I would like to construct a Hidden Markov model with data about online customer journeys. A well-known concept related to the customer journey literature is the sales funnel. Consumers walk through the funnel and the probability to purchase increases while becoming more engaged. The latent variables from the Hidden Markov model are the following states: disengaged > active > engaged > conversion.

The data can roughly be divided into three parts:

  1. Observed behavior from the consumer while preparing for making a purchase, for example, different types of websites viewed, google search etc.
  2. Whether the consumer makes a purchase at the end of his/her customer journey
  3. Various kinds of ads

I was thinking about a multivariate HMM with observed behavior (1) and whether he or she converts (2) as output variables. Moreover, I would like to add the various kinds of ads (3) as covariates in the transition probabilities, however, I find it difficult to understand/interpret those covariates.

Can anyone explain how to incorporate covariates in the transition probabilities and how to interpret the covariates?

  • $\begingroup$ Maybe you should have a look into the book Hidden Markov Models for Time Series: An Introduction Using R, Second Edition (Monographs on Statistics and Applied Probability, Band 150) written by Zucchini and MacDonald! $\endgroup$ – hermanzegerman Apr 18 '18 at 12:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.