Explanation
Every time you use databases, you need to make sure you segment your database first, depending on the requirements. In this case, you need to find the customers with the best lifetime value.
You DON'T want to upload all of the databases, as you won't be able to segment on Facebook.
Once you've segmented the database, you should proceed to upload the database to build a custom audience.
This should allow you to use this audience as a seed audience. Because you've already segmented the database; you know that these are the users with the highest LTV.
You then use this to build a similar audience optimized for similarity, as you are trying to reach potential new customers that have the most similarity with your client's customer base.

Explanation
You should never optimize conversion campaigns unless you have at least 500 events. The number of events you are getting at each stage of your funnel are key here:
* 4,000 Lead Events => you can run conversion campaigns to new audiences.
* 300 Add To Cart Events => you need to run traffic campaigns to custom audiences (web visitors).
* 121 Initiate Checkout Events => you need to run traffic campaigns to custom audiences (web visitors).
* 48 Purchase Events => you need to run traffic campaigns to custom audiences (web visitors).
Therefore, the only two strategies that apply in this case is to run conversion ads to new audiences to the "Lead Event" and traffic ads to people who have stayed at one of the other events to move them ahead in your client's online sales funnel.

Explanation
The only two bidding options that allow you to truly predict impression, reach and frequency is the following:
1. Reach and frequency
2. TRP buying
With all of the other options, you cannot predict results. Even when optimizing a reach campaign through the auction, you won't necessarily get the results expected since you are entering the Facebook auction.
Here is a summary graphic to understand which options to use based on the predictability outcome expected:

Explanation
* Choose an audience size close to one to optimize for similarity.
* Since this is around 1-5% of the total population of the country you are targeting, your Lookalike Audience is likely to be small, with a high resemblance to your seed.
* If you are optimizing for reach, you ll want to choose a number near 10
* This Lookalike Audience will be larger, but less similar to your seed, since you're targeting up to 10% of the total population.
This example would create 3 Lookalike Audiences:
* 0%-1% of the population, which will be the most similar audience to your seed.
* 1%-2% of the population, which is second most similar, but a bit wider reach.
* 2%-5% of the population, which is the least similar but the largest reach