1. National Domestic Violence Hotline
Tisha Fnu
Yan Gao
Jina Kang
Rebecca Lawrence
Wade Treichler
INF 385T - Visualization
2. Initial Data Set
Individual chat sessions
Start (date and time)
Duration
Operator
Real Time Session ID
skill
campaign_source
campaign_search_keywords
Browser
Daily chat aggregates
Day (date)
Chat Requests
Operator Engaged Chats
Chats Started,Abandoned Chats
Avg. Wait Time
Avg. Wait Time to Abandoned
Max Wait Time
Max Wait Time to Abandoned
Avg. Chat Length
Avg. Wrap up Time
Avg. Handle Time
Avg. Contact Time
Chat Rate (%)
Abandon Rate (%)
Online Clicks,
Auto Engage Chats
Operating System
Country
City
World Region
Postal Code
Time Zone
Pre-Chat Survey
Exit Survey
3. Final Data Variables
Individual chat sessions
• Zip code
• Search terms
• Chat duration
Daily chat aggregates
• Wait time
• Abandonment rate
• Total chats
4. Questions
• How do wait times relate to abandonment rates? Can wait-time aggregations
be used to identify windows for likely abandonment?
• Does chat volume tend to be higher on particular days of the week or at
particular times of day? Does the abandonment rate?
• Do chat durations tend to be longer on particular days of the week or at
particular times of the day?
• Do peak-demand time periods correlate with particular search terms?
• How can these patterns help staff manage organizational resources to
increase service levels and lower abandonment rates?
5. Data Visualization
Dashboard
• Total daily chats
• Wait time vs. abandonment
• Chat durations
• Search terms
• Aggregate of total chats
• Interactive time slider
• Location of clients
6. Mappings
• Time variable is global for all dashboard elements; controlled by user via interactive slider
• Total daily chats variable is mapped onto a column chart; column height corresponds to total
number of chats by day
• Wait time vs. abandonment variables are mapped onto a bubble chart; bubble size corresponds to
wait time, vertical height corresponds to abandonment rate, and color indicates acceptability
• Chat duration variable is mapped onto a heat map; color intensity corresponds to chat duration
• Search terms variable is mapped onto a word cloud; word size corresponds to term frequency
• Aggregate chat total variable is displayed in a running chat counter