This client's store was however made in Woocommerce which made things bit complicated. It took about 4 days because we tried several installs that didn't work - first, a purchased module that didn't work and support got us super confused, then a smaller module developed in-house, then I installed the code manually and I suspended the caching module on the pages it was conflicting with..
Eventually, I got around it on the dedicated event pages and the "Add to cart" button, without any modules made by third parties.
Product feed
The second thing I did, in parallel with the Pixel configuration, was to create and install the product feed. There were no surprises here, but we want to mention two extra settings we made compared to a regular feed configuration:
We added more images to each product feed so we could choose between carousel or showcase ads.
We have also set up the product categories the same way they were showing on the site. Because I noticed on other accounts that grouping products by categories they belong to the carousel ads will perform better.
Campaigns:
Then we set up the campaigns. We always start conservatively with new campaigns because new audiences are not yet tested. Here we started with the dynamic remarketing campaign we segmented after the Pixel events. remember this site was getting targeted visits from Google. 
We segmented by Pixel events (actions) because this measures the awareness level of a visitor each step of the way. 
A person who adds products to the cart is more interested in a purchase than a person who just sees a product, and a person who sees a product is more interested in a purchase than another person who leaves the site in the first 5 seconds, and so on
But also how recent the action takes place because a person who adds a product to cart today is more interested than one who performed the same action 30 days ago.
And if we group these two data points, action + timings, we can then design and serve ads that are tailored to their journey and interaction with the product.  
More interested people will see our ads with higher frequency than the ones that are not so interested. This helps a lot with your Facebook ads score. and therefore delivery (Facebook's algorithm takes care of all that)  
Being conservative, we have only set up a general customer acquisition campaign and one retargeting campaign for each bestseller on Google.
All campaigns had medium initial budgets. The ads were dynamic (general) in retargeting and acquisition and with static product images in the bestseller campaigns.
So I set the Pixel, the product feed and the first campaigns. From now own, we were going to gather more data and start scaling optimizations.
Step 3. Scaling
What does it mean to scale an account? It means getting more conversions, as long as the cost per conversion remains profitable, by displaying your ads to more and more people.
For scaling, you need three elements: audience, message (ad) and delivery settings. The actual scaling can be done only in the following order:
                   1, First you need to create a targeted audience, then                   2. You need to create a message (ad) for that audience, then                   3. You have to gradually increase your exposure to that audience
The first two elements form a couple, audience and message. Effective scaling should only be done to the extent that the audience and the message are fully congruent with each other. By "message" or ad, in the context of an online store, we mean an ad promoting a product.
So we promote a product in front of an audience, through an ad. Each posting is a question we ask the person who sees our ad: "Want to have this product?". If many people respond with "Yes" (by buying it) then we know that the product is suitable for that audience, so we have a match between audience and message and we can show the ad to more and more people in that audience. Let's take them one at a time:
1. The audience
The audience I created was from people interested in that product category. I was lucky because for many product categories an exact match interest is not available on Facebook and you'd normally have to test a bunch of indirect interests. 
That is why the first thing to do when setting the audience is to look for people directly interested in that product category. Dedicated audiences always perform better than others. If there is no audience directly interested in the product or product category, your ads will be on to a slow cooking process. 
Regardless of what audiences you create, it is best practice to overlay it with people who historically bought something online before. So we first targeted these people, and only then, as a second layer, other interests or behaviors. so this two go first: 
PIC<<<
2. The Add Copy
We've focused on creating ads that showcase the best selling products, with all their benefits and the best images available on the site. These types of ads perform well in most cases, but it works much better with quality videos, if available. 
So they are a good solution for starters, before creating and testing with video ads. Most of the time if image ads work, video ads will rock. 
So we set up a campaign with few adsets containing different audiences directly interested in our type of products and a separate campaign that presents only the bestsellers.
Did it work? 
It sure did. take a look below: 
PIC<<
In this example, we got 9,200 clicks which resulted in 104 conversions, so a conversion rate of just over 1%. This is the minimum we want from any ad at the testing stage. It is not great but it is above our minimum threshold of 1%.
What is great is the cost per conversion of $13.82 and the ROAS of a bit over 31X. So now we know our audience wants this product and we are on to something big. 
3. Delivery settings
Here we went with the following (conservative) settings:
Budget of $10 per dayDisplay only on Facebook feeds7-day click attribution settings, so we assign conversions no more than 7 days from the last click before conversion. We don't assign a value to impressions though. 
Then we started the campaigns
After we started the campaigns we focused on delivery settings, especially on ad budgets.
To scale the campaigns we looked at what the performance was (Website Purchases, Cost Per Website Purchase and ROAS), how much budget was spent (Amount Spent), and how much budget the ad set had available for spending (Budget) .
Basically three elements must be matched, conversion performance, budget and cost according to the following logic:
If we see profitable conversions and the whole budget is spent then we increase the budget.
Next, we have to ask ourselves, "Were the conversions profitable?" If yes, we ask ourselves, "Did it spent as much as the budget was allowing?" If yes, then we increase the budget. Then we wait for another 7 days conversion window and re-evaluate to make sure results are accurate and constant. 
If the cost per conversion is no longer profitable at this stage or if the ad has spent less than it's budget, we see this as an obvious sign that campaign won't be profitable if scaled hard. So we won't be spending any budget in that direction just to look good. We work on performance only. 
Let's take a concrete example of analysis and optimization:
PIC<<

For this ad set we have conversions at a cost of $28.60 and a ROAS of 15.30. Our customer setted up a target of 10X ROAS, so we are profitable and we can increase the daily budget for this campaign.

We were sitting at $40 per day and the cost was $285.98 per 7 days, so $40.71 per day to be exact. It is clear that we have spent all the budget available in the last 7 days.

So we are getting profitable conversions and we have spent all the available budget. Greenlight for scaling. To get even more conversions we need to increase daily budget. After this action, we will wait a few days and evaluate again.

Let's see what happened next... 
Now, we want to show you the performance of this account from July to October.
July, $9,316 in sales (actually a bit more as Facebook didn't record all purchase events)
PIC<<
August, $37,408 in sales (again, in reality, more)
PIC<<
September, $81,788 (you get it) 
PIC<<
October, $319,392
PIC<<
Here is how it looks on the conversion chart:
PIC<<