TARGETTED SEQUENCE ANALYSIS

TARGETTED SEQUENCE ANALYSIS

When you want to decide whether a sequential event influences a binary event you may use this analysis. As an example you may want to realize churn paths by looking at the customers’ complaint memos. For example customers who first complains about his/her bill, then complains about call center agent and lastly complains about price may be more likely to churn. So a trigger strategy may be applied to these customers at the last point (when he/she complains about price) by giving a discount etc.

For example the data is like this:

CUSTOMER_ID MEMO_DATE MEMO CHURN
1 01.01.2011 BILL PROBLEM 1
1 02.02.2011 CC AGENT BEHAVIOUR 1
1 05.02.2011 PRICING STRATEGY 1
2 05.10.2010 BILL PROBLEM 0
2 05.12.2010 BILL PROBLEM 0
3 05.02.2011 CC AGENT BEHAVIOUR 1
.... .... .... ...

Processes
1. For Target = 1 sequence analysis is made. (select target = 1, then sequnce analysis)
Results
PATH CONFIDENCE etc... NEW COLUMN NAME
BILL PROBLEM, CC AGENT BEHAVIOUR, PRICING STRATEGY 0.95 A0
PRICING STRATEGY, TECHNICAL PROBLEM 0.92 A1
.... ...

2. Paths are discovered and new binary variables are created.
CUSTOMER_ID A0 A1 CHURN
1 1 0 1
2 0 0 0
3 0 1 1
... ... ...

These operations may be done
in SPSS: by using @OFFSET()
in SAS: by using lag
Accompanying with Case When or IF statement.
3. Now a model can be built by using the last table and our new variables. So it is now possible to decide whether a sequenced path really impact Churn behaviour.
4. A trigger strategy is designed.

Güven Kızıltaş