A. Cards Campaigns Analytics
Debit Card FRM Segmentation (video summary)
- Collated debit card transaction data from multiple sources and generated recency, frequency and monetary variables from this base on a customer-level. Also culled activity and vintage data for further study.
- Performed outlier removal and normalization of these variables based on visual inspection of variable distribution.
- Created elbow plots from iterative clustering procedures (k-means) to identify optimal number of meaningful clusters.
- Validated above clulsters with basic statistics and canonical discriminant plots. Also deployed SAS code to draw Andrews plots for further validation.
- Profiled resultant clusters with radar charts to identify the main differentiating property for each cluster.
- Conducted further analysis by plotting clusters on three dimensions judged to be most significant to better understand cluster behavior.
Campaign Tracking/Monitoring (github link)
- Calculated basic marketing metrics like activity, conversion and lift for credit/debit card campaigns.
- Prepared target and conversion bases to facilitate aforementioned calculations.
- Composed SAS code that automates repetitive processes used for culling and preparing report data.
In-depth Campaign Analysis
Certain business-sensitive campaigns every quarter were selected for deeper analysis. After derivation of standardized metrics, deep dives were conducted for these campaigns. Two notable examples are:
- Credit Card Mobile App Cross-Sell Campaign (github link): The campaign analyzed aimed to cross-sell the bank's mobile app to credit card customers. Post extraction of basic campaign metrics, demographic and vintage splits along with average utilization rate were analyzed to better differentiate responders from non-responders.
- End of Season Sale Campaign (github link):: This campaign involved tie-ups with multiple merchants for credit/debit card offers during a 45 day period. Pre-post analysis on an overall level and on merchant level were calculated to identify profitable merchants. The proportion of inactive customers activated and their activity post campaign were also studied.
B. Liabilities Trigger Campaigns
Triggers are campaigns that are directed to customers when specific events occur in their life-cycles with respect to their relationships with the bank. Major accountabilities shouldered for trigger campaigns analytics are:
T-test algorithm automation (video summary)
- Successful development and deployment of automated macro to perform T-test like-to-like algorithm used to perform outlier treatment. Employed extensively for trigger campaigns to maintain similarity between test and control groups.
- Generation of a representative pseudo-control based on design rules and exclusions that were used to program the trigger
- Deployment of rigorous outlier treatment on test and psuedo-control datasets using T-tests to bring about likeness between the two groups on the basis of an appropriate metric
- Extraction of conversion data from internal databases to map final conversions on the modified test and control data
- Calculation of incremental conversions and lift on the basis of which the economic value of the campaign is established.
C. Loans Reporting
- Development of robust SAS code that is capable of preparing and validating loans data for other teams' consumption.
- Automation of existing Excel-based reports using SAS PROC REPORT code and rigorous macro language.
- Migration of report and data preparation processes from older data sources to new normalized data mart.