1
$\begingroup$

I have a dataset of user demographics (i.e. region, age, gender etc.) and each step of a conversion funnel that each user reaches (i.e. site visit, placing item in cart, checkout page, purchase).

I want to understand if there are certain features of a user that are associated with them reaching each stage of the funnel. From what I've gathered, I should be using a classification model? To do that, I created a modified dataset where I have a single row for each user with their demographic data and then a column for each stage of the funnel with a 1 if they reached the stage or 0 if they didn't.

Do I run a separate classification model for each of the funnel step columns with all of the demographic variables and then assess which variables have a statistically significant impact on each step of the funnel? Is it ok to use all of the data rather than testing/training with it? I'm guessing my approach is overly simplified but am I on the right path? What else should I do? Thanks!

$\endgroup$

1 Answer 1

0
$\begingroup$

I think it would be sufficient to model the probability of passing through the funnel conditional on having passed through the previous step (e.g. probability of placing an item in the cart conditioned on you coming to the site, or the probability of purchasing conditional on having put something in your cart). In principle, this should mean running a logistic regression at each step adjusting for any covariates you intend to adjust for.

Do not split the data into train and test, this is an estimation problem not a prediction problem per se.

$\endgroup$
1
  • $\begingroup$ Hi Demetri, thanks for your reply. So it sounds like I was on the right path. To clarify further: I'm removing any numerical variables that have a correlation greater than 0.75 (based on some tutorial I found), and then for each stage of the funnel, I run a logistic regression (glm binomial regression in R), and I look at the statistically significant coefficients to decide which variables were meaningful in assessing users who reach that stage. Is that the right way to go about it? $\endgroup$ Nov 28, 2021 at 21:32

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.