Media mix modeling

Question

Hello,

Has anyone created a media mix modeling?  I'm trying to learn about the concept and create a workflow in Alteryx.

Some of the questions I have are:

* I'm trying to find what are the variables that needs to be used for the model.
* When the model can be used during the sales funnel. 
* Can it used with the live data or it works only with historic data.

Anyone who has experience in it, could you please suggest some material or if you can spend 30 mins to kick start my learning, I would really appreciate it?

Regards,

Saravanakumar

BrandonB · Accepted Answer

@Saravanakumar media mix modeling is a complex topic in general and your ability to create and leverage a model depends a lot on the data sources and tracking elements that you have access to.

In the most simple of attribution approaches where a tracking element is leveraged, you can look at last touch attribution to see where people converted from. For example, a specific landing page that ties directly to an advertisement could be leveraged to understand where people are converting. Some companies will go as granular as using specific tracking elements or promotional codes that tie to specific creative messaging as well so they can more accurately understand effectiveness of messaging in addition to the media channels/placements themselves. Then we must investigate the website traffic and conversions that are not explicitly tied to tracking elements to understand why individuals are converting and appropriately attribute conversions where individuals were driven to a website from an advertisement but didn't adhere to the tracking component. This requires some form of lift analysis to understand sales above and beyond regular traffic to isolate true lift from regular business.

To implement a true media mix model, you also need to consider all of the different touch points in the sales funnel where a customer may have interacted with some form of marketing media that influenced their decision. Additionally, different forms of media have different response curves. Someone listening to the radio, for example, likely isn't going to buy a product while they are driving and may purchase something an hour later (just a potential example), where someone sitting on the couch watching TV may open up their smart phone and buy within the next 10 mins (another potential example, not necessarily fact).

As a starting point it may be good to build a straightforward model using the attribution data that you currently have access to and then continue to tweak and improve the model as you dig deeper into how the different media channels interact with each other and how you want to develop your strategy. I know that this isn't a direct answer, but that is because there are nuances between companies.  Therefore the way that a model is built is dependent on a variety of different things which even relate to customer's buying habits as it relates to the kind of product.

Hopefully this at least gets the wheels turning and helps you with the beginning of your journey, but many of the built in assumptions and attribution decisions will be based on the nuances of the data that you are working with.