tweet tracking app design document
TRANSCRIPT
Runninghead:TwitterTrackingandAnalysisWebAppDesignDocument 1
Twitter Tracking and Analysis Web App Design Document
Bessie Chu
University of Southern California
TwitterTrackingandAnalysisWebAppDesignDocument 2
Abstract
This is a design document to providing a product roadmap for web
application to track Twitter conversations during live events, such as a television
show or a sports game, and output parsable data for post-analysis. This
document serves to complete the requirements of IML 502: Techniques of
Information Visualization.
The web-based application would ideally be used by non-technical
personnel at an advertising agency, research group, academic institution, and
other organizations to track conversations around a certain topic, such as
individual users or a brand. The potential design and visualization functionality I
detail in this document works in conjunction with accepted design principles
drawing from theoretical frameworks and best practices. A basic rundown of
needed personnel, estimated timeline, and tools is also included with the
caveat that a specialized technical team may change these steps and
timelines depending on resources.
This document should accomplish the goal of working as a roadmap for
development in terms of design, functionality, and process to a functional alpha
version of this software for testing before a wider beta rollout to a limited number
of clients.
TwitterTrackingandAnalysisWebAppDesignDocument 3
Expanded Project Background:
I am working on long-term research project researching international
sports fandom and brands using Twitter conversations during the UEFA
Champion’s League and the World Cup as subjects of study as a part of the
Annenberg Innovation Lab. We are working in conjunction with IBM’s jStart
Team, an emerging technologies group in the company, and international
communications conglomerate HAVAS Sports and Entertainment. I have been
working with IBM to develop a platform for the type of research we’re trying to
accomplish, namely a 1) web-based application for collecting live tweets
during a soccer game to be used by non-technical personnel, in this case
account planners at advertising agencies such at HAVAS SE. 2) a platform to
directly plug into collected tweets for limited analysis using data visualization as
an exploratory tool.
They have currently developed a Java-based desktop app that is not
scalable for distributed use. We have proposed a browser-based web
application that they are currently beginning to build.
This design document moves beyond the more basic feature set I’ve
been working with IBM. What I’m attempting to accomplish here with this
design doc is an expansion of that idea into an idealized platform.
I choose this project as I eventually want to go into Product Management,
and this is a good exercise for me to get feedback on how to organize different
moving parts, how to communicate them visually, and how to integrate
TwitterTrackingandAnalysisWebAppDesignDocument 4
functionality with design. Part of the feature set of this idealized web
application includes plug and play data visualizations, similar to how Tableau
has pre-set options, with four specific presets in the application.
Objective
Build web-based solution to track and analyze tweets for a non-technical
audience.
Current Datasets and Collection Method:
Twitter conversations are tracked in real time using Twitter’s Public
Streaming API. The current collection methodology is a Python-based script run
through a shell script pre-set with access tokens and variable fields for each
data collection run. The script collects metadata about each Tweet, such as
User ID, handle name, text of Tweet, etc. and compiles them into dictionaries.
We then run another script to parse the dictionaries out into csv files.
TwitterTrackingandAnalysisWebAppDesignDocument 5
Current Process:
TwitterTrackingandAnalysisWebAppDesignDocument 6
This methodology is currently not scalable to non-technical staff,
especially given that we are tasked with having an application to be used by
staff scattering in various international markets with varying technical capability
and system configurations. Working with spreadsheets is not scalable due data
exceeding Excel’s capabilities. The data needs to be moved into a MongoDB
or PostGreSQL database with connectors and software in place to be accessed
by non-technical staff.
Data has already been collected, with nearly a million tweets on servers
from the UEFA Champion’s League Series with plans to collect data in several
key games about HAVAS SE client brands during the FIFA World Cup this
summer. This web app will allow us to scale this data collection to non-technical
staff worldwide in HAVAS SE’s offices and among non-technical USC
researchers.
Informed Design Principles: Theoretical Framework
In creating this web application, beautiful and functional design is
paramount. Working with an international audience of marketers and getting
their buy into this project will require an intuitive and aesthetically pleasing portal
to establish the credibility of our team. In this process, I must insist on using
design principles and acknowledge some perils in creating essentially preset
TwitterTrackingandAnalysisWebAppDesignDocument 7
data visualization outputs as well as the need for an iterative process rather than
a single platform release set in stone.
Design Guidelines
1. Simple Design Principles
The current set-up with IBM’s proprietary (hence screenshots not included)
platform is not friendly for non-technical users and is desktop based. Part of that
is because it uses more than a dozen colors on the interface that resembles
more of a Windows desktop application than the browser-based applications
that access a cloud stack that our target user-base would be more familiar with.
Hence, the web app I’m proposing uses design principles in color using what
William Lidwell, Krstina Holden, and Jill Butler (2003) call “using colors
conservatively” (Color section, para. 2). I’ve done so on the wireframes that
the site colors will evolve from. The eye can only process about five colors in
one glance (Color section, para. 2). With that in mind, the website should be
created using opposing colors on the color wheel and use pure hues to attract
attention to functional elements, such as buttons to save and move data (Color
section, para. 3 & 4).
I also took care to keep elements balanced on the wireframes and will
continue to do so in the design of the web application, particularly paying
attention to symmetry. This website will be used by a relatively discerning and
design savvy yet not technical crowd, having an application pleasing to the
eye paired with functionality will go a long way. Ahamed Altaboli and Yingzi Lin
TwitterTrackingandAnalysisWebAppDesignDocument 8
(2012) found during their experiments that at “high levels of symmetry, unity of
form… has significant effects on perceived visual aesthetics of website interface
design” (p. 732). The design should focus on coherent layout and symmetric
layout for user-friendly experience. Large conglomerates often forget these
principles that start-ups have embraced in the analytics space to their success.
I’ve kept this in mind as I’ve wireframed this app.
2. Acknowledging Perils
Alexander Galloway argues that, “Data have no necessary form” (p. 82)
meaning “any visualization of data requires a contingent leap from the mode of
the mathematical to the mode of the visual. This does not mean that
aestheticization cannot be achieved” (p. 82) thus “every interface must try to
overcome its own unworkability” (p. 85). In other words, data should be made
beautiful and understandable. This is the challenge of how to strip down the
interface to be usable by laypersons but not remove too many potential
features for research, being mindful of not to “sacrifice the aesthetic for the
algorithmic” nor “sacrifice the algorithmic in favor of the aesthetic” (p. 97).
Neither the art nor the science should be shortchanged.
In both interface design for the site and the data visualization outputs, I
chose the simple visual outputs to acknowledge these issues. This audience
should be able to understand planned outputs from the naked eye: namely a
network graph, bar charts, and spreadsheets. This could be a particular issue
with the network visualization outputs. Development will have to pay careful
TwitterTrackingandAnalysisWebAppDesignDocument 9
attention to the network mapping and making sure that connections can be
zoomed in and out of and panned without loosing the density provides deep
insides into what Julie Steele & Noah Illiinsky (2010) call “social intimacy” (p. 23)
in the connections among users. The outputs should allow us to see the social
landscape easily.
3. Iterative Design
One challenge and design with this idealized app is to make it a constant
work in progress. I concur with Jer Thorp (2003) describing taking “small
visualization steps” as “sketch points” that don’t initially consider aesthetics.
The way this design doc is initially designed, as well as the alpha and beta
versions could be taken as adhering to that process. Some companies want to
deploy everything done at once, but I believe in the iterative approach and
allowing the end users to weigh in their input and going back and making
incremental changes before a large release. A large release should not
constrain future improvements.
TwitterTrackingandAnalysisWebAppDesignDocument 10
Design Document
Key Functionality
• Login page
o Twitter API key configurations tied to logins
• Simple navigation page
• Tracking page connected to Twitter Public API Stream
o Field to copy and paste keywords and hashtags to track
o Field to copy and paste user IDs to track
• Analysis Dashboard
o Network analysis showing how users are interconnected with links
and nodes
o Summary Bar Chart analysis on which URLs or web links were
tweeted the most
o Summary Bar Chart showing which handles had the most retweets
o Tweet and User Search
! Look up individual users and tweets by metadata elements
! Look up tweets containing keyword, hashtag, or other search
value
TwitterTrackingandAnalysisWebAppDesignDocument 11
Wireframes
Site mockups created with Balsamiq
Core functionality outline:
Index Page:
• Login page that will load a post login page
• It is expected that users will be pre-generated accounts from company
without a registration step
TwitterTrackingandAnalysisWebAppDesignDocument 12
TwitterTrackingandAnalysisWebAppDesignDocument 13
Post Login Page:
• Simple navigation page
TwitterTrackingandAnalysisWebAppDesignDocument 14
Tracking Page
• Stripped down functionality as much as possible as a motivation
• Current IBM tool is extremely cluttered with too many options, which has
proved frustrating for even technical staff at lab
• Once tracking starts the “Tracking Output Page” will appear
• Users will be able to stop and add more keywords or user ids, which will
automatically saved in the fields
• When user clicks “Finished,” values will clear
TwitterTrackingandAnalysisWebAppDesignDocument 15
User Configuration Page:
• Simple configuration page to copy over Twitter credentials
• Field order and text mimics Twitter Developer Account layout
• This step has proven one of the most difficult in beta testing
• Duplicated link to documentation with the text “Help” since there will likely
be confusion here
TwitterTrackingandAnalysisWebAppDesignDocument 16
Tracking Output Page:
• Streaming page with just with text output from Twitter stream to watch
tracking
TwitterTrackingandAnalysisWebAppDesignDocument 17
Analysis Dashboard
• Search for file (files available to all members of team) and select method
of analysis
TwitterTrackingandAnalysisWebAppDesignDocument 18
Network Graph
• Example graph taken from Gephi
• Ideal scenario where users can zoom into nodes and click individual
nodes for user metadata
• See how users might be linked to certain influencers and social eco-
systems on Twitter
TwitterTrackingandAnalysisWebAppDesignDocument 19
Retweet Bar Chart
• Bar chart of most Retweeted users. Reasonable magnitude algorithms will
need to be built to cut-off too values too small to add to analysis
TwitterTrackingandAnalysisWebAppDesignDocument 20
URL Breakdown Bar Chart
• Bar chart of most Tweeted URLs. Certain television shows have Instagram
heavy audiences, for example, that could be interesting for advertisers
TwitterTrackingandAnalysisWebAppDesignDocument 21
Spreadsheet Browse and Search
• Search by individual user or terms in Tweet text to load into interface for
csv export
TwitterTrackingandAnalysisWebAppDesignDocument 22
Production Timeline and Personnel Needed
Personnel
• Backend Developer with knowledge of EC2 Amazon Cloud Servers
• Frontend Developer with knowledge of moving data from site inputs to
back-end smoothly
• Product Manager gathering requiresment and feedback
Development Timeline
Next Steps
After the completion of this course, I plan to deliver a stripped down
version of this design document for our purposes to IBM to build the tool that will
only include the login and collection functionality as an initial version.
The rest of the functionality roadmap will be cleaned up and expanded
upon accordingly with feedback to be loaded on a personal online portfolio of
my graduate work.
TwitterTrackingandAnalysisWebAppDesignDocument 23
UPDATE: This application was built and used in six international offices at HAVAS
(Brazil, France, Italy, Mexico, Russia, United Kingdom) and by three USC
researchers as part of a data collection of nearly 10 million tweets.
TwitterTrackingandAnalysisWebAppDesignDocument 24
References
Altaboi, A., & Lin, Y. (2012). Effects of unity of form and symmetry on visual aesthetics of
website interface design. Proceedings of the Human Factors and Ergonomics
Society Annual Meeting, 56(1), 728-732. doi:10.1177/1071181312561152
Galloway, A. R. (2013). The interface effect. Hoboken: Polity.
Lidwell, W., Holden, K., & Butler, J. (2003). Universal principles of design: 100 ways to
enhance usability, influence perception, increase appeal, make better design
decisions, and teach through design [Kindle Edition]. Gloucester, Mass: Rockport.
Steele, J., & Iliinsky, N. P. N. (2010). Beautiful visualization: Looking at data through the
eyes of experts. Sebastopol, CA: O'Reilly.
Thorp, J. (2013). Visualization as a process., 2014, from
http://blogs.hbr.org/2013/04/visualization-as-process/
Twitter Developer Documentation. (2014). Frequently asked questions., 2014, from
https://dev.twitter.com/docs/faq#6861