1
$\begingroup$

I recently found a package ChannelAttribution which is pretty cool for attributing the marketing channels used during customer's journey (I exported data from Google Analytics). The package is super simple, and that is the problem.

I have a data set containing records in the following format:

Columns:

Path: A > B > C > D > A > Z
The number of conversions: 12
The conversion value: 12,632 USD

See how it looks in Google Analytics

I would like to attribute the value to each channel using Markov model, however, more in detail so I can understand the whole process behind. Can you recommend any packages, scripts and steps I should take in order to achieve it?

Improve this question
$\endgroup$

4 Answers 4

Reset to default
2
$\begingroup$

Anderl, E., Becker, I., Wangenheim, F. V., & Schumann, J. H. (2014). Mapping the customer journey: A graph-based framework for online attribution modeling. Available at SSRN: http://ssrn.com/abstract=2343077 or http://dx.doi.org/10.2139/ssrn.2343077

This should help to understand how Markov models are being used to analyze multichannel customer path.

Improve this answer
$\endgroup$
2
$\begingroup$

Sergey Bryl' does a fantastic job explaining those here using R - http://analyzecore.com/2016/08/03/attribution-model-r-part-1/

He simulates the data and explains it from scratch.

Improve this answer
$\endgroup$
2
  • 1
    $\begingroup$ This does not answer the question. $\endgroup$
    – Michael R. Chernick
    Commented Apr 16, 2017 at 14:55
  • $\begingroup$ even though it does not answer the question - it does link to an exceptionally good article that explains in detail exactly what the question asks...there is no point or space here to elaborate as much as the (2-part) article. $\endgroup$
    – davidski
    Commented May 2, 2018 at 15:06
0
$\begingroup$

In your typical example

A > B > C > D > A > Z

The package calculates importance of each node by calculating the reduction in total income you generate when you remove each node. 12,632 USD. For example if you remove A, you will lose all your revenue, because users start their journey with channel (node) A. to calcuate the probability you need mode paths like the above including those who didn't convert. hope it helps.

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

After you done a better look in the paper "http://dx.doi.org/10.2139/ssrn.2343077" to understand the topics, If you will use google analytics you will have to create a view and set a session as a goal, this way you'll be able to acquire the path leading to failures too. Other important step is determine the size of the window in days for the analysis.

But what i recommend is a self hosted analytics to gather all clickstream data, like piwik.org.

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.