| Predictive Analytic
| Database Intelligence
Some refer to it as data mining, others know it as KDD (Knowledge Development in a Database), and some use the term CRM (Customer Relationship Management).
At Decision Analyst, we simply talk about Database Intelligence; that is, using every tool available to maximize and exploit the information
and profit potential of marketing databases.
For companies engaged in direct marketing, their customer database is their greatest asset. Exploiting this asset by maximizing knowledge about database customers
(buying habits, perceptions, attitudes, motivations, future purchase intent) and, by extension, about other potential customers—and putting that knowledge
in the most actionable form—is what database intelligence is all about.
Data Cleansing and Data Integrity
Database quality has a direct impact on how much (even whether or not!) intelligence can be mined from a database; our own experience indicates that
the typical customer database, even after list cleansing, is composed of 20%-55% worthless records (duplicates, records with erroneous contact
information, bad addresses, missing data, etc.). This means direct marketing costs for most companies could be reduced (by up to half) just by better
cleaning of the database.
Database intelligence at Decision Analyst begins with making sure the database is as good as it possibly can be. Using a stochastic
partitioning model, the Predictive Analytic Services Team assesses and ranks the viability of each record in the database. Marketing efforts can then be concentrated
on the records with the highest viability, while clean-up and restoration efforts can be focused on the records with lower viability. At the same time, we can help
you develop a plan to keep the new, improved database at a uniformly high level of accuracy as new records are added.
Data Warehouse Design/Architecture
The potential value of a database can be greatly enhanced by the design of the data warehouse. Massive volumes of data are of little value if they cannot
be accessed quickly and manipulated in meaningful ways without backbreaking (or mind-breaking) effort. That’s why, at Decision Analyst, data
warehouse design and architecture are part of database intelligence. With experience on databases ranging in size from a few kilobytes to several terabytes,
Decision Analyst’s Predictive Analytic Services Team has the experience and know-how to design a data warehouse of virtually any size and complexity.
Exploratory Analysis of Data
Intelligent data mining depends on knowing the behavior of the variables that make up each record in the database. That is, the parameters of the variables
(summary statistics such as min, max, mean, standard deviation, etc.), what each variable represents in the real world, which variables are potentially
useful dependent variables, and how variables in the database relate to each other and (especially) to potential dependent variables. Univariate, bivariate,
and multivariate techniques—both tabular and graphical—are the primary techniques here.
Predictive Analytics and Lead Selection
Once the data are well understood, modeling can begin. If there is a stable and robust response variable present, then a logistic regression
or other kind of model is a logical next step, ranking the leads based on likelihood of being in the “responder” (or “payor,”
or “persistor,” etc.) category, or assigning a loyalty or lifetime value score to each record. Depending on the modeling objective and
the available response variable(s), modeling efforts may result in:
- Response models
- Revenue/Profit models
- Retention/Loyalty/Persistency models
- Lifetime value models
- Net lift models
In the case where a marketing history is not available for a given product or product class, latent variable analysis (an extension
of exploratory analysis) can also be used to generate a propensity model.
One of the fundamental and time-honored ways of targeting distinct portions of the marketplace is through segmentation. Segmentations can range from the very
simple, such as divisions of the market by age or income, to the very complex segmentations built on many variables and optimized for the number of segments.
The Predictive Analytic Services Team at Decision Analyst is ready and equipped to fill your database segmentation needs, whether they be strategic or tactical.
Information in a database is of no value if it cannot be extracted and reported promptly and accurately. Decision Analyst has developed powerful
multilevel reporting software with graphics and drill-down capabilities to race important information to decision makers.
Predictive Analytic Services
For more information on Predictive Analytic Services, please contact Rod
Carver, Vice President, Director of Predictive Analytic Services (firstname.lastname@example.org),
or call 1-800-ANALYSIS (262-5974) or 1-817-640-6166.
Additional Resources from Decision Analyst
Predictive Analytic Services
Advanced Analytics Brochures
Predictive Analytic Case Histories