October 2002 Campaign Tracking & Analysis Module (CTAM) ~ Concept Design Hugh McKay
Mar 29, 2015
October 2002
Campaign Tracking & Analysis Module (CTAM) ~ Concept Design
Hugh McKay
2ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
Why CTAM ?
• Post Campaign Analysis and tracking has always suffered from reprioritisation when there have been resource limitations
•The need to remove the ad hoc nature of campaign analysis to regular reporting for efficiencies
•Accessibility of Campaign results by all of Portfolio Management and wider Cards in a consistent format
3ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
What is in CTAM
Accessed through Essbase there will be 4 modules once fully implemented
Phase I - implementation 1/10/02
Base CampaignPortfolio
Phase II – Implementation due ??Complex CampaignNiche
Demographic measuresMore complex key measures
4ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
Phase I implementation
Base CampaignAllows the end user to view campaigns and drill down by dimensions :
Total Campaign, Cell, Offer and compare Mailed to responders to control groups(where available)
Data is available as either tabular or graph time series for key performance Measures
as requested in the BRD
PortfolioAllows the user to view the same key performance measures as for base campaign but at the portfolio level. The portfolio dimensions that can be selected are : product, revolving status, market segment, loyalty, scheme, acquisition channel, customer segment, bill code
By providing this information users can compare campaign performance versus portfolios in addition to controls.
5ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
Monthly Updating
CampaignsData on a campaign becomes available once responders are loaded, a parameter file drives the number months of data presented pre and post campaign. Typically, the campaign will be tracked for 6 months of historic data pre mail date and 2 yrs post mail date. All campaigns will be updated monthly, with data typically available mid month.
PortfoliosPortfolio measures will be updated monthly in line with campaigns going forward. The Statement measures are available for 27 months.The End of Month measures will start from July 2002
NicheTypically 12 months of historic data (where available) will be loaded and subsequently tracked forward according to requirements
6ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
Building a system – Key milestones
1. Design and build key cyclic measures table (AUSSCORE) including initial population of 24 months data and monthly load process - 1 record per account,(1.87m records) - 1923 fields per record - 27 months of rolling data - approx 36K cylinders of data storage ($8k per month)
- approx CPU cost of $60K to load initial data 2. Develop Weekly loading procedures to AUSPROMO3. Develop monthly extract and merge process from AUSPROMO and AUSSCORE
to SAS 4. - approx cost of $30K to extract to date, approx $10K per month ongoing5. Develop Driver file for campaign monitoring (start and end dates)6. Develop in SAS summarised data for Essbase cube load7. Develop Essbase reporting suite8. Load past campaigns to AUSPROMO
7ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
CTAM structure
Market Pulse Campaign Module
(CTAM Phase II)
Contact Selection (Market Pulse, DMIS, SAS)
Selected
Mailhouse & Systems Excludes
Contacted
Response Rules (Market Pulse, DMIS, SAS)
Response
Banking & Demog Data x Campaign / Portfolio x Account
AUSSCORECurrent
AUSSCOREPrevious...
M O N T H L Y
Summarise by Segment / Cell
REPORTINGCustomer Segments x Month x Performance
AUSPROMOAll Customer Contact x Campaign x Cell x
Response
WEEKLY
8ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
Loading campaigns to AUSPROMO
9ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
Campaign selection and contact …
CAMPAIGN BRIEF (Business Case, Extract,
Analysis, Fulfil)
SelectedTarget & Control
FORMAT
Extract targets from Market Pulse / DMIS
Manipulate data in SAS and assign cells
Exclusions – Address,
block code, delinquency
Contacted
Target
Direct Mail
PIF
Statement
SYSTEMSMAILHOUS
E
Triad
AcquireInsuranc
e
SYSTEMSAUSSTMT fields
– Statement mailed, hopper
code
MAILHOUSEReturned mailfile - formatted with required fields
FORMAT
FORMAT
AUSPROMOAll Customer Contact x Campaign x Cell x
Response
WEEKLY
UPDATE
WEEKLY APPEND
Weekly Change Report
BACKO
UT CO
DE
Historical Contacted
Target
Historical SelectedTarget & Control
MAILFILE DATASETS
SELECTION DATASETS
10ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
Extracting and manipulating performance
data
11ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
Collating performance data by segment …
AUSSCOREAccount
dimensionsDemography
Account status
AUSPROMOAll Customer Contact x Campaign
x Cell x Response
AUSSCOREStatement dimensionsSpend,Fees
InterestLoyalty
Autotag records if Campaign analysis required for processing month
Automatically identify Portfolio or Niche membership and extract data
DIMENSION: Campaign / Portfolio / Niche DATA: Banking / Demographic / Time
LEVEL: Account
Ma
tch
e
ach
C
am
pa
ig
n /
P
ort
foli
o /
N
ich
e
reco
rd
in
AU
SP
RO
MO
w
ith
ba
nki
ng
b
eh
avi
ou
r a
nd
d
em
og
ra
ph
ic
pro
file
a
nd
ta
g
with
p
roce
ssin
g
mo
nth
DIMENSION: Campaign / Portfolio / Niche DATA: Banking / Demographic / Time LEVEL: Dimension Statement & EOM
SUMMARISE / AVERAGE FOR EACH DIMENSION
The whole process requires automation, driven by Campaign analysis periods and
Portfolio / Niche segment definitions. It will run automatically when calendar month data is available each month.
12ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
Reporting to Portfolio
Management
13ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
14ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
Control (Type = C) Mailfile (Type = M)
Responded (Y, N or P) ~ Campaign specific
View results
Choose campaign
Choose cell
Choose period
Reporting Campaign
Choose view
15ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
16ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
17ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
18ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
19ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
AUSSCORE The file
Purpose:Predictive Datamart Extracts
Initial subset of around 20,000 accounts Validation of key algorithms pre campaignCTAM ExtractApplication of key algorithms over whole base Scoring and cycling of bad and doubtful debts
Features:Combined account view covering transferred accounts
There is no central account number.Identifies and uses the correct statement
Changes in cycle date can cause statement skipsTransferred before / after statement date
20ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
Preselection process
AUSSCORE Measures by Dimensions
SAS Dataminer produces Algorithm
Extract
20000
Records
Convert to M204 UL
SAS Results
M204 ResultsCompare ResultsAUSSCORE
Results stored
21ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
AUSSCORE The file
File size:Uses 36730 Cylinders 70 fields have 27 months history 33 fields are either Static or Invisible
Design:Monthly update
Single or many recordsSingle record chosen
Multiple or Single OccurrencesDue to potential Data Quality we chose Single
Move monthly 1 to 2 ,2 to 3 etcUser Language only available
22ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
AUSSCORE The file
Features:About 1200 fields are Non Occurs
Scanning time and space used (3600 bytes)
Define fields as occurs fields limit of 743 individual fieldsMultiply Occurs could solve this
Maximum of one page of occurs fields
Define first occurrence as occurs front of recordFast access as the latest is access most often
Using Float 4 could cause rounding errors
Account number is 16 digits cannot be Ordered NumberPacked with DP would save space
Extension records using Prime IRNsIncreased
23ACCOUNT MANAGEMENT TEAM, PORTFOLIO ANALYTICS
AUSSCORE The file