googleAnalyticsR v1.0.0 - GA4 API, Automatic Shiny Dashboarding, Improved UI

Workflows with GA4

Below is an example workflow using the GA4 APIs - the use case here is to analyse the first and last touch of users. It incorporates some of the new features of the API, such as the filters, creating your own dimensions and ordering the results. Personally I find it much easier to use, so it encourages more ‘flow’ in my analysis of the data.

library(googleAnalyticsR)
# ✓ Setting client.id from /Users/mark/clients/my_client.json
ga_auth()
# ℹ 2021-02-21 16:23:02 > Using email from GARGLE_EMAIL env var: [email protected]

# I've forgot my GA4 property Id so lets fetch them
al <- ga_account_list("data")

## Auto-refreshing stale OAuth token.

# ah yes this is my website
my_id <- 206670707

# what metrics/dimensions can I fetch again? Include my custom fields
custom_meta <- ga_meta("data", propertyId = my_id)

## ℹ 2021-04-17 13:57:39 > Metadata for propertyId 206670707

## ℹ 2021-04-17 13:57:40 > Caching new meta data

# last year
last_year <- c("2020-01-01","2021-01-01")

# I want to cross reference first user source and last user source
ga_data(my_id, 
        metrics = "sessions",
        date_range = last_year,
        dimensions = c("firstUserMedium","sessionMedium"))

## ℹ 2021-04-17 13:57:41 > Downloaded [ 29 ] of total [ 29 ] rows

## # A tibble: 29 x 3
##    firstUserMedium sessionMedium sessions
##    <chr>           <chr>            <dbl>
##  1 organic         organic          17557
##  2 (none)          (none)            5750
##  3 referral        referral          3982
##  4 organic         referral          1165
##  5 (none)          referral           951
##  6 (none)          organic            656
##  7 referral        organic            362
##  8 email           email              246
##  9 organic         (not set)          148
## 10 (none)          (not set)          127
## # … with 19 more rows

# what about only users from twitter? First check the sources
ga_data(my_id, 
        metrics = NULL, # to only return unique dimensions
        date_range = last_year,
        dimensions = "firstUserSource")

## ℹ 2021-04-17 13:57:43 > Downloaded [ 100 ] of total [ 193 ] rows

## # A tibble: 100 x 1
##    firstUserSource
##    <chr>          
##  1 (direct)       
##  2 127.0.0.1:17923
##  3 127.0.0.1:19409
##  4 127.0.0.1:20644
##  5 127.0.0.1:21594
##  6 127.0.0.1:23734
##  7 127.0.0.1:23960
##  8 127.0.0.1:25447
##  9 127.0.0.1:25537
## 10 127.0.0.1:28513
## # … with 90 more rows

# oh dear lots of ip address - lets filter those out
ip_filter <- ga_data_filter(!firstUserSource %begins% "127.0.0.1")
ga_data(
  my_id, 
  metrics = NULL,
  date_range = last_year,
  dimensions = "firstUserSource", 
  dim_filters = ip_filter, # apply filter
  limit = -1) # all results

## ℹ 2021-04-17 13:57:44 > Downloaded [ 167 ] of total [ 167 ] rows

## # A tibble: 167 x 1
##    firstUserSource        
##    <chr>                  
##  1 (direct)               
##  2 185.199.111.153:6080   
##  3 34.66.89.215           
##  4 ActiveCampaign         
##  5 ConversionWorksBlog    
##  6 Data_Elixir_222        
##  7 Iterable               
##  8 aboutdata.atlassian.net
##  9 adelachrist.top        
## 10 affecheafe             
## # … with 157 more rows

# twitter filter: list shows 't.co', 'twitter' and 'twitter.com' 
twit <- ga_data_filter(firstUserSource == c("twitter","twitter.com","t.co"))

# did anyone who first arrived by Twitter later come back via other channels?
ga_data(
  my_id, 
  metrics = "sessions",
  date_range = last_year,
  dimensions = 
    c("firstUserSource",
      lastMediumSource = "sessionMedium/sessionSource"), #dynamic dimension
  dim_filters = twit)

## ℹ 2021-04-17 13:57:45 > Downloaded [ 15 ] of total [ 15 ] rows

## # A tibble: 15 x 3
##    firstUserSource lastMediumSource               sessions
##    <chr>           <chr>                             <dbl>
##  1 t.co            referral/t.co                       801
##  2 t.co            referral/code.markedmondson.me       73
##  3 t.co            organic/google                       41
##  4 t.co            (none)/(direct)                       8
##  5 t.co            referral/disq.us                      7
##  6 twitter         social/twitter                        7
##  7 t.co            referral/github.com                   4
##  8 t.co            (not set)/(not set)                   2
##  9 t.co            social/meetedgar.com                  2
## 10 t.co            referral/alexbirkett.com              1
## 11 t.co            referral/linkedin.com                 1
## 12 t.co            referral/mark.shinyapps.io            1
## 13 t.co            referral/markedmondson.me             1
## 14 t.co            referral/rweekly.org                  1
## 15 twitter.com     social/twitter.com                    1

# lets take out self referrals and those coming back by twitter
me <- ga_data_filter(!sessionSource %contains% "markedmondson.me")
nottwit <- ga_data_filter(!sessionSource == c("twitter","twitter.com","t.co"))

# and change the metrics/dimensions
ga_data(
  my_id, 
  metrics = c("engagementRate","activeUsers"),
  date_range = last_year,
  dimensions = 
    c("unifiedScreenName",
      lastMediumSource = "sessionMedium/sessionSource"), #dynamic dimension
  dim_filters = ga_data_filter(twit & me & nottwit), # combine filters
  orderBys = ga_data_order(-activeUsers) # order by activeUsers metric
  )

## ℹ 2021-04-17 13:57:46 > Downloaded [ 40 ] of total [ 40 ] rows

## # A tibble: 40 x 4
##    unifiedScreenName                 lastMediumSource engagementRate activeUsers
##    <chr>                             <chr>                     <dbl>       <dbl>
##  1 Mark Edmondson                    organic/google              1             6
##  2 Welcome to bigQueryR              organic/google              0.5           5
##  3 Launch RStudio Server in the Goo… organic/google              0.5           4
##  4 Google Tag Manager Server Side o… organic/google              0.4           4
##  5 Introducing googleCloudRunner - … organic/google              0.4           4
##  6 Shiny on Google Cloud Run - Scal… (none)/(direct)             1             3
##  7 Five Ways to Schedule R scripts … organic/google              0.8           3
##  8 BigQuery Queries                  organic/google              0             3
##  9 Google Tag Manager Server Side o… (none)/(direct)             1             2
## 10 Turning GA360 BigQuery exports i… organic/google              1             2
## # … with 30 more rows

So for Twitter first time referrals, it seems most people used Google to come back to visit my home page, which is not surprising.

Sending in Measurement Protocol hits and seeing them in real-time

The Measurement Protocol v2 for GA4 is less open and so can’t be spammed. It works using a server key so it seems useful to include in the package. Thank Martijn van Vreeden for the encouragement in making this feature.

googleAnalyticsR includes functions to call the measurement Protocol API, but it itself depends on a smaller library (new on CRAN v0.1.0) dedicated to measurementProtocol hits. This is so you can use the measurement protocol without needing to import all the other dependencies googleAnalyticsR uses, which can be quite heavy as its focused on usability of the data.

Once you have the measurement ID and stream secret then you can send in events associated with a clientId like so:

# preferably set this in .Renviron - take from the webUI
Sys.setenv(GA_MP_SECRET="MY_SECRET")

# your GA4 settings
my_measurement_id <- "G-43MDXK6CLZ"
my_connection <- ga_mp_connection(my_measurement_id)

a_client_id <- ga_mp_cid()

event <- ga_mp_event("an_event")
ga_mp_send(event, a_client_id, my_connection, debug_call = TRUE)

As we also have the real-time API incorporated, it can be quickly verified which hits are going in:

This has wide applications to send in analytics hits with R, such as adding customer segments to GA4 BigQuery exports using R to model the clusters.

It is also compatible with GTM server-side, which can hold a measurement protocol client. Simply change your mp_connection object to point at your GTM endpoint instead:

# custom GTM server side endpoint
my_custom_connection <- mp_connection(
   my_measurement_id,
   endpoint = "https://gtm.example.com",
   preview_header = "ZW52LTV8OWdPOExNWFkYjA0Njk4NmQ="
 )

This holds many possibilities for enriching your data streams.

ℹ 2021-03-05 09:36:46 > MP Request: https://www.google-analytics.com/debug/mp/collect?measurement_id=G-XXX&api_secret=_XXXXXX 
 {
  "client_id": "XXXXXX",
  "non_personalized_ads": true,
  "events": {
    "name": "googleanalyticsr_loaded",
    "params": {
      "r_version": "R version 3.6.3 (2020-02-29)",
      "r_platform": "x86_64-apple-darwin15.6.0 (64-bit)",
      "r_locale": "en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8",
      "r_system": "macOS  10.16",
      "package": "googleAnalyticsR 0.8.0.9000"
    }
  }
}
ℹ 2021-03-05 09:36:46 > Response:  200

ga_trackme_event(debug_call=TRUE, say_hello = "Hello from Denmark!")
...
 {
  "client_id": "XXXXX",
  "non_personalized_ads": true,
  "events": {
    "name": "googleanalyticsr_loaded",
    "params": {
      ...
      "say_hello": "Hello from Denmark!",
      ...
    }
  }
}