Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Data Model for Database Marketing Activity Alberto Saccardi, Frank Fiocca - NUNATAC Abstract What kind of data do we have collect to optimise C.R.M.? Which data do we need to understand the market and our competitors? Which logical subjects should we consider in a marketing database? How detailed and, how summarised should our data be? How should it be organised? These are a few questions that have to be dealt with to design a marketing database for decision support. We’ll illustrate the basic guidelines to define the correct data model for a powerful database marketing activity. In particular we’ll present some cases we have encountered and tried to resolve. Introduction. Nunatac is a consulting firm, which is made up of specialists in statistics, marketing and information technology. Its current business solutions fall into the following categories: data warehousing, data mining, business reporting and campaign management. In particular we are experts in database marketing activities, comprising data warehousing and data mining to do business. Nunatac focuses on supplying tailor made business solutions for its clients, using the SAS SYSTEM and is certified as a SAS Quality Partner. The most recent projects are in the publishing, manufacturing, banking and insurance sectors. In the first part of this paper we’ll introduce the business problem and what’s necessary for a database marketing activity. In the second part we will discuss the main points which characterise a marketing database model. At the end some concluding remarks. 1. Database Marketing Activity. Some of the most important business questions a marketing manager must answer are: · Who are our customers? · What are they like? · How many behavioural groups are there among them? · Which are the most profitable? · How can we maximise customer lifetime value? Answering these questions in various projects has allowed Nunatac to refine its approach to Database Marketing Activity (DBMA). The DBMA puts the customer at the centre of interest and the approach may be summarised in the following steps. Business question; Defines the goal of the analysis. Data Organisation; Once the business question has been defined, the second step is to collect and to arrange the data needed to solve the problem. In this phase it’s necessary to: · define the logical units · census the data present in the Information System · Transform and aggregate data in to data marts for analysis. Data Analysis; The third step is the analysis of the data using, for example, data mining techniques to find similar groups of customers considering their behaviour not related to a specific phenomenon or to identify the most important attributes which explain a specific phenomenon: e.g. a policy subscription. Test & Implementation; The fourth step is to test the models, which have been estimated, and to implement the rule on the marketing database. Results; The fifth and last step is to measure the results that are obtained using this rule. When possible, we should compare these results with the results obtained from a control sample selected using traditional criteria. What do we need to implement this approach? 1 First; an organised work group with diverse expertise: Marketing, Statistics, and Information Technology. Second; data access and management technology. Third; a data model by which it is possible to organise and to prepare data for data analysis activities: reporting, OLAP and data mining. 2. Marketing Database Model. In this paragraph we’ll list the main elements involved in the construction of the data model. Logical Subject definition; Generally speaking the logical subjects that characterise a marketing database are: · Client · Product · Sales Force · Territory Internal and External data; Once the logical subjects have been identified it is possible to associate: · Information gathered from operational data sources; invoices, payments, claims, etc. that are related to each of the logical subjects. · Relevant external information; differentiated by the level of detail and quality. Physical organisation of data; The internal and external sources supply the environment prepared for the DBMA. This environment is composed of fact tables (detailed and certified data) and subject tables (summarised data related to the logical subjects, e.g. the customer). The main characteristic of the fact tables is that they contain data at the maximum detail: at least one record per variation of the fact considered. In this phase data are not summarised and they are most of all used for extraction or for reporting activities: standard tabulate, OLAP and so on. The main characteristic of the subject tables is that they contain summarised data regarding: · The analysis variables, in data warehousing terminology the facts, e.g. the number of products purchased; · The classification variables, the dimensions, e.g. product type; · The interaction between facts and dimensions; · The fixed time lag for summarising the data. The data at this level are organised for data mining activity, where e.g. the customer is the logical dimension of interest and the data are prepared in data marts where one customer corresponds to one record. This is the standard checklist that should be adhered-to to build a marketing database, but what does all this mean? Our presentation will attempt to highlight how we have put this theory into practice. Alberto Saccardi - Frank Fiocca NUNATAC Via Crocefisso 5, 20122 Milano Italy Tel. +39 02 86996848 Fax. +39 02 89012074 E-mail: [email protected] [email protected] 2