A/B test results in Tableau: Analyzing Wikipedia’s new search functionality

Hey there, thank you for signing up for my workshop! Here are some materials that will help you prepare.

To Do before the workshop:

1. Install Tableau Public from here
2. Download the zip folder with the dataset from this repository (available in .csv and .xlsx formats at the top of this page)
3. (Optional) Read A Practical Guide to A/B Testing by VWO
4. (Optional) Watch Tableau training videos, specifically:
   • Getting Started : Getting Started & The Tableau Interface
   • Calculations: Calculation Syntax, Intro to Table Calculations & Introduction to LOD expressions
   • Why is Tableau doing that? : Understanding Pill Types

Case Study Background

The case was inspired by Wikimedia Foundation Data Analyst Technical Assignment. The original dataset was transformed and adapted for the purposes of the workshop.

Today you will become part of Discovery department at Wikimedia Foundation for a day. Your team just made a big change to the Wikipedia Search experience and ran an A/B test to determine whether the new version of the search is better than the old one. It is your task to analyze the data and prepare a brief explaining the results of the test.

Discovery team relies on event logging (EL) to track a variety of performance and usage metrics to help the team make decisions. Specifically, Discovery is interested in:

• clickthrough rate: the proportion of search sessions where the user clicked on one of the results displayed
• zero results rate: the proportion of searches that yielded 0 results

and other metrics outside the scope of this task. Event Logging uses JavaScript to asynchronously send messages (events) to Discovery servers when the user has performed specific actions.

Data

The dataset comes from a tracking schema that Discovery uses for assessing user satisfaction. Desktop users are randomly sampled to be anonymously tracked by this schema which uses a "I'm alive" pinging system to estimate how long users stay on the pages they visit. The dataset contains just a little more than a week of EL data.

Column	Value	Description
uuid	string	Universally unique identifier (UUID) for backend event handling.
timestamp	integer	The date and time (UTC) of the event.
session_id	string	A unique ID identifying individual sessions.
group	string	A label ("a" or "b").
action	string	Identifies which action triggered the event. See below.
checkin	integer	How many seconds the page has been open for.
page_id	string	A unique identifier for correlating page visits and check-ins.
n_results	integer	Number of hits returned to the user. Only shown for searchResultPage events.
result_position	integer	The position of the link to the visited page on the search engine results page (SERP).

The following are possible values for the action field of an event:

• searchResultPage: when a new search is performed and the user is shown a search results page.
• visitPage: when the user clicks a link in the results.
• checkin: when the user has remained on the page for a pre-specified amount of time.

Example Session

uuid	timestamp	session_id	group	action	checkin	page_id	n_results	result_position
4f699f344515554a9371fe4ecb5b9ebc	20160305195246	001e61b5477f5efc	b	searchResultPage	NA	1b341d0ab80eb77e	7	NA
759d1dc9966353c2a36846a61125f286	20160305195302	001e61b5477f5efc	b	visitPage	NA	5a6a1f75124cbf03	NA	1
77efd5a00a5053c4a713fbe5a48dbac4	20160305195312	001e61b5477f5efc	b	checkin	10	5a6a1f75124cbf03	NA	1
42420284ad895ec4bcb1f000b949dd5e	20160305195322	001e61b5477f5efc	b	checkin	20	5a6a1f75124cbf03	NA	1
8ffd82c27a355a56882b5860993bd308	20160305195332	001e61b5477f5efc	b	checkin	30	5a6a1f75124cbf03	NA	1
2988d11968b25b29add3a851bec2fe02	20160305195342	001e61b5477f5efc	b	checkin	40	5a6a1f75124cbf03	NA	1

This user's search query returned 7 results, they clicked on the first result, and stayed on the page between 40 and 50 seconds. (The next check-in would have happened at 50s.)

View full recording of the workshop presented at DSGO Virtual conference in April 2021.