MANTA N 29

From this Address we get R is the street type, MANTA is the street name and 29 is the street number. We saw a special usage of the initial 'N' that stands for 'Numero'...sometimes 'N' precedeeds the street number. This understanding is codified in DataGYM through parameters.

Case 3: The Street Address has multiple alpha characters and numeric data not anticipated before

Example: R MANTA N 29 L 2

From this address we get R is the street type, MANTA is the street name, 29 is the street number. We find 'L' to stand for Loja and so the apartment number is then identified as 2.

Case 4: Street Address not in any known format.

Example: Q2 LOJA 12

From this Address (we found that street addresses in Brasilia are mentioned using quadrants) we get Q2 is identified as the street name (or it is a significant component of the street) and 12 is the apartment number identified by an apartment type.

Name Handling:

Case 1: The Name field contained many initials

Example: JOSE C B M JORGE

Case 2: There are more than one strings as First Name or Middle Name

Example: ALIETE MARIA A DE MELO

From this Name we get ALIETE MARIA as the First Name and A as the Middle Name

Example: SIMONE A ZAMORA DE M ARRUDA

From this Name we get SIMONE as the First Name and A ZAMORA as the Middle Name

Case 3: Last Name Prefix used to identify the Last Name and Suffixes:

Example: SERGIO P MARTELLO DE FILHO

From this Name we get DE FILHO as Last Name

Example: ANTONIO M Z COSTA FILHO

From this Name we get COSTA as the Last Name and FILHO as the Last Name Suffix

Example: JOSE G DE CASTRO FILHO

From this Name we get DE CASTRO as the Last Name and FILHO as the Last Name Suffix

 Back to top of page