Using the Facebook Marketing API

Sofia Gil May 20, 2019

• Requirements
• Retrieving Data
  • 1. Basic URL
  • 2. Retrieving in a Programmatic Way
  • 3. Total Population broken down by age, gender and country
  • 4. Total Population that match certain characteristics broken down by age, gender and country
• So Far, So Good?
  • Basic Demographic Information
  • Specific Characteristics
  • The Solutions

The aim of this course is to give a quick overview on what it is and how to use the Facebook Marketing API. In this course you will learn how to query and retrieve aggregated data regarding different users' demographic characteristics.

Requirements

1. Have a Facebook account
2. Set up a Facebook Marketing App
3. Obtain the Token and Creation Act of your Facebook Marketing App
4. Check the version of the API that you are using
5. Install the next R packages:
   • tidyverse
   • jsonlite
   • httr
6. Knowledge of R

For the steps 1 to 4 you can check First_Step.pdf, once you have them save them in a txt, we will be using that information for accessing the Facebook API.

If you are not a R user, you can find a python tutorial here.

Retrieving Data

Would you like to check more information about the Facebook Marketing API or about the JSON syntax? Then take a look on HandsOn.pdf.

1. Basic URL

---

First lets try using a browser, replace your data in the next URL:

https://graph.facebook.com/<<vX.X>>/act_<<creation_act>>/delivery_estimate?access_token=<<TOKEN>>&include_headers=false&method=get&pretty=0&suppress_http_code=1&method=get&optimization_goal=REACH&pretty=0&suppress_http_code=1&targeting_spec={"geo_locations":{"countries":["MX"]},"genders":[1] ,"age_min":16, "age_max":24}

2. Retrieving in a Programmatic Way

---

In order to retrieve and transform the data to a data frame we will use the packages tidyverse and jsonlite.

library(tidyverse)
library(jsonlite)

The way we will pass our credentials to Facebook is through the string that we will save in Credentials, so save your token into the variable token and your creation act into act:

token="Your Token"

act="Your Creation Act"

version="vX.X" # replace the X with your values
    
Credentials=paste0('https://graph.facebook.com/',version,'/act_',act,'/delivery_estimate?access_token=',token,'&include_headers=false&method=get&optimization_goal=REACH&pretty=0&suppress_http_code=1')

3. Total Population broken down by age, gender and country

---

Let's set up our initial variables, they will be save in R and then we will concatenate them in a string.

Age1=25
Age2=55

g=1 # 1:men and 2:women

C='"DE"' # Country code

The parameters we will use are in a JSON format, but we will handle them in R through a string:

• age_min: is a value
• age_max: is a value
• genders: is an array
• geo_locations: is a JSON object where country is an array

query <- paste0(Credentials,'&
                targeting_spec={
                "age_min":',Age1,',
                "age_max":',Age2,',
                "genders":[',g,'],
                "geo_locations":{"countries":[',C,'],"location_types":["home"]},
                "facebook_positions":["feed","instant_article","instream_video","marketplace"],
                "device_platforms":["mobile","desktop"],
                "publisher_platforms":["facebook","messenger"],
                "messenger_positions":["messenger_home"]}')


(query_val<-url(query)%>%fromJSON)

Since age_min and age_max are JSON values their input is always a single value, in this case a integer value between 16 and 65, where 65 means 65 and over.

In the case of genders, it is an array, that means that it can receive more than one value, but the values must be the same type (integer, float, character, etc). So, if we want to query the number of women and men that use Facebook, we would have to set genders to [1, 2].

Finally, geo_locations is a JSON object, therefore, it can contain all the JSON objects already described. In this case, we are specifying countries and location_types and both are arrays.

You can find more information about these and other parameters here.

3.1 Exercise

Change the parameters in the code in order to retrieve the next data:

The number of women and men between 20 and 55 years old that live in Spain and Germany and are Facebook users.

4. Total Population that match certain characteristics broken down by age, gender and country

---

The first step is to know the name of all the possible variables that we can query. There are three different classes:

• demographics
• interests
• behaviors

Let's retrieve all the demographics variables:

library(httr)

DF_CHARTICS<-GET(
  
  "https://graph.facebook.com/v3.2/search",
  
  query=list(
    
    type='adTargetingCategory',
    
    class='demographics',
    
    access_token=token,
    
    limit=2000
    
  )) %>%content(as="text")%>%fromJSON%>%.[[1]]

View(DF_CHARTICS)

Now we will prepare a basic query, for this you just need to choose one variable and save the next information:

ROW=1

(TYPE=DF_CHARTICS$type[ROW])
(ID=DF_CHARTICS$id[ROW])
(NAME=DF_CHARTICS$name[ROW])

For targeting populations that match specific characteristics we will use the parameter flexible_spec from the Facebook Marketing API, this parameter is a JSON object. In order to incorporate it to our initial string, we will save the string in the variable CHARTICS.

CHARTICS<-paste0(',"flexible_spec":[{"',TYPE,'":[{"id":"',ID,'","name":"',NAME,'"}]}]')

A basic query including this parameter is:

query <- paste0(Credentials,'&
                targeting_spec={"age_min":',Age1,',
                "age_max":',Age2,',
                "genders":[',g,']',
                CHARTICS,',
                "geo_locations":{"countries":[',C,'],"location_types":["home"]},
                "facebook_positions":["feed","instant_article","instream_video","marketplace"],
                "device_platforms":["mobile","desktop"],
                "publisher_platforms":["facebook","messenger"],
                "messenger_positions":["messenger_home"]}')


(query_val<-url(query)%>%fromJSON)

In the case of the specific characteristics, you can make the next type of queries:

• one characteristics and other:

  '"flexible_spec":[{
      "TYPE_1":[{"id":"ID_1","name":"NAME_1"}]
    },
    {
      "TYPE_2":[{"id":"ID_2","name":"NAME_2"}]
  }]'

• one characteristics or other:

  '"flexible_spec":[{
    "TYPE_1":[{"id":"ID_1","name":"NAME_1"}],
    "TYPE_2":[{"id":"ID_2","name":"NAME_2"}]
  }]'

In the case of OR we need to group by TYPE. Check the next example: People that are travelers OR like soccer OR movies.

  '"flexible_spec": [{ 
    "behaviors": [
      {"id":6002714895372,"name":"All travelers"}
    ], 
    "interests": [ 
      {"id":6003107902433,"name":"Association football (Soccer)"}, 
      {"id":6003139266461,"name":"Movies"} 
    ] 
  }]'

More info here: https://developers.facebook.com/docs/marketing-api/targeting-specs#broadcategories

4.1 Exercise 2

Code the next query: The number of women between 50 and 60 years old that live in Spain that are "Away from hometown" and "Close friends of people with birthdays in a month" and are Facebook users.

4.1 Exercise 3

Code the next query: The number of women between 50 and 60 years old that live in Spain that are either "Away from hometown" or "Close friends of people with birthdays in a month" and are Facebook users.

So Far, So Good?

Let's challenge your understanding on retrieving data. In the next steps you will recreate part of the code that was used for the paper Demographic Diferentials in Facebook Usage Around the World, but just for some of the countries in Country_Codes.csv.

Basic Demographic Information

1. Upload Country_Codes.csv into the R environment.
2. Create a data frame where you will save all the information.
3. Create a nest loop where you can change the next variables in your queries:
   • Country: each country in Country_Codes.csv.
   • Age: 16-24, 25-54, 55-64
   • Gender: female and male

If you already have the steps 1 to 3, then you will notice a problem. What problem are you encountering?

Specific Characteristics

Now we are going to restrict the population to those that match specific characteristics:

• Away from hometown
• Close friend of users with birthdays in a month

The Solutions

You can find the complete code for replicating the Demographic Diferentials in Facebook Usage Around the World here.