Counting on confusion Why no one knows how many people are visiting their websites
Counting on confusion
Why no one knows how manypeople are visiting their websites
What if everyone is wrong?
The mystery emerges
• Boston.com, BostonHerald.com complain of lowballing
The mystery emerges
• Boston.com, BostonHerald.com complain of lowballing
• Internal counts are three times higherthan Nielsen ratings
The mystery emerges
• Boston.com, BostonHerald.com complain of lowballing
• Internal counts are three times higherthan Nielsen ratings
• Whose count is accurate? Agencies’ or content-providers’?
Panel-based metrics
• Nielsen and comScore sample what sites audience is visiting
Panel-based metrics
• Nielsen and comScore sample what sites audience is visiting
• Afternoon traffic is undercounted, local sites are a challenge
Panel-based metrics
• Nielsen and comScore sample what sites audience is visiting
• Afternoon traffic is undercounted, local sites are a challenge
• Nielsen once gave washingtonpost.com 10m and comScore 17m
Server-side metrics
• An actual count of incoming traffic.What could go wrong?
Server-side metrics
• An actual count of incoming traffic.What could go wrong?
• Visitors are counted by device. Less accurateif they clear cookies.
Server-side metrics
• An actual count of incoming traffic.What could go wrong?
• Visitors are counted by device. Less accurateif they clear cookies.
• No way of separating out robotic hitsfrom search engines
What is a unique visitor?
• Most agree it is someone who visited once during time span
What is a unique visitor?
• Most agree it is someone who visited once during time span
• Unique visitors per month is most common Internet traffic measure
What is a unique visitor?
• Most agree it is someone who visited once during time span
• Unique visitors per month is most common Internet traffic measure
• Other measures: return visits, number of page views, time spent
Dearth of free data
• Compete.com is best known, samples a panel of 2 million people
Dearth of free data
• Compete.com is best known, samples a panel of 2 million people
• Quantcast gets good reviews from some, but code must be installed
Dearth of free data
• Compete.com is best known, samples a panel of 2 million people
• Quantcast gets good reviews from some, but code must be installed
• Alexa provides a lotof data, but results seem very odd
Was newspaper data better?
• No one knows how many people actually open up the paper
Was newspaper data better?
• No one knows how many people actually open up the paper
• Multipliers of “readers” versus “subscribers”is highly suspect
Was newspaper data better?
• No one knows how many people actually open up the paper
• Multipliers of “readers” versus “subscribers”is highly suspect
• Chicanery to make numbers look higheris not unusual
Credits
• Much of this presentation is based on “Confusion Online: Faulty Metrics and the Future of Digital Journalism,” by Lucas Graves, John Kelly and Marissa Cluck, Tow Center for Digital Journalism, Columbia University Graduate School of Journalism, September 2010.