Inside Nielsen’s VideoCensus Methodology

Dave Osborn, SVP, Product Leadership, Nielsen Online

Nielsen’s VideoCensus provides timely statistics and insights into how consumers use video online.  This includes the size and demographic composition (age, gender, race, etc) of the viewing audience for each website, as well as important measurements like the total number of streams viewed and the time spent watching by the average viewer. All this information helps content providers and websites more effectively sell their assets and audience, while providing tools for advertisers trying to decide where to place an online video campaign.  In assembling this information, we use a number of different measurement technologies, each with its own advantages.

Why Nielsen uses samples to collect its research

• The only way to really report audience demographics (the age, gender, race and other characteristics of the person viewing video) is to actually measure what people – not computers – watch.  Because it is not possible to track the viewing of every user on every computer in the U.S., the best way to tell how many people are watching online video is to select a representative cross-section of the entire Internet population, monitor their viewing, and project the results to the population as a whole.  Just as a doctor only draws a small sample of blood to measure red and white blood cell counts, so too does Nielsen use samples (or panels) to measure Internet use.  Nielsen uses the same principles to measure TV ratings and consumer buying patterns.
• The key to accurate sample measurement is to create a panel in which every member of the population has an equal chance of being selected.  This ensures that the panel proportionately represents men and women, teens and adults, high and low income individuals, employees of large and small companies, and so on.  It’s also essential to represent both heavy and light users in proportion to the entire population.  If, for example, only heavy users were represented in our samples, our estimates would be too high and if only light users were selected our estimates would be too low.

How Nielsen’s VideoCensus panels are assembled

• Nielsen recruits its panelists through a number of methods, including email, online advertising, telephone calls and postal mailings.. Other Internet measurement companies rely exclusively on online recruitment, which skews the sample towards heavy users and typically overstates Internet activity.  Nielsen’s combination of recruitment tactics captures a broader spectrum of demographics and consumer behavior.
• The VideoCensus panel is composed of two separate samples:

— Nielsen recruits a very large sample through e-mail and online advertising.  The sample includes hundreds of thousands of households and allows for very granular measurement.  To make sure the information from this sample reflects the population as a whole, we adjust the data collected from this sample with a second, more representative Calibration Sample.

— The Calibration Sample is created by randomly identifying and actively recruiting panelists through their street addresses and phone numbers.  We are in direct contact with these panelists and pay them for their participation.  These are the same methodologies used to recruit TV ratings panelists. The “randomness” of this sample means it includes both heavy and light Internet users across all demographics, and is therefore more representative of the entire Internet universe. The combination of these two panels gives us the depth of a large sample balanced by the industry’s only truly random calibration sample.

How Nielsen collects Internet data from panels

• After panelists agree to participate in our panel, they install a Nielsen software “meter” on their computer, which enables Nielsen to measure their online and computer usage. When a panel member views a video, the meter communicates information about that activity to Nielsen. The panelist’s demographic information, the stream URL, and other site information are processed and reported within the VideoCensus system. (Note: Nielsen does not sell or publish user information.)

How Nielsen collect actual ‘Census’ viewing behavior

• Nielsen also directly measures the number of times a particular video is played. To improve the accuracy of this measurement, video networks or broadcast sites can embed a code or “tag” in their video players.  This process causes all viewers of content to send an anonymous ‘ping’ to Nielsen.  This so-called ‘census’ measurement makes it easy for Nielsen to identify and report the actual content or program being consumed, and allows us to report an actual count of times that content is played back.

Why Nielsen combines panel and website data

• Nielsen’s panel data provides the audience demographics of viewers, which is the industry’s ‘currency’ for media planning.  Census measurement provides the actual count of the times content is consumed and is typically the method used by online publishers and advertisers for selling and buying impressions.
• Nielsen combines the datasets and provides reporting that includes unique viewers, total streams, demographic composition, and time spent viewing.

Why Nielsen’s data differs from the clients’ internal server-based counts

• The most frequently debated number in online measurement is unique audience.  Website analytics systems (what we call ‘internal’ data) count cookies or Internet browsers, but not people.  These systems overstate audience for a number of reasons.
• If you visit a website from home and from work, generally that website counts you as two visitors.  If you regularly delete your cookies, each time you visit that website you’ll be counted as a new visitor.  Panel data collected by Nielsen, by contrast, measures actual people and projects their activity to a carefully enumerated Internet universe.

Why Nielsen’s numbers are usually lower than those from other measurement companies

• Because our randomly selected Calibration Panel includes both heavy and light users, our data is a more accurate reflection of the entire online universe.  Other measurement companies create their panels from people who answer online solicitation and who tend to be heavy users.  As a result, other measurement companies sometimes report numbers that are even higher than the internal server counts from the websites they are measuring.