As Nielsen pushes its “total audience” measurement system as the new currency for a hybrid TV and video advertising marketplace, some fundamental issues remain -- especially the fact that Nielsen will effectively be adding many more sources of video viewing to the measured universe, and consequently, will fragment the TV/video advertising marketplace beyond anything Madison Avenue has experienced to date.
Before that happens, a number of big video platforms still need to integrate Nielsen’s digital measurement software in order to become part of its universe.
“For Nielsen to produce true cross-platform data, their SDK [software development kit] must be installed in each publisher’s digital platform,” Alan Wurtzel, president of measurement and media development at NBCUniversal, explains about the chicken-and-egg issue impeding Nielsen’s total audience measurement effort.
“There have been issues getting that done,” Wurtzel says understatedly.
“Every client has differing viewing environments that they have to work through,” concedes Kelly Abcarian, senior vice president of product architecture for Nielsen’s Watch unit, which oversees media audience measurement, adding: “This is a heady task and we have had regular meetings with clients in regard to implementation and engineering of this technology in order to support the progress of a currency-quality metric.”
How many clients? “Dozens,” is all Abcarian will say.
She says the process is subjective to each organization and the process varies by platform.
“In some cases we have seen clients proceed through certification in a matter of days,” she notes, adding that Nielsen’s main focus has been to work with each organization to integrate its software as “smoothly” as possible.
That process is taking too long as far as some big customers are concerned, especially those on the “buy-side.” GroupM Managing Partner of Research and Marketplace Analysis Lyle Schwartz, for example, points out that Nielsen originally promised to release its Total Audience measurement system in time for 2016-17 upfront negotiations, which are expected to begin soon.
In fact, one big player -- Interpublic’s Mediabrands’ Magna Global unit -- has already announced a nine-figure “newfront” deal with Google’s YouTube, though terms like audience guarantees and measurement methods were not disclosed.
“It really doesn’t exist until the marketplace can analyze it and project it,” GroupM’s Schwartz says of Total Audience ratings becoming a marketplace currency. “You need months of data, you need time.”
Currency is exactly what Nielsen is hoping its Total Audience ratings will become, and it is its belief -- and that of many industry observers -- that Nielsen’s historic role as the official currency of the TV advertising marketplace will enable it to extend that position into the hybrid, and hyper-fragmented TV/video world.
It’s not alone in that pursuit, of course. comScore is also moving toward a more comprehensive measurement of total audiences, and its recent acquisition of and integration with Rentrak’s TV audience measurement data is expected to accelerate that. comScore has already made some key announcements about plans to provide even more integrated and holistic measurement of its panel’s complete media consumption -- TV, digital, you name it -- by measuring usage on every device used by those households.
But Nielsen’s big leverage continues to be the fact that it is the established baseline currency for the apex medium in the mix -- television.
Nielsen Total Measurement system relies on four discrete components:
1 - Measuring consumer media exposure across platform.
2 - Creating comparable measurement.
3 - Measuring advertising and content separately.
4 - Providing a ratings solution for video, audio and text.
Total Audience measurement will have two main outputs: “total ad ratings” (Nielsen’s current C3 and C7 TV ratings and digital ad ratings through 35 days) and “total content ratings” (TV program ratings and digital content ratings through 35 days).
“The traditional C3 or C7 doesn’t tell the complete story, which has been the network's’ argument for many years,” says Nielsen’s Abcarian.
C3 stands for average commercial minute audience plus three days of playback viewing. C7 includes up to seven days of playback. Most big agencies continue to negotiate deals based on C3, but at least one, GroupM, has publicly stated it is using C7 for its deals.
Nielsen national TV sample now has 40,000 households, representing about 100,000 people, using about 50,000 connected devices in their homes. Nielsen says its software will pick up data embedded on publishers platforms for phone, smartphone, PC and tablets.
What’s new and different? Nielsen’s preliminary Total Audience data not only shows all TV and digital viewing through 35 days, but can reveal unique episode-by-episode viewing on specific days.
Early results show that while live, video-on-demand and DVR viewing account for the most viewing of recent episodes aired, time-shifted viewing continues to grow on digital platforms with older episodes.
Nielsen offered up on analysis of an undisclosed TV program looking at three episodes of a series that ran in a given week -- one original and two repeats. The original episode of the series being analyzed generated 91% of its viewing from live and DVR playback, but an earlier repeat episode re-running that same week generated 52% of its audience from VOD or digital streaming.
The net finding, says Nielsen’s Abcarian, is that older episodes of TV shows are beginning to see a shift toward more digital consumption.
She says it is Nielsen’s ability to analyze these detailed behaviors that gives it a leg up vs. comScore.
“We have granularity on the episode level across all platforms,” she explains, adding, “I’m not certain of anyone else is tying it altogether on an episode basis.”
Meanwhile, Nielsen is also offering Total Ad Ratings as a means of measuring the effectiveness of advertising campaigns, which is the bottom line for brands and agencies buying TV and digital video.
Separating advertising from program content make sense, Abcarian explains, noting: “The relationship between an ad and a program is becoming less and less -- because of dynamic ad insertion and other [advertising efforts].”
One of the biggest issues Nielsen has had to confront in reporting the data is accounting for the different “languages” -- or ways of looking at -- audience exposure among its different customer types.
For example, Nielsen’s Abcarian says traditional TV executives speak about programming and commercials in “average minute viewing” timeframes, while digital video executives speak in “views” which have little/no relation to any time/duration.
“We spent a lot of time creating comparable metrics,” Abcarian says -- adding, however, that Nielsen does not plan to become the arbiter of which metric is appropriate. Over time, she says, Nielsen believes the industry will sort out a dominant marketplace metric. Meanwhile, it will focus on measuring all of it.When exactly will those measurements be available to the marketplace? Abcarian says Total Audience data currently is scheduled to be released for both TV content providers and media agencies sometime in the third quarter of this year, meaning it will effectively be available after 2016-17 upfront negotiations are completed. But there’s always next year.
Metric---schmetric! It's either viewing or it's not. If it's "viewing" for "linear TV" but only device usage for other venues then the advertisers and their time buyers have a big problem while the sellers will rejoice as they will get really big numbers.
We are having exactly the same issues deploying the SDK on publisher websites and apps. The purpose of this data is to establish 'traffic' which is then used with the mobile / smartphone / laptop / desktop panel(s) to convert that traffic to audience data.
Regarding the metric we have the same issue of digerati quoting 'views' (which is actually stream starts - yes, including auto-play starts - an important measure that only the panel can estimate), while TV people still use 'average minute audience'.
Our TV ratings system (OzTAM) has deployed an SDK for the top broadcasters and our major subscription TV service (i.e. not the entire universe but a whopping chunk of it). The SDK collects (from the browser) the duration of the play (N.B. this is not necessarily viewing but is a very close cohort for short-form video). All these minutes are aggregated and the divided by the duration of the content to create the Video Player Measurement (VPM) rating.
This is 'stand-alone' data (i.e. not de-duplicated from the broadcast audience) and is Total People only (based on the assumption that 'a person' initiated the video play and that only one person viewed that device). It is not perfect but it is a good start. Clearly duplication panels and data will be needed to generate de-duplciated audience by demographics.
As an example see ... http://www.oztam.com.au/documents/vpm/OzTAM_VPM_7Day_TopPrograms_Daily_20160509.pdf
This is for 7-day viewing but 28-day viewing is also available. Limited data is being released into the public domain but each participating broadcaster gets to see VPM on all their content. The Top 10 programmes are released and the Top 5 for each of the six key broadcasters.
In the PDF you can see that the highest VPM was for Seven's May 03 episode of 'Seven Year Switch' which amassed 3,465,000 minutes within 7 days. The episode was 49.9 minutes long which converts to a VPM of 69,000.
To give that some perspective the Average Minute Audience was 820,000 in our metro markets and 383,000 in our regional markets for a total 1,203,000. Of this 1,101,000 was watched Live (to the second) while 101,000 watched 'As Live' (i.e. in the Overnights). This doesn't include the 7-day PVR payback yet (out tomorrow).
John, how do you plan to determine who was "watching" per minute for all of these devices and, in the case of commercial minutes, whether the assumed or claimed program content "viewer" actually "saw" the average commercial?
Ed you only believe in samplings and panels of a subset of the total population. So its about 40,000 out of 320,000,000. Curation at its Best!
Good questions Ed (as always).
Let's say someone clicks on an online TV catch-up service and selects a one-hour programme. The clock starts running. Let's also say that they get bored after exactly 36 seconds and stop the stream (e.g. Stop button or click away). That person's viewing would count as one percent of a view (36 seconds in 3,600 seconds). If one-thousand people bailed out after 36 seconds that would equate to just one person's 'view'. This is analogous to how TV's Average Minute Audience is calculated.
The difference is that with TV the People Meter is a multi-person device for which we have a priori information on geography, age, gender etc. for each button on the meter. VPM uses a browser plug-in that doesn't collect audience profile information - it is universal and passive. There is an inherent assumption that there is just a single person 'viewing' that session - that is someone had to have clicked on the Play button, but we assume no-one is peering over their shoulder (even though these 'personal devices' are often shared). This is also analogous to the way the People Meter works - when the TV goes on or when there have been long periods without any change we ask for audience confirmation, but in the interim we assume no audience change unless someone pushes their individual button. So, in essence these assumptions are more likely to produce a 'low-ball' VPM estimate.
TBC in Part 2 due to the character count limit.
Regarding commercial minutes this is relatively easy to establish. What the viewer sees is actually non-linear. The programme content is served from the broadcaster's server which is identifiable by the URL (most programmes are segmented into a number of files but they share the base URL). The inserted ads however, are served by a third party ad-server which has a completely different URL. So while someone may have watched for (say) 60 minutes on their clock, around 48 minutes comes from the broadcaster’s server/URL and 12 minutes comes from various ad servers/URLs.
This system (VPM) is actually a programme content rating and not an ad rating. Broadcasters have traditionally sold based on average minute across Start/End time differentials. We all know that the ad breaks are lower rating than the programme minutes so what has traditionally been reported as a "programme rating" was a low-ball figure as it included the ad breaks.
Here in Australia, we have a system with Nielsen where advertisers can get their media agency to tag their (online) ad content and we are able to ping third-party sources (can't name them but we're talking the majority of Australians being signed-up) and get age, gender and geography back along with the usual tag data. This is entirely optional and is on an advertiser by advertiser basis.
As I said it's not a bad start and has been readily accepted by the market place readily. Of course we realise we still have a lot of work to do - especially on de-duplicating the TV rating with the online rating. We also need demographic profiles which will need an online panel to provide the profile of active online viewers which can then be applied to the VPM 'Total People' quantum.
And Leonard. I assume you come from the 'sampling doesn't work' subset of the population. I sincerely hope you never come down with any serious ailment. Because when you go to the doctor or the hospital they will take a blood test. Generally it is 5ml out of the 8 litres of blood in your body (around 1 in 1600). I'm yet to meet the person who says ... Doc, I don't believe in samples - drain me dry just to make sure.