Adobe Analytics / Google Analytics API Speed

Below are some tests comparing the speed of the Adobe Analytics API to the Google Analytics API. To do the comparison, we’ve looked at two sites – one small and one quite large – that are both running both Adobe Analytics and Google Analytics. To keep the comparison, well, comparable, the queries stick to standard dimensions and metrics so that the “same” data is being pulled from both systems. The queries all pull for the last 365 days.

This is a test comparing the speed of the Adobe Analytics Reporting API to the Google Analytics Reporting API. A few notes about that:

• This is focused on the free version of Google Analytics; in the case of the large site, that means sampling comes into play, and, to make it an apples-to-apples comparison, the query uses the anti_sample = TRUE flag in googleAnalyticsR. That’s, basically, a bit of a hack to work around Google Analytics sampling. That shows in the results.

• A separate comparison could be comparing the crunching of Adobe Analytics data feed data to Google Analytics 360 data that has been pushed into BigQuery. That’s wayyyy beyond the scope of this assessment, though.

The current test just runs each query once. Eventually, I may update this to run each query multiple times (and possibly try at different days of the week and times of day) to get a more robust set of data. But, I’m starting simple! I haven’t included the setup code here, as that’s a little tedious. But, I’ve tried to include all of the salient code for transparency purposes.

The Queries

This is certainly something that can be fiddled with a bit, but, for the initial test, I set up two queries:

• Daily unique visitors (users), visits (sessions), and page views (pageviews)
• Visits (sessions) broken down by device type and browser

For now, I haven’t added any segments to either query – either inline (dynamic) or named. But, that may be something to add in the future.

The functions below can then be run on different accounts to make these queries:

## COMPARISON #1: DAILY METRICS
# Adobe Analytics
daily_data_aa <- function(rsid, date_from, date_to){
  aa_data <- QueueOvertime(rsid,
                         date.from = date_from,
                         date.to = date_to,
                         metrics = c("uniquevisitors", "visits", "pageviews"),
                         date.granularity = "day")
}
# Google Analytics
# This includes the "unsampled" flag to ensure more of an apples-to-apples comparison
daily_data_ga <- function(view_id, date_from, date_to){
  ga_data <- google_analytics_4(view_id,
                                date_range = c(date_from, date_to),
                                metrics = c("users", "sessions", "pageviews"),
                                dimensions = "date",
                                anti_sample = TRUE)
}
## COMPARISON #2: DEVICE TYPE + BROWSER AGGREGATION
# Adobe Analytics
ranked_data_aa <- function(rsid, date_from, date_to){
  aa_data <- QueueRanked(rsid,
                           date.from = date_from,
                           date.to = date_to,
                           metrics = "visits",
                           elements = c("mobiledevicetype", "browsertype"),
                         top = c(25,25))
}
# Google Analytics
# This includes the "unsampled" flag to ensure more of an apples-to-apples comparison. One
# extra step is required here to get the data to come back similarly -- specifying the order.
order_ga <- order_type("sessions", sort_order = "DESCENDING", orderType = "VALUE")
ranked_data_ga <- function(view_id, date_from, date_to){
  ga_data <- google_analytics_4(view_id,
                                date_range = c(date_from, date_to),
                                metrics = "sessions",
                                dimensions = c("deviceCategory", "browser"),
                                order = order_ga,
                                anti_sample = TRUE)
}

Small Site First!

We’ll start with a small site. The first query is the daily data for 365 days. A bit of the data itself is included below to show that they two platforms return comparable results. All that is being recorded for the time is the time to actually make and retrieve the results of the query.

Adobe Analytics: Daily Trend - Small Site

# Adobe Analytics
time_daily_aa <- system.time(aa_data <- daily_data_aa(aa_rsid, start_date_year, end_date))

[1] "Requesting URL attempt #1"
[1] "Requesting URL attempt #2"
[1] "Received overtime report."

aa_data <- aa_data %>% 
  select(datetime, uniquevisitors, visits, pageviews) %>% 
  mutate(datetime = as.Date(datetime))
kable(head(aa_data))

datetime	uniquevisitors	visits	pageviews
2016-07-12	1493	1600	2156
2016-07-13	1484	1597	2124
2016-07-14	1527	1608	2163
2016-07-15	1108	1187	1572
2016-07-16	338	365	494
2016-07-17	542	583	779

Google Analytics: Daily Trend - Small Site

# Google Analytics
time_daily_ga <- system.time(ga_data <- daily_data_ga(ga_view_id, start_date_year, end_date))
kable(head(ga_data))

date	users	sessions	pageviews
2016-07-12	1507	1628	2188
2016-07-13	1479	1620	2126
2016-07-14	1505	1596	2132
2016-07-15	1096	1187	1589
2016-07-16	335	365	490
2016-07-17	549	598	786

This query took 13.12 seconds for the Adobe Analytics API and 1.66 seconds for the Google Analytics API.

Now, let’s pull the device type and browser data – also for a year.

Adobe Analytics: Ranked Data - Small Site

# Adobe Analytics
time_ranked_aa <- system.time(aa_data <- ranked_data_aa(aa_rsid, start_date_year, end_date))

[1] "Requesting URL attempt #1"
[1] "Requesting URL attempt #2"
[1] "Received ranked report."

# We're not counting this in the time calculation -- just making it a bit more apples-to-apples with
# Google Analytics.
  aa_data <- aa_data %>% 
    arrange(-visits) %>% 
    mutate(mobiledevicetype = ifelse(mobiledevicetype=="Other", "Desktop", mobiledevicetype)) %>%
    select(mobiledevicetype, browsertype, visits)
kable(head(aa_data))

mobiledevicetype	browsertype	visits
Desktop	Google	284833
Desktop	Microsoft	60961
Desktop	Mozilla	39910
Mobile Phone	Apple	16831
Desktop	Apple	13570
Mobile Phone	Google	13483

Google Analytics: Ranked Data - Small Site

# Google Analytics
time_ranked_ga <- system.time(ga_data <- ranked_data_ga(ga_view_id, start_date_year, end_date))
kable(head(ga_data))

deviceCategory	browser	sessions
desktop	Chrome	284983
desktop	Internet Explorer	50327
desktop	Firefox	39297
mobile	Safari	14351
mobile	Chrome	13939
desktop	Safari	13184

This query took 12.8 seconds for the Adobe Analytics API and 1.59 seconds for the Google Analytics API.

Now, A Much Larger Site

This is a much larger site – primarily an Adobe Analytics shop, but they also have Google Analytics implemented (lightly), so it suits our purposes. Again, the first query is the daily data for 365 days. We’re not going to show the code here, because it’s identical to the code above.

Adobe Analytics: Daily Trend - Large Site

[1] "Requesting URL attempt #1"
[1] "Requesting URL attempt #2"
[1] "Requesting URL attempt #3"
[1] "Received overtime report."

datetime	uniquevisitors	visits	pageviews
2016-07-12	90375	114619	692623
2016-07-13	97333	122789	737202
2016-07-14	94911	119724	708964
2016-07-15	76312	94919	547828
2016-07-16	11084	12459	70326
2016-07-17	9013	10003	48970

Google Analytics: Daily Trend - Large Site

date	users	sessions	pageviews
2016-07-12	92946	122265	711739
2016-07-13	99901	130705	740745
2016-07-14	97663	127825	721988
2016-07-15	78651	101623	564974
2016-07-16	11226	12812	71551
2016-07-17	9145	10250	49598

This query took 22.55 seconds for the Adobe Analytics API and 2.15 seconds for the Google Analytics API.

Now, let’s pull the device type and browser data – also for a year.

Adobe Analytics: Ranked Data - Large Site

[1] "Requesting URL attempt #1"
[1] "Requesting URL attempt #2"
[1] "Received ranked report."

mobiledevicetype	browsertype	visits
Desktop	Microsoft	14624108
Desktop	Google	13528832
Desktop	Mozilla	4192844
Tablet	Apple	1132420
Desktop	Apple	734011
Mobile Phone	Apple	298861

Google Analytics: Ranked Data - Large Site

deviceCategory	browser	sessions
desktop	Chrome	14255211
desktop	Internet Explorer	13774114
desktop	Firefox	4308947
desktop	Edge	1522778
tablet	Safari	1174571
desktop	Safari	754205

This query took 11.5 seconds for the Adobe Analytics API and 80.47 seconds for the Google Analytics API.

Results Summary

Below is a summary of the results.

Query	Adobe Analytics Time (seconds)	Google Analytics Time (seconds)
Small Site: Daily Trends	13.1	1.7
Small Site: Ranked Values	12.8	1.6
Large Site: Daily Trends	22.5	2.1
Large Site: Ranked Values	11.5	80.5

Google Analytics got killed on the ranked data. For a large site, and with the “anti-sampling” flag being used in the query, that is because there were many calls to the API. If the site was a GA360 site with a custom table set up, it would have come back much faster (but custom tables only go back 30 days, so, even with GA360, this may be a tough one.)