An introduction to Google's Lightweight MMM & code walkthrough

We discussed about MMM using Facebook's Robyn in our previous videos.

In this video, Virendra Singh Shekhawat, Senior Performance Marketing Manager at Rocketship HQ, dives into MMM using Google's Lightweight. Lightweight takes a Bayesian approach to building an MMM.

Related links:

-----

Check out our Robyn series here:

**

Check out our podcasts:
The Mobile Acquisition Show

**

Get more mobile user acquisition goodies here - or engage us to run your mobile user acquisition activity:

**

[TylerRich-mstd]

Thank you! How would I go about creating arrays if I have multiple geos?

[ankushdeshmukh]

This contribution is for train data records only? How can we have the contribution for test data records also

[tomaspapantos]

Hi! I would like to get the predicted values from plot_model_fit, how can I get those? I've tried using he predict method also doing the inverse_transform over the values but the output is really strange (multiple negative sales amounts which make no sense).

[qadeerahmed]

Thank for this great video and resources shared.
The only question I have is that why do I have different results given I am using the same data input and parameters. Is this normal and if so, how can one be confident with the results and recommendations

[rajeshghosh]

Hi, great job. Can you tell how to get the contribution if each variable including the intercept

[saitarunyadalam]

Hi there. Thank you for sharing this explanation. How do I leverage the costs parameter in case I'm using impressions/clicks instead of the spend as media variable? Do I use the average CPMs or total spend for the costs?

[eurojourney]

I watched the entire video and I must say, great job! having said that, as mentioned before, it would be great to have the code you explained in the video, I can't see the link you mentioned. Now, in regards to the code, which I was able to replicate to some extent, I am having issues trying to run lightweight without any issues with the Matplotlib library. Forcing to install version 3.1.3 is not resolving the issue, in fact, it prints a message to upgrade to 3.6.1, which, once I updated, it also prints an error message for a different reason and indicates to revert to 3.1.3. It is quite confusing to be honest, lightweight seems to have several conflicts with other Python libraries. Also, going back to the video, would you be able to elaborate a bit more in the "optimisation" process? I really appreciate the time you put into developing this content, thank you sir!

[eurojourney]

I can't see the code you indicated in the video.