Using Data from Digital Traces
Bradley R. StaatsVisiting Associate Professor, The Wharton School
University of PennsylvaniaAssociate Professor, UNC Kenan-Flagler
Trace data: A sign or evidence of some past thing
http://hadoopilluminated.com/hadoop_illuminated/Big_Data.html
1. Introduction 2. Examples 3. Issues to Consider
1. Digital trace: Process data
1. Introduction 2. Examples 3. Issues to Consider
Additional Application
Capture (1 & 2)
Document Tagging
No
YesPolicy fail?
Application received &
scanned
Document Tagging
Application Capture (1 & 2)
Preliminary Information (1 & 2)
Materials received &
scanned
Automatic rejection letter
sent
Application Complete?
NoAutomatic incomplete
application letter
Yes
Credit Check (1 & 2)
Policy fail?
Automatic rejection letter
sent
Yes Automatic rejection letter
sent
NoAutomatic request for additional
materials
Application Complete?
No Automatic incomplete
application letter
Yes
Policy fail?Income Tax
(1 & 2)
NoPolicy fail?
Real Estate (1 & 2)
No
Yes Yes
Credit Approval
NoAutomatic
approval letter sent
Yes
Routed to credit expert
for negotiationMarginal
CustodianYes
No
CustodianYes
No
111 workers over 2 ½ years598,393 individual transactions
A perennial problem in industry has been that of sustaining human productivity over extended periods of time.
–Scott 1966: 4
Staats & Gino (2012). Specialization and variety in repetitive tasks: Management Science.
Specialization(Smith 1776; Taylor
1911; Skinner 1974; Boh et al. 2007; Schultz et al.
2003)
Variety(Hackman & Oldham 1976; Schilling et al.
2003; Narayanan et al. 2009)
Task Allocation Strategy?
Findings• During a day: Specialization > Varied
assignment• Across days: Varied assignment >
Specialization • Workers exhibit learning in setups
2. Digital trace: Click data
1. Introduction 2. Examples 3. Issues to Consider
2. Digital trace: Click data
1. Introduction 2. Examples 3. Issues to Consider
How does the team encourage (or discourage) individual knowledge sourcing behavior?• Knowledge repository data– Per-click data by person
• Software development project data– Project outcomes, characteristics– 487 projects
• HR system data– E.g., demographics
• Do temporal landmarks motivate aspirational behavior
Google searches for “diet” (average)
3. Digital trace: Search data
1. Introduction 2. Examples 3. Issues to Consider
Dai, Milkman & Riis (Forthcoming). The Fresh Start Effect. Management Science.
Mon Tue Wed Thu Fri Sat Sun505560657075
Day of the Week
-4 -3 -2 -1 0 1 2 3 450
55
60
65
70
75
Days Since the First Workday After a Federal Holiday
First workday after federal holidays
4. Digital trace: Tracking data
1. Introduction 2. Examples 3. Issues to Consider
Dai, Hengchen, Milkman, Katherine L., Hofmann, David A., & Staats, Bradley R. The Impact of Time at Work and Time off from Work on Rule Compliance: The Case of Hand Hygiene in Healthcare.
RFID tags monitoring hand washing compliance Personalized messagingData transfer to a central server 13,773,068 hand hygiene opportunities• Generated by 4,157 caregivers at 35 hospitals
from January 2010 to March 2013
4. Digital trace: Tracking data
1. Introduction 2. Examples 3. Issues to Consider
Dai, Hengchen, Milkman, Katherine L., Hofmann, David A., & Staats, Bradley R. The Impact of Time at Work and Time off from Work on Rule Compliance: The Case of Hand Hygiene in Healthcare.
How do job demands affect individual compliance over the course of a single shift?
• Access– Relationship building– Pipeline management
• Structure to the data• Mechanisms– Mixing other methods– Or maybe leveraging
more trace data…• Informed consent
Issues to Consider
1. Introduction 2. Examples 3. Issues to Consider
Myers, Chris, Staats, Bradley R., & Gino, Francesca. “My Bad”: The Impact of Internal Attribution and Ambiguity of Responsibility on Learning from Failure.