1
$\begingroup$

I have a large data set of sales leads that are in the form of a lead_id, a sequence of binary integers that denote the order of emails and phone calls made to a sales lead, and the binary outcome of that lead (whether the lead was converted to a sale or was lost).

As an example, rows may look like:

LEAD_ID- 1 , PHONE/EMAIL_ORDER- [01101001], OUTCOME- 0

LEAD_ID- 2 , PHONE/EMAIL_ORDER- [01110], OUTCOME- 1

My goal is to find the sequence of emails and calls that maximizes the probability of getting a sale while minimizing the cost of emails/ calls. Does anyone have any ideas or algorithms to use that could output the best sequence given these parameters?

Improve this question
$\endgroup$
1
  • 2
    $\begingroup$ A simple logistic model would be to plot success vs. # of customer touches (email or phone call.) From this, you'll find if there is an optimal number over all. Then I would consider a set of conditionals: conditional on $n$ touches, what is the optimal number of phone calls, etc? I would then use the $b$-bit encoding, e.g., $101$ for $ b=3$ to specify nominal variable pattern ids and to see if there is a pattern that is statistically significant in a) the last 3 touches before winning/losing the sale or b) the first 3. Lots of things to try here! Take care not to over-fit! $\endgroup$
    – Peter Leopold
    Jul 26, 2019 at 20:46

0

Reset to default

Your Answer

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