CV/SO Community: I am probably skirting (or crossing) the line of the preference for questions that can be answered vs. those that can (only) be discussed.
That said, I'm trying to wrap my head around various approaches to what is variously referred to as:
- Marketing mix model(s/ing)
- Market response model(s/ing)
- Sales response model(s/ing)
- Drivers analysis
- Due-to analysis
For ease, I will refer to these as Marketing Mix Models (MMM). I have seen numerous mentions of the following approaches to MMM, but very few (i) hands-on, (ii) empirical, (iii) side-by-side, (iv) reproducible comparisons on (v) the same data in (vi) the public domain:
- Regression (of various flavors)
- Bayesian approaches: most resources are consumer-based, like Rossi's "Bayesian Statistics and Marketing"; see below.
- Systems of equations
- Attribution modeling (usually digital)
- Time series analysis (this is forecasting (?))
A lot of vendors have approaches, but for obvious reasons aren't sharing them. I am primarily interested in understanding what (marketing) inputs caused a change in Sales and Share in the past, and being able to create a What-If tool to play around with different levels of inputs (that is, be interpret-able). I'm less interested in forecasting a trend, although I'm willing to listen to arguments in favor of TSA.
The inputs can be construed broadly to be traditional mass media and promotions, newer digital media, and even operational inputs like longer opening hours.
The data should be real "passively-collected" data / data exhaust, NOT panel/cross-sectional qualitative data (although I am a fan of qualitative data).
- Before I recreate the wheel, could somebody point me to hands-on, empirical, reproducible implementations of the various approaches to MMM? (note: I'm relaxing the desire for side-by-side and same data!)
- If the answer is "no", who is interested in working on this collaboratively? I don't mind doing the work, but would be very interested in gathering a "team of collaborators" to review and constructively critique the work.
- Does anybody have a recommendation for a publicly available dataset?
Thanks in advance for thoughts, suggestions, and recommendations. Finally, I'm an R guy, but have access to SAS.
P.S. To avoid confusion, I'm NOT speaking about the following:
- NOT mixed effect modeling
- NOT mixture modeling (sub-population analysis)
I have also reviewed the following: