Doubts about which MI methods to use (linear regression, segmentation, neural networks)

Question

I apologize for sounding like a silly question, I started studying AI and machine learning this year and I’m still studying about it. The main reason is I want to elaborate a project and I challenged myself to it. It’s hard for me to "understand how to design the model and what to use", I’m reading articles, but even so it doesn’t enter my mind.

Exemplifying what I want to do, have the following tables:

• Table customers.
• Order table for these customers(With the list of products and date they made the purchase).
• Product table.

It is also possible to obtain average sales, best selling products etc...

But let’s say I want to know if a certain propaganda action product directed to a particular group of customers will have what effect on the sales.

Now I’ll tell you what I thought I’d do:

• Take orders from these customers and make an average purchase of these products (per month or day), so I will have information about the amount that these customers buy. I could then make a linear regression of this customer in relation to the product and "predict" his next purchase. And then make a "segmentation?" to divide into groups based on which products tend to buy in the next period, monthly value spent and predict which other products tend to buy "together" with the main product.
• I apply "linear regression?" about customer groups and their buying tendencies, so I see what was the "percentage" of recurrent average sales change by advertising.
• So I can predict what will be the percentage of relevance of a product advertisement to a group of customers.

So what am I missing and what am I hitting? What would be better?

I thought about using neural networks with tensorflow, but I couldn’t make the examples available because most of them are based on image recognition, which is something I’m not interested in right now, and I wanted something more business-like, using bank data like sales. (if you have an example, please)

About linear regression, is it possible to use strings and assign values? let’s say the type "teachers" buy more chalk than the type "students".

If you have any article that can help, thank you! Anyway I will still have to study a lot...