1 optimizing marketing campaigns by the use of data mining methods for the hamburg-mannheimer...
Post on 20-Dec-2015
214 views
TRANSCRIPT
1
Optimizing Marketing Campaigns by the Use of Data Mining Methods
for the Hamburg-Mannheimer Insurance
Die Kaiser-Rente®
Glück ist planbarDie Kaiser-Rente®
Glück ist planbar
Thomas Rauscher - ITERGO Informationstechnologie GmbH
2
Overview:
1. Commercial Goals: Why Data Mining ?
2. Setting up a Data Mining Project
3. Into the Mining Process: Statistical Challenges
4. Doing the Campaigns & Controlling of Results
Br
3
1. Commercial Goals: Why Data Mining?
2. Setting up a Data Mining Project
3. Into the Mining Process : Statistical Challenges
4. Doing the Campaigns & Controlling of Results
Br
4
Why does Hamburg-Mannheimer Insuranceuse Data-Mining-Methods?
Use valuable information from the customer database Better targeting of sales and backoffice activities Customer segmentation
The Projects: 1999/2000 cancelation reduction for life insurance 2001 campaign management for the Kaiser-Rente 2002 recruitment controlling for new agents for HMI
sales organisation from 2001 on: customer selection for several mailings
Br
5
The Basic Concept:The basic idea about the usage of data mining methods is the targeting of valuable customersIn this context ‚valuable‘ means that these customers are likely to respond to a particular offer or activity
6
The Project „Kaiser-Rente® “„Riester-Rente“ = private pension with additional governmental funding (amount of funding based on income and number of kids) „Kaiser-Rente®“ = name of the product offered by the Hamburg-Mannheimer
7
The Target Group for the „Riester-Rente“Governmental funding would be availble for all employees paying social security fees: 30 Million German inhabitants 2,7 Million Hamburg-Mannheimer customers
Doubling the market share in the new market 4% existing market share for classical like insurance 8% expected market share as target for ‚Riester-Rente‘
The Commercial Goal
The Slogan: „Glück ist planbar“ „Luck can be planned“
Br
8
Optimization of Marketing Campaigns for the Kaiser-Rente®
Question:Which customers are most likely to sign a contract for the Kaiser-Rente?
Action: Selection of those customers who must be first contacted for the whole sales organisation (mandatory!) directly after product launch of the Kaiser-Rente Tracking of results, selection of customers for follow-up campaigns
Br
9
1. Commercial Goals: Why Data Mining?
2. Setting up a Data Mining Project
3. Into the Mining Process : Statistical Challenges
4. Doing the Campaigns & Controlling of Results
Br
10
4 Major Campaigns July 2001: 1. Campaign (with product launch) October 2001: 2. Campaign (after product launch) March 2002: 3. Campaign January 2003: 4. Campaign
Each Campaign should cover ~ 300.000 - 400.000 customer contacts
The Big Challenge Whole project was started in February 2001, product launch and the first campaign were targeted to 1. of July 2001.
Campaigns for the Kaiser-Rente®
R
11
Project organization: Who was involved?1 Marketing Expert (Hamburg-Mannheimer)
Modeling and quality control 2 external Programmers
Data management and sampling1 Data-Mining-Expert (ITERGO)
Data mining and scoring1 Programmer (ITERGO)
Customer selection and printing 1 Sponsor (Hamburg-Mannheimer)
• Basis conception and coordination of sales activities
12
Amount of campaign activities (in days)
20
10
0
0
25
10
10
5
35
15
50
15
0 10 20 30 40 50 60
Campaign
Mining
Data Management
Modeling
1.Campaign
2.Campaign
3.Campaign
13
Model 1: First Campaign (with product launch)
One big Problem: No experience, no historical data !The solution: Two particular groups of customers:
R
9.000 Contracts with ‚Anpassungsgarantie‘: Option to change from a classical private pension to the Kaiser-Rente in July 2002 after Certification
2.000 Customers who responded to a mailing with information about the Kaiser-Rente
14
Analysis of first Contracts for the Kaiser-Rente®
from July and August 2001
Process (same as first campaign)
Contract for a Kaiser-Rente
No Contract
30.6. 2001:
Collection of potential predictors from the customer database (sample of total population)
31.8. 2001: Collection of target variable, (Contract Kaiser-Rente) and Sampling
1.9. - 15.10.2001: Data Mining Process
15.10 2001: Scoring for the complete customer database,
Customer Selection for the campaign
R
Model 2: Second Campaign (after product launch)
15
1. Commercial Goals: Why Data Mining?
2. Setting up a Data Mining Project
3. Into the Mining Process:Statistical Challenges
4. Doing the Campaigns & Controlling of Results
Br
16
Technical Environment
Database: HM Customer Database (DB2).
Data Management Tool: SAS Data Selection from DB2 into SAS-Datasets Data Manipulation and Merging Download to a NT-Server for the Data Mining Process
Mining-Tool : SAS- Enterprise Miner automatically generates SAS-Code for scoring of the complete customer database
The complete Workflow was done using SAS-Software
R
18
Quality of Datamost important issue (!) that can only be controlled properly by perfect knowledge or backtracing analysis of data sources
Choice of Method: Regression vs. Tree-Algorithm none of both is dominant in performance. Tree: Needs less variables, easier to interprete for non-statisticians, more robust to outliers Regression: easier to interprete for statisticians, better control about variable selection and multicollinearity For the Kaiser-Campaigns both decision trees and regression were used for different campaigns and subgroups
R
Statistical Challenges
19
Time since last contact to any agent Contacting Sales organization Classical life-insurance-contract (yes/no) Status of contacting sales agent Number of kids Type of Bank account Age
R
Influential Variables
A selection of variables predicting the probabilty of signing a contract for the Kaiser-Rente:
20
1. Commercial Goals: Why Data Mining ?
2. Setting up a Data Mining Project
3. Into the Mining Process : Statistical Challenges
4. Doing the Campaigns & Controlling of Results
Br
21
Product launch for the Kaiser-Rente®
Customer selection for sales contact - Campaign 1: 400.000 selected customers - Campaign 2: 290.000 selected customers defined contact forms printed for the sales agents
Br
23
Target and Control Groups
Campaign 1: 1/3 of customers as control group: random selection regardless of scoring value
Important: Control group of Campaign 1 came to be the base population needed for campaign 2 modeling !
Campaign 2 - 4: 1/5 of customers as control group
Br
24
Results of Campaign 1 (from control group):
Customers in the first percentile had a response rate which was 3.4 times higher than the response for the total population
Percentile of ‚best‘ customers
Ratio of response rate below percentile / total population
R
25
Campaign 1 (Response Rate by Score)
0+ 20+ 40+ 60+ 80+ 100+ 150+ 200+ 250+ 300+ 350+ 400+ 450+ 500+ 550+ 600+ 650+ 700+
% o
f S
old
Co
ntr
ac
ts
Sales Organisation A Sales Organisation B
Average Sales Org. A
Average Sales Org. B
26
Consequences and Results for Campaign 2
The different behaviour of the two sales organization led to the development of different models for those organisations during the mining process for Campaign 2
Results: Again good seperation between high and low score intervals, but: much weaker lift in response rate between target and control group Why ?
Br
27
The ‚Wave‘-Problem
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0+ 50+ 100+ 150+ 200+ 250+ 300+ 350+ 400+ 450+ 500+ 550+ 600+ 650+ 700+
% o
f t
ota
l p
op
ula
tio
n b
y S
co
re-I
nte
rva
l
Campaign I Campaign II Rest
28
Consequences for Campaign 4
Following the original concept Campaign 4 should cover a seclection of those customers who had not been selected for Campaign 1 to 3
Change of Concept: Campaign 4 was focused on recontacting the highest-scored customers from campaign 1 to 3 who had not yet signed a contract for the Kaiser-Rente
Br
29
Conclusions
When using Data Mining in a commercial context, not the statistical quality of modeling and analysis is of primary interest, but three other issues:
Data Quality, good knowledge of data sources
Well defined target variable: What is the question that shall be answered by Data Mining methods?
Well defined actions: What shall actually be done with the results of the Data Mining process?
Br
30
Thanks for your attention !
Contact
Thomas RauscherAnwendungsentwicklung Data WarehouseITERGO Informationstechnologie GmbHÜberseering 35, D - 22297 HamburgTel. (++49) (0)40 6376-6613E-mail: [email protected]
VG-QS/ITERGONovember 2002