Internet R/F and Cross-Site
Duplication Project
April 9, 2003
ARF 49th Annual Convention
Roger Baron; FCB/Chicago
Leslie Wood; LWR
Thank You’s
Manish Bhatia Nielsen Net/Ratings
Jason Bigler DoubleClick
Lynn Bolger comScore
Jim Daniels Church & Dwight
Doug Honnold comScore
Stephen Kim comScore
Allister Lam DoubleClick
Stacy Malone Universal McCann Interactive
Brian Monahan Universal McCann Interactive
Kelly Niehoff Mullen
Scott Penniston FCB/Southern California
Susan Russo Nielsen Net/Ratings
Marc Ryan Nielsen Net/Ratings
Jennifer Shunk R.J. Palmer
Denise Siedner DoubleClick
Dave Smith Mediasmith
Young-Bean Song AtlasDMT
Jim Vail R.J. Palmer
Randy Wootton AtlasDMT
Background
Internet R/Fs were presented at a March 2002 R/F committee
meeting
Dramatically different results from each supplier
Leslie Wood and Roger Baron proposed comparing campaign
reach/frequency and cross-site duplication estimates provided
by each supplier.
David Smith and Gabe Samuels created coalition of data
providers
Coalition investigated and developed a methodology
Ten advertisers have agreed to participate
Preliminary data from one-week of tracking will be presented
The Team
Researchers
Gabe Samuels, The ARF – Project Coordinator
David Smith, Mediasmith – Project leader
Leslie Wood, Leslie Wood Research – Research
Designer and Researcher
Roger Baron, SVP, Director of Media Research,
FCB/Chicago – Researcher
Team – continued
Data Suppliers
Server Centric Measures
Denise Siedner, Allister Lam; DoubleClick
Young-Bean Song; Atlas DMT
User Centric Measures
Marc Ryan, Susan Russo; Nielsen//NetRatings
Doug Honnold, Lynn Bolger; comScore
Key Objectives
Compare R/F and duplication measures from different
data suppliers
Server centric
User centric
Create standards for reporting duplication
Better understand the possibilities and obstacles to
developing a new methodology combining user and
server centric measures into a single R/F measure.
Make recommendations for next steps to modeling
R/F
Some definitions
Server Centric Measures (SCM)
Count of computers (cookies) that were served an
advertising message.
Includes every served impression - the basis for
media billing
Need to differentiate US versus international
exposure
No demographics
Participating suppliers
DoubleClick
Atlas/DMT
Definitions (cont.)
User Centric Measures
Website exposure from a panel of Internet users
who allow researchers to electronically monitor
their browsing behavior.
Statistically projectable to U.S. Internet users
Individual user login allows demographic detail
Syndicated service reports websites, not ads
Must be provided ad URL’s for this study
Participating suppliers
Nielsen//NetRatings
ComScore
Definitions (cont.)
Reach: Traditional media definition
“The number of different persons or homes exposed to a
specific media vehicle or schedule at least once. Usually
measured over a specified period of time (e.g. four weeks).
Also known as cume, cumulative, unduplicated or net
audience.”
- Advertising Media Planning, 6th ed - Sissors & Baron
Reach: Internet media definition
“The number of different cookies or United States persons
2+ at home or at work, who have been exposed one or
more times to a complete Web advertising message
over the campaign period. "Reach" is synonymous with
unique audience.”
- ARF Reach/Frequency Committee
Two-site Duplication
The number of persons or cookies exposed to both
sites as a percent of those who are exposed to
either site.
Not
Reached
% Reach
Site 1
Random duplication:
Assumes people who browse site 1 are as likely to browse
site 2 as anyone in the population.
Not
Reached
% Reach
Site 1 Duplication
% Reach Site 2
Actual duplication is always more
Think: people who browse IDG.NET also browse ZiffDavis
Actual Not
Duplication Reached
% Reach
Site 1 Duplication
Random always
overstates Reach
% Reach Site 2
Definitions (cont.)
Calculating percent duplication
Site A A&B Site B
Calculated as: A&B / (A+B-A&B)
Study Design
Report the same measures for the same advertising
campaigns from four data suppliers
Server centric suppliers give ad campaign URLs to
user centric researchers.
User centric researchers develop custom
methodology to report when these ads appear on
their sample’s browsers.
Required because their syndicated reports just track
websites, not the embedded ads.
Custom methodology is still being refined
Findings reflect both of the server centric suppliers.
Study Design continued
Advertisers across industries
Large ad campaigns
Variety of ad strategies
Frequency caps
Targeted
ROS
Matched variables from both server and user centric
suppliers
Reported Measures
Total campaign impressions
Total campaign reach
Advertiser: “How many people saw my ad?”
For each website in the campaign
Impressions
Total Reach
Exclusive reach
Duplication matrix of all sites taken two at a time.
Progress
Determine technical logistics
What address could everyone read
What address made sense across suppliers
Full agreement from suppliers to participate
Testing of IAB’s Ad URLs completed
Recruiting advertisers to allow server centric
suppliers to share data
NDA’s from team for advertisers
Several advertisers have signed agreements
Data collection just beginning
Issues along the way
Advertiser resistance and need for confidentiality
DoubleClick limited to “DART For Advertisers”
Time consuming (gratis) project
Large number of creative units
On-going need to notify UCM’s of creative changes
Complexity of Rich media creative
Handling of geographically focused and “tracking” ads
Inconsistent website name granularity
Custom definitions and report formats
Data / Findings
Data / Findings
Data / Findings
Data / Findings
Data / Findings
Data / Findings
Duplication between two websites is small (typically less than 0.25%)
but slightly greater than random
Data / Findings
Learnings
At this early date, there is similarity among the methods
Net reach as a percent of sum of site reaches
Average site percent exclusive reach
Actual campaign reach vs. random
Duplication between websites
Duplication between websites is greater than random, but only
slightly
As in traditional media, random overstates campaign reach
Determining User based ad campaign reach and frequency is labor
intensive and methodologically immature.
Next steps
Continue tracking pilot advertiser campaigns
Add additional advertisers
Report detailed findings at June ARF Internet R/F
Committee meeting