Author: Os Keyes
License: MIT
Status: Stable
A well-studied part of web analytics and human-computer interaction is the concept of a “session”: a series of linked user actions. This is used for anything from evaluating the impact of design or engineering changes on users, to providing common, high-level metrics such as time-on-page or bounce rate.
reconstructr is a library designed to efficiently
reconstruct sessions from a series of user events, and then generate
common metrics from that session-based data, including bounce rate,
session length and time-on-page. It features heavy internal use of C++
to make it lightning-fast over datasets containing millions or tens of
millions of events, along with a wide range of options with each
function, allowing you to heavily customise what data is produced and
what data is evaluated. For more information, see the introductory
vignette.
The package is under active development: if you find bugs or have suggestions for new features, please feel free to report them.
So you’ve got a session dataset, which we’ll call, well,
session_dataset. It looks like this:
library(reconstructr)
data("session_dataset")
str(session_dataset)
# 'data.frame': 63524 obs. of 3 variables:
# $ uuid : chr "47dc43895814861e21a2edf93348c826" "a736822df1890011694e7049cb3abef3" "674d2d00e096a3319874a4347caa1f4a" "f62d315398e6d04a3f2fa02e8ae42d49" ...
# $ timestamp: POSIXlt, format: "2014-01-07 00:00:15" "2014-01-07 00:01:11" "2014-01-07 00:01:54" ...
# $ url : chr "https://www.nasa.gov/history/mercury/mercury.html" "https://www.nasa.gov/images/ksclogosmall.gif" "https://www.nasa.gov/elv/hot.gif" "https://www.nasa.gov/facts/faq04.html" ...You have timestamps, you have UUIDs for each user, and you have the
URL (or any other metadata you might need!). What you really want to do
is divide the data up into ‘sessions’ - distinguishable blocks of
browsing activity by a single user. For this we use
sessionise, passing it the dataset, the column names for
timestamps and user IDs, and a threshold - the number of seconds after
which to decide a user has entered a new session. By default this is
3600 (1 hour):
sessionised_data <- sessionise(session_dataset, "timestamp", "uuid")
str(sessionised_data)
# 'data.frame': 63524 obs. of 5 variables:
# $ uuid : chr "0005839b3e8483d50870f61f50307fa7" "000b047bad36484451f12c114ab5eb28" "000b047bad36484451f12c114ab5eb28" "000b047bad36484451f12c114ab5eb28" ...
# $ timestamp : POSIXlt, format: "2014-01-14 12:47:59" "2014-01-07 14:25:11" "2014-01-09 12:47:17" ...
# $ url : chr "https://www.nasa.gov/history/apollo/images/footprint-logo.gif" "https://www.nasa.gov/ksc.html" "https://www.nasa.gov/biomed/threat/gif/beachmousefinsmall.gif" "https://www.nasa.gov/shuttle/resources/orbiters/atlantis.html" ...
# $ session_id: chr "9c77ea18bbef377253be1b22957071c1" "eda2ec544d96f0f1e3271902cbb693b7" "ee6d08bdaf1fb3c28edd0ac3290b82f5" "ee6d08bdaf1fb3c28edd0ac3290b82f5" ...
# $ time_delta: int NA NA NA 45 4 75 274 47 NA 28 ...This adds two new columns - a unique ID for each session, and (for each event) the time elapsed between that event and the next one in a session.
From this we can calculate a lot of commmon session-related metrics:
# Number of sessions per user
sess_count <- session_count(sessionised_data, "uuid")
str(sess_count)
# 'data.frame': 10000 obs. of 2 variables:
# $ user_id : chr "0005839b3e8483d50870f61f50307fa7" "000b047bad36484451f12c114ab5eb28" "000b2bc1a5438d8d54d4fbec139a2fd5" "001b6e80a14ba8d809c4ff18cdbade40" ...
# $ session_count: int 1 2 1 1 1 6 1 1 1 1 ...
# Length of each session
sess_length <- session_length(sessionised_data)
str(sess_length)
# 'data.frame': 20820 obs. of 2 variables:
# $ session_id : chr "0000664732878ba3409c138d4870a42d" "00029b1cd83040b8e14d7d65e057029e" "0002e5a2e75610bfb6c0598ea228a9d1" "00097364d131b6d6580d3c69a3e0a868" ...
# $ session_length: int 0 62 101 0 83 7 3113 0 4071 0 ...
# The 'bounce rate' (overall or per user!)
sess_bounce <- bounce_rate(sessionised_data)
str(sess_bounce)
# num 18.9
sess_bounce <- bounce_rate(sessionised_data, "uuid")
str(sess_bounce)
# 'data.frame': 10000 obs. of 2 variables:
# $ user_id : chr "0005839b3e8483d50870f61f50307fa7" "000b047bad36484451f12c114ab5eb28" "000b2bc1a5438d8d54d4fbec139a2fd5" "001b6e80a14ba8d809c4ff18cdbade40" ...
# $ bounce_rate: num 100 14.3 0 100 100 ...
# And many othersFor the current release version:
install.packages("reconstructr")For the development version:
library(devtools)
install_github("ironholds/reconstructr")