Getting Started: Adobe Analytics Clickstream Data Feed

“Well, first you need a TMS and a three-tiered data layer, then some jQuery with a node backend to inject customer data into the page asynchronously if you want to avoid cookie-based limitations with cross-domain tracking and be Internet Explorer 4 compatible…”

Blah Blah Blah. There’s a whole cottage industry around jargon-ing each other to death about digital data collection. But why? Why do we focus on tools, instead of the data? Because the tools are necessarily inflexible, so we work backwards from the pre-defined reports we have to the data needed to populate them correctly. Let’s go the other way for once: clickstream data to analysis & reporting.

In this blog post, I will show the structure of the Adobe Analytics Clickstream Data Feed and how to work with a day worth of data within R. Clickstream data isn’t as raw as pure server logs, but the only limit to what we can calculate from clickstream data is what we can accomplish with a bit of programming and imagination. In later posts, I’ll show how to store a year worth of data in a relational database, storing the same data in Hadoop and doing analysis using modern tools such as Apache Spark.

This blog post will not cover the mechanics of getting the feed delivered via FTP. The Adobe Clickstream Feed documentation is sufficiently clear in how to get started.

FTP/File Structure

Once your Adobe Clickstream Feed starts being delivered via FTP, you’ll have a file listing that looks similar to the following:


What you’ll notice is that with daily delivery, three files are provided, each having a consistent file naming format:

  • \d+-\S+_\d+-\d+-\d+.tsv.gz

This is the main file containing the server call level data

  • \S+_\d+-\d+-\d+-lookup_data.tar.gz

These are the lookup tables, header files, etc.

  • \S+_\d+-\d+-\d+.txt

Manifest file, delivered last so that any automated processes know that Adobe is finished transferring

The regular expressions will be unnecessary for working with our single day of data, but it’s good to realize that there is a consistent naming structure.

Checking md5 hashes

As part of the manifest file, Adobe provides md5 hashes of the files. There are at least two purposes to this, including 1) making sure that the files truly were delivered in full and 2) that the files haven’t been manipulated/tampered with. In order to check that your md5 hashes match the values provided by Adobe, we can do the following in R:


#Read in Adobe manifest file
manifest <- read.table("zwitchdev_2015-07-13.txt", stringsAsFactors=FALSE)
names(manifest) <- c("key", "value")

#Use digest library to calculate md5 hashes
servercalls_md5 <- digest("01-zwitchdev_2015-07-13.tsv.gz", algo="md5", file=TRUE)
lookup_md5 <- digest("zwitchdev_2015-07-13-lookup_data.tar.gz", algo="md5", file=TRUE)

#Check to see if hashes contained in manifest file
servercalls_md5 %in% manifest$value #[1] TRUE
lookup_md5 %in% manifest$value #[1] TRUE

As we can see, both calculated hashes are contained within the manifest, so we can be confident that the files we downloaded haven’t been modified.

Unzipping and Loading Raw Files to Data Frames

Now that our file hashes are validated, it’s time to load the files into R. For the example files, I would be able to fit the entire day into RAM because my blog does very little traffic. However, I’m going to still limit the rows brought in, as if we were working with a large e-commerce website with millions of visits per day:

#Get list of lookup files from tarball
files_tar <- untar("zwitchdev_2015-07-13-lookup_data.tar.gz", list = TRUE)

#Extract files to _temp directory. Directory will be created if it doesn't exist
untar("zwitchdev_2015-07-13-lookup_data.tar.gz", exdir = "_temp")

#Read each file into a data frame
#If coding like this in R offends you, keep it to yourself...
for (file in files_tar) {
  df_name <- unlist(strsplit(file, split = ".tsv", fixed = TRUE))
  temp_df <- read.delim(paste("_temp", file, sep = "/"), header=FALSE, stringsAsFactors=FALSE)
  #column_headers not used as lookup table
  if (df_name != "column_headers"){
    names(temp_df) <- c("id", df_name)
  assign(df_name, temp_df)

#gz files can be read directly into dataframes from base R
#Could also use `readr` library for performance
servercall_data <- read.delim("~/Downloads/datafeed/01-zwitchdev_2015-07-13.tsv.gz",
                       header=FALSE, stringsAsFactors=FALSE, nrows = 500)

#Use column_headers to label servercall_data data frame using first row of data
names(servercall_data) <- column_headers[1,]

If we were to be loading this data into a database, we’d be done with our processing; we have all of our data read into R and it would be a trivial exercise to load the data into a database (we’ll do this in a separate blog post). But since we’re going to be analyze this single day of clickstream data, we need to join these 14 data frames together.

SQL: The Most Important Language for Analytics

As a slight tangent, if you don’t know SQL, then you’re going to have a really hard time doing any sort of advanced analytics. There are literally millions of tutorials on the Internet (including this one from me), and understanding how to join and retrieve data from databases is the key to being more than just a report monkey.

The reason why the prior code creates 14 data frames is because the data is delivered in a normalized structure from Adobe. Now we are going to de-normalize the data, which is just a fancy way of saying “join the files together in order to make a gigantic table.”

There are probably a dozen different ways to join data frames using just R code, but I’m going to do it using the sqldf package so that I can use SQL. This will allow for a single, declarative statement that shows the relationship between the lookup and fact tables:


query <-
browser.browser as browser_name,
connection_type.connection_type as connection_name, as country_name,
languages.languages as languages,
resolution.resolution as screen_resolution,
from servercall_data as sc
left join browser on sc.browser =
left join browser_type on sc.browser =
left join connection_type on sc.connection_type =
left join country on =
left join javascript_version on sc.javascript =
left join languages on sc.language =
left join operating_systems on sc.os =
left join referrer_type on sc.ref_type =
left join resolution on sc.resolution =
left join search_engines on sc.post_search_engine =

denormalized_df <- sqldf(query)

There are three lookup tables that weren’t used: color_depth, plugins and event. The first two don’t have a lookup column in my data feed (click link for a full listing of Adobe Clickstream data feed columns available). These columns aren’t really useful for my purposes anyway, so not a huge loss. The third table, the event list, requires a separate processing step.

Processing Event Data

As normalized as the Adobe Clickstream Data Feed is, there is one oddity: the events per server call come in a comma-delimited string in a single column with a lookup table. This implies that a separate level of processing is necessary, outside of SQL, since the column “key” is actually multiple keys and the lookup table specifies one event type per row. So if you were to try and join the data together, you wouldn’t get any matches.

To deal with this in R, we are going to do an EXTREMELY wasteful operation: we are going to create a data frame with a column for each possible event, then evaluate each row to see if that event occurred. This will use a massive amount of RAM, but of course, this is a feature/limitation of R which wouldn’t be an issue if the data were stored in a database.

#Create friendly names in events table replacing spaces with underscores
event$names_filled <- tolower(gsub(" ", "_", event$event))

#Initialize a data frame with all 0 values
#Dimensions are number of observations as rows, with a column for every possible event
event_df <- data.frame(matrix(data = 0, ncol = nrow(event), nrow = nrow(servercall_data)))
names(event_df) <- event$id

#Parse comma-delimited string into vector
#Each vector value represents column name in event_df, assign value of 1
for(row in servercall_data$post_event_list){
        for(event_ in strsplit(row, ",")){
          event_df[as.character(event_)] <- 1

#Rename columns with "friendly" names
names(event_df) <- event$names_filled

#Horizontally join datasets to create final dataset
oneday_df <- cbind(denormalized_df, event_df)

With the final cbind command, we’ve created a 500 row x 1562 column dataset representing a sample of rows from one day of the Adobe Clickstream Data Feed. Having the data denormalized in this fashion takes 6.13 MB of RAM…extrapolating to 1 million rows, you would need 12.26GB of RAM (per day of data you want to analyze, if stored solely in memory).

Next Step: Analytics?!

A thousand words in and 91 lines of R code and we still haven’t done any actual analytics. But we’ve completed the first step in any analytics project: data prep!

In future blog posts in this series, I’ll demonstrate how to actually use this data in analytics, from re-creating reports available in the Adobe Analytics UI (to prove the data is the same) to more advanced analysis such as using association rules, which can be one method for creating a “You may also like…” functionality such as the one at the bottom of this blog.

Example Files:

RSiteCatalyst Version 1.4.4 Release Notes

It’s been about six months since the last RSiteCatalyst update, and this update is really just a single bug fix, but a big bug fix at that!

Sparse Data = Opaque Error Messages

Numerous people have reported receiving an error message from RSiteCatalyst similar to the following:

‘names’ attribute [1] must be the same length as the vector [0]

This is about the least helpful message that could’ve been returned, but it was an R message indicating an internal function trying to overwrite the column names vector (which had non-zero length) with a vector of length zero (which is an error in the context of a data frame). Thankfully, Willem Paling was able to squash this bug (hopefully) once-and-for-all; the error occurs when a user tries to do a Queue*` report with multiple breakdowns, where NULL data is returned by the Adobe API for one of the breakdowns.

So hopefully, if you’ve run into this error before (which I have to imagine was quite frustrating), you shouldn’t see this again with v1.4.4 of RSiteCatalyst. Additionally, tests will be added to the test suite to attempt to trigger this warning, so that this horrible monster of a bug doesn’t appear again.

Authentication Messaging

The only other change of substance was to modify the message returned after calling SCAuth(); some users were having issues with API calls not working, after RSiteCatalyst having returned 'Authentication Succeeded' to the console. RSiteCatalyst never actually validates that your credentials are correct, just that they are stored within the session. The console message has been updated to reflect this.

Proper Punctuation Prevents Poor Documentation!

The Eagle-Eyed among you might have noticed that my DESCRIPTION file was out of CRAN spec for many months. This has now been fixed, so that the meaning is as clear as possible.

Feature Requests/Bugs

As always, if you come across bugs or have feature requests, please continue to use the RSiteCatalyst GitHub Issues page to submit issues. Don’t worry about cluttering up the page with tickets, please fill out a new issue for anything you encounter (with code you’ve already tried and is failing), unless you are SURE that it is the same problem someone else is facing.

Contributors are also very welcomed. If there is a feature you’d like added, and especially if you can fix an outstanding issue reported at GitHub, we’d love to have your contributions. Willem and I are both parents of young children and have real jobs outside of open-source software creation, so we welcome any meaningful contributions to RSiteCatalyst that anyone would like to contribute.

Vega.jl, Rebooted

pie donut

Mmmmm, baked goods!

Rebooting Vega.jl

Recently, I’ve found myself without a project to hack on, and I’ve always been interested in learning more about browser-based visualization. So I decided to revive the work that John Myles White had done in building Vega.jl nearly two years ago. And since I’ll be giving an analytics & visualization workshop at JuliaCon 2015, I figure I better study the topic in a bit more depth.

Back In Working Order!

The first thing I tackled here was to upgrade the syntax to target v0.4 of Julia. This is just my developer preference, to avoid using Compat.jl when there are so many more visualizations I’d like to support. So if you’re using v0.4, you shouldn’t see any deprecation errors; if you’re using v0.3, well, eventually you’ll use v0.4!

Additionally, I modified the package to recognize the traction that Jupyter Notebook has gained in the community. Whereas the original version of Vega.jl only displayed output in a tab in a browser, I’ve overloaded the writemime method to display :VegaVisualization inline for any environment that can display HTML. If you use Vega.jl from the REPL, you’ll still get the same default browser-opening behavior as existed before.

The First Visualization You Added Was A Pie Chart…

…And Followed With a Donut Chart?

Yup. I’m a troll like that. Besides, being loudly against pie charts is blowhardy (even if studies have shown that people are too stupid to evaluate them).

Adding these two charts (besides trolling) was a proof-of-concept that I understood the codebase sufficiently in order to extend the package. Now that the syntax is working for Julia v0.4, I understand how the package works (important!), and have improved the workflow by supporting Jupyter Notebook, I plan to create all of the visualizations featured in the Trifacta Vega Editor and other standard visualizations such as boxplots. If the community has requests for the order of implementation, I’ll try and accommodate them. Just add a feature request on Vega.jl GitHub issues.

Why Not Gadfly? You’re Not Starting A Language War, Are You?

No, I’m not that big of a troll. Besides, I don’t think we’ve squeezed all the juice (blood?!) out of the R vs. Python infographic yet, we don’t need another pointless debate.

My sole reason for not improving Gadfly is just that I plain don’t understand how the codebase works! There are many amazing computer scientists & developers in the Julia community, and I’m not really one of them. I do, however, understand how to generate JSON strings and in that sense, Vega is the perfect platform for me to contribute.

Collaborators Wanted!

If you’re interested in visualization, as well as learning Julia and/or contributing to a package, Vega.jl might be a good place to start. I’m always up for collaborating with people, and creating new visualizations isn’t that difficult (especially with the Trifacta examples). So hopefully some of you will be interested in enough to join me to adding one more great visualization library to the Julia community.

  • RSiteCatalyst Version 1.4.16 Release Notes
  • Using RSiteCatalyst With Microsoft PowerBI Desktop
  • RSiteCatalyst Version 1.4.14 Release Notes
  • RSiteCatalyst Version 1.4.13 Release Notes
  • RSiteCatalyst Version 1.4.12 (and 1.4.11) Release Notes
  • Self-Service Adobe Analytics Data Feeds!
  • RSiteCatalyst Version 1.4.10 Release Notes
  • WordPress to Jekyll: A 30x Speedup
  • Bulk Downloading Adobe Analytics Data
  • Adobe Analytics Clickstream Data Feed: Calculations and Outlier Analysis
  • Adobe: Give Credit. You DID NOT Write RSiteCatalyst.
  • RSiteCatalyst Version 1.4.8 Release Notes
  • Adobe Analytics Clickstream Data Feed: Loading To Relational Database
  • Calling RSiteCatalyst From Python
  • RSiteCatalyst Version 1.4.7 (and 1.4.6.) Release Notes
  • RSiteCatalyst Version 1.4.5 Release Notes
  • Getting Started: Adobe Analytics Clickstream Data Feed
  • RSiteCatalyst Version 1.4.4 Release Notes
  • RSiteCatalyst Version 1.4.3 Release Notes
  • RSiteCatalyst Version 1.4.2 Release Notes
  • Destroy Your Data Using Excel With This One Weird Trick!
  • RSiteCatalyst Version 1.4.1 Release Notes
  • Visualizing Website Pathing With Sankey Charts
  • Visualizing Website Structure With Network Graphs
  • RSiteCatalyst Version 1.4 Release Notes
  • Maybe I Don't Really Know R After All
  • Building JSON in R: Three Methods
  • Real-time Reporting with the Adobe Analytics API
  • RSiteCatalyst Version 1.3 Release Notes
  • Adobe Analytics Implementation Documentation in 60 Seconds
  • RSiteCatalyst Version 1.2 Release Notes
  • Clustering Search Keywords Using K-Means Clustering
  • RSiteCatalyst Version 1.1 Release Notes
  • Anomaly Detection Using The Adobe Analytics API
  • (not provided): Using R and the Google Analytics API
  • My Top 20 Least Useful Omniture Reports
  • For Maximum User Understanding, Customize the SiteCatalyst Menu
  • Effect Of Modified Bounce Rate In Google Analytics
  • Adobe Discover 3: First Impressions
  • Using Omniture SiteCatalyst Target Report To Calculate YOY growth
  • ODSC webinar: End-to-End Data Science Without Leaving the GPU
  • PyData NYC 2018: End-to-End Data Science Without Leaving the GPU
  • Data Science Without Leaving the GPU
  • Getting Started With OmniSci, Part 2: Electricity Dataset
  • Getting Started With OmniSci, Part 1: Docker Install and Loading Data
  • Parallelizing Distance Calculations Using A GPU With CUDAnative.jl
  • Building a Data Science Workstation (2017)
  • JuliaCon 2015: Everyday Analytics and Visualization (video)
  • Vega.jl, Rebooted
  • Sessionizing Log Data Using data.table [Follow-up #2]
  • Sessionizing Log Data Using dplyr [Follow-up]
  • Sessionizing Log Data Using SQL
  • Review: Data Science at the Command Line
  • Introducing Twitter.jl
  • Code Refactoring Using Metaprogramming
  • Evaluating BreakoutDetection
  • Creating A Stacked Bar Chart in Seaborn
  • Visualizing Analytics Languages With VennEuler.jl
  • String Interpolation for Fun and Profit
  • Using Julia As A "Glue" Language
  • Five Hard-Won Lessons Using Hive
  • Using SQL Workbench with Apache Hive
  • Getting Started With Hadoop, Final: Analysis Using Hive & Pig
  • Quickly Create Dummy Variables in a Data Frame
  • Using Amazon EC2 with IPython Notebook
  • Adding Line Numbers in IPython/Jupyter Notebooks
  • Fun With Just-In-Time Compiling: Julia, Python, R and pqR
  • Getting Started Using Hadoop, Part 4: Creating Tables With Hive
  • Tabular Data I/O in Julia
  • Hadoop Streaming with Amazon Elastic MapReduce, Python and mrjob
  • A Beginner's Look at Julia
  • Getting Started Using Hadoop, Part 3: Loading Data
  • Innovation Will Never Be At The Push Of A Button
  • Getting Started Using Hadoop, Part 2: Building a Cluster
  • Getting Started Using Hadoop, Part 1: Intro
  • Instructions for Installing & Using R on Amazon EC2
  • Video: SQL Queries in R using sqldf
  • Video: Overlay Histogram in R (Normal, Density, Another Series)
  • Video: R, RStudio, Rcmdr & rattle
  • Getting Started Using R, Part 2: Rcmdr
  • Getting Started Using R, Part 1: RStudio
  • Learning R Has Really Made Me Appreciate SAS