Subscribe to DSC Newsletter

I've recently come across Attribution Modeling and it seems to be the talk of the town in many areas. I tried googling about it, and figured out much about the basic skeleton of what it is.  But what I'd like to know is,

  • How different is it from Market Mix Modeling (MMM) ?!
  • Where is it more applicable that MMM?
  • What are the advantages of one method over the other?

Tags: Attribution, MMM, MMO, Market, Mix, Modeling

Views: 19443

Reply to This

Replies to This Discussion

Hi David!

Thanks for reading the e-book and the question.  First off the Intelligent Response Attribution solution will keep track of campaigns that are delivered and then a customers action after the delivery.  So if multiple emails are sent to a customer in between purchases the solution keeps track of the deliveries and also the campaign these belong to.  If the customer opens and clicks the emails these actions are recorded and the campaigns they belong to as well.  Using a model based approach we can model these interactions of delivered, opened and clicked and that will differentiate from just a delivery.  These actions result in the higher likelihood of a purchase and therefore higher influence in driving that purchase.  That and the fact that there were more touches within email for that campaign will give that campaign more attribution from that sale then the other touches.  This makes sense anyway because the campaign that drove a customer to open and then click should be the better campaign then the one that was just delivered and did not drive an open or click.  The Intelligent Response Attribution solution also identifies what I call channel fatigue.  Basically at what point are the touches before a purchase not providing any more influence.  Using a model based approach identifies thresholds for the channels that identify the typical number of touches that occur before a purchase for different customer segments.  When the threshold is met for a channel there typically isn't much lift from the excess touches. 

Overall using a model based approach identifies the different scenarios of purchases from people just receiving emails then going directly to the site to purchase or doing a bunch of organic searches then clicking on a banner then purchasing.  I hope this answers your question without me giving away too many of the nuts and bolts.  Let me know if you have any more questions!


Thanks Kevin,

Sounds like you do some sort of campaign-level regression modeling to establish appropriate weights to touches and their frequency. There is, therefore, an implicit penalty for touching too frequently if it doesn't lead to sales events. Am I thinking about this correctly?

What about the impact of offline channels? How is it possible to collect enough data points at the campaign level and tie everything together to build that model? In your example above you made reference to online touches that are somewhat trackable via a cookie (ignoring the cross-channel and overall challenges that come with cookies).

Your e-book talks about Intelligent Response Attribution at the channel level, but I'm still not able to get my head around campaign-level performance. Within the same channel, the performance of a campaign is impacted by the creative, offer, audience, et cetera. What am I missing here?

Hey Kevin,

I just added you on linked in and wanted to ask if you had a current link for your eBook?  The one here seems to be broken.

Also, do you have any experience / direction to head in deploying these methods practically in R / Python / (or perhaps even SQL)?


Increasingly when I read about Attribution modeling, I get the feeling it's not even modeling per se. Here's an article that brings out some quick gaps in the method. I'm sure there are some folks who've developed more rigorous regression based allocation methods, but for the most part, it is as elementary as exponential distribution of credit across various channels.


On Data Science Central

© 2021   TechTarget, Inc.   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service