2
$\begingroup$

I have data on my customers (age, location, gender) and number of interactions with customer by channel (#calls, #sms, #letters) and data on whether they have bought a product (yes/no).

I would like to predict what combination of interaction (e.g. 3 calls, 2 letters) is the best approach when contacting different types of customers. What would be an appropriate statistical/ machine learning technique to use to calculate this?

Improve this question
$\endgroup$
2
  • $\begingroup$ I have a similar problem and have been looking into probablistic graphs without too much progress. Time between interactions/purchase could also be useful $\endgroup$
    – TBSRounder
    Commented Jan 4, 2016 at 18:27
  • $\begingroup$ Is a purchase a one time event or does a customer potentially make multiple purchases over time? $\endgroup$
    – Paul
    Commented Jan 5, 2016 at 0:06

2 Answers 2

Reset to default
0
$\begingroup$

I would do the following:

First, cluster your customers into several groups, based on age, location, gender etc. Second, for each group, use different combination of interaction as features to predict 'yes/no', and select which combination of interaction performs best.

Improve this answer
$\endgroup$
0
$\begingroup$

This is how you can approach this problem:

  1. Customer Segmentation: Segment each customer as a separate data point, and if possible have another project for clustering the customers depending on various factors (Example: Depending on geographical regions, age, etc). This would help in simplifying the problem statement and narrowing it too.
  2. Now choose the best combination: Now after segmentation, choose the best(average) combination of interaction mechanisms needed to be implemented for customer conversion. The mean would answer you the question: What is the average number of interactions by different media taken by a customer before he/she is converted

Bonus: In addition to calculating the mean, calculating the median also helps, as it helps in fixing skew due to overly compulsive customer reps who interact a bit more frequently than some, and difficult customers. Useful reference

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.