Analyzing Facebook campaigns performance with R

This is a test assignment I’ve done a while ago. The client is a media company producing healthcare and fitness apps. During summer 2018, they were running Facebook campaigns to increase their apps installs worldwide.

Given the 3-month historical data, and approximate financial metrics (ARPU = $4 and ARPPU = $50), I was asked to analyze campaigns’ performance and give some recommendation on which campaigns / ads / countries should be scaled and which should be stopped to increase ROI.

Helicopter view on the campaigns’ performance

Summer marketing activities were successful: for $3.4M we got 1.1M installs which led to $4.5M in revenue, which is a 30% return on investment.

To analyze campaigns’ performance, let’s look at them in two dimensions: efficiency and scalability:

The closer a campaign to the upper-right corner, the more scalable it is and the better return on investment it yields. Conversely, the closer a campaign to the bottom-right corner, the more money it loses.

We can see that two campaigns — vik-vid-23 and vik-vid-WW — are clearly standing out: with a total spend of over $1.1M they were able to yield over 80% return on investment.

If we take a look at the vik-vid-WW’ ads performance, we’ll see the same 80-20 pattern: there are 3 best-performing ads (one of them is paused for some reason) generating majority of revenue:

The second category of campaigns are those with total spend between 50K and 300K:

Here we can see a bunch of campaigns that are losing money. For example, rud – vid – tests old (the rightest one), has spent almost 300K with ROI of -6.5%.

Among rud – vid – tests old’s ads, there are two barely profitable ones, and four that constantly losing over 25% of its cost:

Those ads should be stopped.

Finally, there are smaller campaigns which spend less than $50K:

Campaigns with negative ROI should be further investigated.

CPA vs ARPPU analysis

Another way to look at campaigns’ efficiency is to compare its CPA to ARPPU:

Campaigns with CPA > ARPPU should be further investigated.

Campaigns’ performance by country

Finally, we can take a look at performance by countries:

We can see that the top 4 countries by marketing budget — US, GB, CA, and AU — have slightly negative ROI. 88% of the countries are profitable.

It doesn’t mean we should stop advertising to those top 4 countries though.

Different countries have different CPI and ARPU. In the calculation, we consider different CPIs, but imply the same ARPU for all countries (which is not true).

That’s why countries with low customer acquisition cost seem so profitable here, and conversely, countries with high acquisition cost seem to be losing money.

 114   2019